ArticlePDF Available

Cohesion and Performance in Groups: A Meta-Analytic Clarification of Construct Relations


Abstract and Figures

Previous meta-analytic examinations of group cohesion and performance have focused primarily on contextual factors. This study examined issues relevant to applied researchers by providing a more detailed analysis of the criterion domain. In addition, the authors reinvestigated the role of components of cohesion using more modern meta-analytic methods and in light of different types of performance criteria. The results of the authors' meta-analyses revealed stronger correlations between cohesion and performance when performance was defined as behavior (as opposed to outcome), when it was assessed with efficiency measures (as opposed to effectiveness measures), and as patterns of team workflow became more intensive. In addition, and in contrast to B. Mullen and C. Copper's (1994) meta-analysis, the 3 main components of cohesion were independently related to the various performance domains. Implications for organizations and future research on cohesion and performance are discussed.
Content may be subject to copyright.
Cohesion and Performance in Groups:
A Meta-Analytic Clarification of Construct Relations
Daniel J. Beal
Purdue University
Robin R. Cohen
Avon Products
Michael J. Burke and Christy L. McLendon
Tulane University
Previous meta-analytic examinations of group cohesion and performance have focused primarily on
contextual factors. This study examined issues relevant to applied researchers by providing a more
detailed analysis of the criterion domain. In addition, the authors reinvestigated the role of components
of cohesion using more modern meta-analytic methods and in light of different types of performance
criteria. The results of the authors’ meta-analyses revealed stronger correlations between cohesion and
performance when performance was defined as behavior (as opposed to outcome), when it was assessed
with efficiency measures (as opposed to effectiveness measures), and as patterns of team workflow
became more intensive. In addition, and in contrast to B. Mullen and C. Copper’s (1994) meta-analysis,
the 3 main components of cohesion were independently related to the various performance domains.
Implications for organizations and future research on cohesion and performance are discussed.
Throughout the history of organizational research, an important
goal has been to identify the factors and processes that give rise to
increased group performance. In the pursuit of this goal, research-
ers often have focused on the social and motivational forces that
exist between group members. The theoretical and intuitive hy-
pothesis has been that these forces create a bond, or cohesion,
among the members of the group, and that the stronger the bond,
the greater the productivity of the group. Presumably, when cohe-
sion is strong, the group is motivated to perform well and is better
able to coordinate activities for successful performance (Cart-
wright, 1968; Davis, 1969). Although most researchers have ac-
knowledged the plausibility of the relation between group cohe-
sion and group performance, empirical observations of the relation
have varied greatly, causing some authors to doubt the generaliz-
ability of the effect (Stogdill, 1972; Tziner, 1982) or to dismiss it
altogether (Steiner, 1972).
Contributing to the ambiguity of the cohesion–performance
relation are a wide variety of conceptualizations for both con-
structs. A prominent confusion, for example, concerns the appro-
priate level of analysis. Many researchers have measured group
cohesion as individual perceptions of the group and related them to
individual aspects of performance. It is not our position that such
relations are insubstantial; however, because the construct of co-
hesion refers to the resultant of forces acting on the group, it seems
appropriate to conceptualize the constructs of interest at the group
level. Indeed, Gully, Devine, and Whitney (1995) tackled this very
issue and found that relations between cohesion and performance
were stronger when both constructs were measured at the group
level. Moreover, many researchers have noted that discrepancies
between conceptual and operational levels of analysis can result in
ambiguous and inaccurate findings (Klein, Dansereau, & Hall,
1994; Ostroff, 1993; Scullen, 1997). Therefore, our meta-analyses
include only those studies that measured cohesion and perfor-
mance at the group level.
Because of the apparent ambiguity in the relation between group
cohesion and performance, several meta-analyses have attempted
to highlight situations in which the effect is stronger or weaker
(e.g., Carron, Colman, Wheeler, & Stevens, 2002; Evans & Dion,
1989; Gully et al., 1995; Mullen & Copper, 1994; Oliver, Harman,
Hoover, Hayes, & Pandhi, 1999). These studies have succeeded in
identifying several moderators of cohesion–performance relation-
ships, including group size, group reality, level of analysis, and
group interdependence.
One limitation of the previous cohesion–performance meta-
analyses is a lack of clarity concerning the conceptual nature of the
constructs. A primary purpose of the current meta-analysis was to
evaluate what we mean when we say cohesion, what we mean
when we say performance, and how these construct domains relate
to each other. The question of what is meant by cohesion has
indeed been addressed in one of the existing meta-analyses.
Daniel J. Beal, Military Family Research Institute, Purdue University;
Robin R. Cohen, Avon Products, New York; Michael J. Burke, A. B.
Freeman School of Business, Tulane University; Christy L. McLendon,
Department of Psychology, Tulane University.
This work was partially supported by a Robert E. Flowerree Summer
Research Fellowship. We thank Howard Weiss and Janice Kelly for their
comments on drafts of this article, and Jenn Kaufman, Beth Deitch, Cathy
Maraist, Kim O’Farrell, and Carmen Turillo for their assistance in acquir-
ing and categorizing studies. Portions of this article were presented at the
1st Annual Conference of the Society for Personality and Social Psychol-
ogy, Nashville, Tennessee, February 2000.
Correspondence concerning this article should be addressed to Daniel J.
Beal, Military Family Research Institute, Purdue University, 14 North 2nd
Street, Suite 300, Lafayette, Indiana 47901. E-mail:
Journal of Applied Psychology Copyright 2003 by the American Psychological Association, Inc.
2003, Vol. 88, No. 6, 989–1004 0021-9010/03/$12.00 DOI: 10.1037/0021-9010.88.6.989
Mullen and Copper (1994) examined the long-held notion that
cohesion is composed of interpersonal attraction, group pride, and
task commitment. They concluded that task commitment was
significantly related to performance and that interpersonal attrac-
tion and group pride were not independently related to perfor-
mance. For several reasons, however, we believe that these con-
clusions may have been unwarranted. Thus, one purpose of this
meta-analysis is to reexamine the role of components of cohesion.
In addition, we give consideration to an often neglected topic in
research on cohesion and performancethe criterion domain. We
have chosen to focus on these issues because a closer examination
of predictor and criterion construct issues is essential for gaining a
better understanding of cohesionperformance relationships. Fi-
nally, we address the particular patterns of workflow within teams
and how these patterns might affect cohesionperformance
Thus, the purposes of this study are to (a) conceptually recon-
sider the structure and content of criteria used within group cohe-
sion studies, (b) meta-analytically test hypothesized cohesionper-
formance relationships with respect to more refined criterion
categories, (c) constructively reexamine the independent contribu-
tions of interpersonal attraction, group pride, and task commitment
in relation to criteria employed within group cohesion studies, and
(d) examine the potential influence of workflow patterns on cohe-
sionperformance relations. The remainder of the introduction
unfolds as follows: First, we discuss issues concerning the struc-
ture and content of criteria in group cohesion studies. Second, we
discuss psychometric and statistical reasons for reconsidering
Mullen and Coppers (1994) conclusion regarding components of
cohesion. Third, we describe more fully the workflow construct
and explain why it might be relevant for research on cohesion and
performance. Finally, we present the specific study hypotheses
concerning the influence of each construct on existing cohesion
performance effects.
Cohesion and the Criterion Problem
One common feature of most of the moderators examined thus
far in the cohesionperformance literature is the focus on the
predictor side of the relation. This bias should come as no surprise
to researchers in applied psychology; admonishments for an over-
emphasis on the predictors of performance, as opposed to the
performance domain itself, have appeared at regular intervals
throughout the history of this discipline (Austin & Villanova,
1992; Flanagan, 1956; Smith, 1976). In essence, cohesion re-
searchers have relegated performance or criteria used within group
cohesion studies to nothing more than outcomes of group cohe-
sion. To some extent, this fractured view of the criterion domain is
understandable. That is, a particularly strong argument for the
benefits of group cohesion could be made if the criterion domain
were inclusive of all definitions and types of criteria. Unfortu-
nately, as we have already noted, this result has not been the case.
As such, we feel that the cohesionperformance literature has a
definite need for a critical examination of the criterion domain.
Toward this goal, our critique of the criterion domain in the
cohesion literature focuses on two main areas: whether the crite-
rion is conceived of as a behavior or an outcome and whether
output criterion measures are adjusted for inputs.
Behavior Versus Outcome
Recent treatments of the criterion domain in the applied psy-
chological literature have noted that traditional, global conceptu-
alizations of performance are fraught with difficulties (e.g., see
Campbell, McCloy, Oppler, & Sager, 1993). Appropriate specifi-
cation of criteria in applied studies rarely is given consideration
beyond tacit acknowledgment as the outcome variable. In the
cohesion literature, if multiple criterion measures are used, the
result typically is confusion that cohesion was related to one
measure and not to another. Conclusions drawn from such research
frequently cite the tenuous nature of the cohesionperformance
relation as the cause of the mixed results (e.g., Keyton & Spring-
ston, 1990), or the differential findings are overlooked (e.g., Co-
hen, Whitmyre, & Funk, 1960; Deep, Bass, & Vaughan, 1967). For
these reasons, we felt that using a more fine-grained approach for
the criterion domain might shed some needed light on how vari-
ables such as cohesion vary in their relation to different types of
In particular, Campbell and his colleagues (Campbell, 1990;
Campbell et al., 1993) argued for a distinction between perfor-
mance as behavior and performance as outcome. Put simply,
performance is in the doing, not in the result of what has been
done. The latter view of performance as outcome is fairly common
in many areas of applied psychology, including the literature on
cohesion and performance. As Campbell and others have pointed
out, this latter view of performance does not take into consider-
ation the many potential impediments to performance that are
outside the control of the individual or group of individuals. For
example, group sales outcomes might depend heavily on location,
time of year, and economic conditionsnone of which are indi-
cators of a groups ability to perform. Because of the possibility of
such impeding factors, we hypothesized that cohesion would have
a stronger relation to performance behaviors than performance
Effectiveness Versus Efficiency
In addition to differentiating between behaviors and outcomes,
we also wanted to examine how the consideration of inputs af-
fected cohesionperformance relationships. When comparing
groups, efficiency measures, which adjust for group inputs, often
are more informative of performance than effectiveness measures,
which only permit a comparison of group outputs (e.g., see
Borucki & Burke, 1999; Wilderom, Glunk, & Maslowski, 2000).
For example, if a retail organization with stores of many different
sizes wanted to evaluate the performance of each store, overall
effectiveness measures (e.g., gross sales) would not provide an
accurate assessment of performance. A larger store most likely
would rate higher simply because of its greater capacity to pro-
duce. A smaller store, despite extremely efficient use of its re-
sources, might never be able to reach the overall level of perfor-
mance of the larger stores. Efficiency measures, which take inputs
as well as outputs into account, might better reflect the true nature
of store performance.
Aside from the influence of measurement issues of effectiveness
and efficiency, there are theoretical reasons to believe that cohe-
sion might bear a stronger relation to efficiency measures than to
effectiveness measures. A variety of group researchers have pos-
ited that cohesion is an important variable linking group processes
and group outcomes. Although the temporal placement of cohesion
in this causal process is uncertain, researchers have found cohesive
groups to have increased efficiency of language behavior (Mick-
elson & Campbell, 1975), greater team mental model convergence
(Mathieu, Heffner, Goodwin, Salas, & Cannon-Bowers, 2000),
and greater use of transactive memory systems (Hollingshead,
1998, 2000; Wegner, Erber, & Raymond, 1991). In sum, cohesive
groups should be able to use their groups resources more effi-
ciently because they know the members of the group better and are
motivated to complete the task successfully. Effectiveness mea-
sures may be somewhat insensitive to this aspect of performance,
but efficiency measures are designed to hone in on these sorts of
processes. As such, we hypothesized that relations between cohe-
sion and efficiency measures would be greater than relations
between cohesion and effectiveness measures.
The particular reasons for expecting this pattern of cohesion
performance relations highlights another major purpose of this
meta-analytic integration. Our arguments supporting moderating
roles for behavior versus outcome and efficiency versus effective-
ness are based on the assumption that interpersonal attraction, task
commitment, and group pride are all important aspects of group
cohesion. This view is at odds with a previous meta-analysis of the
components of cohesion (Mullen & Copper, 1994). These au-
thors examined several components of cohesion and concluded
that task commitment (as opposed to interpersonal attraction
and group pride) was the only component that independently
contributed to the cohesionperformance relation. This finding
was surprising, especially considering that most of the research
on cohesion and performance has operationalized cohesion
almost completely in terms of interpersonal attraction (Lott &
Lott, 1965). After briefly reviewing some previous perspectives
on the dimensionality of cohesiveness, we present three argu-
ments suggesting that Mullen and Coppers (1994) conclusions
may have been premature.
Components of Cohesion and the Need to Reconsider
Mullen and Coppers (1994) Conclusions
The recognition of cohesion as a multidimensional construct
dates back to the seminal work of Festinger (1950). He discussed
cohesion as a culmination of factors, such as attraction to the
members of a group, the activities of a group, and the prestige of
the group. Despite this early emphasis on acknowledging the
components of cohesion, researchers often have measured only
one aspect of the construct (e.g., Seashore, 1954) or have used an
omnibus measure that cannot determine the independent contribu-
tions of each component (e.g. Gowda, 1988). This fact is not to
say, however, that the importance of components of cohesion has
been completely ignored. In response to several articles lamenting
the unidimensional approach to measuring cohesion (Carron,
1982; Mudrack, 1989; Tziner, 1982), numerous researchers have
explored relations between its components and performance (Car-
ron, Widmeyer, & Brawley, 1985; Zacarro & Lowe, 1986; Zacarro
& McCoy, 1988).
Considering the original emphasis of components in the cohe-
sion construct and the continued emphasis on this approach (Car-
less & De Paola, 2000; Cota, Evans, Dion, Kilik, & Longman,
1995), Mullen and Coppers (1994, p. 224) conclusion, that
efforts to enhance group performance by fostering interpersonal
attraction or pumping up group pride are not likely to be effec-
tive, was somewhat surprising. Recent advances in meta-analytic
methods, however, suggest several reasons why this conclusion
needs to be reconsidered. Below, we discuss how issues related to
levels of analysis, stochastically dependent effects, and the use of
regression weights for determining the relative contributions of
components of cohesion each contribute to uncertainty regarding
Mullen and Coppers conclusion.
Mixed Levels of Analysis
One interpretational difficulty concerning the results of Mullen
and Coppers (1994) meta-analysis is the inclusion of studies in the
same distribution of effect sizes that measured the group cohesion
and performance variables at an individual level, as well as studies
that measured them at a group level. Gully et al. (1995) docu-
mented how this practice can lead to an underestimation of the
effect because the cohesionperformance relation is weaker at the
individual level. Because individual-level studies were included in
the component of cohesion moderator tests, the results are ambig-
uous to some extent.
The problem of mixing levels of analysis is compounded if the
effects are sample-size weighted, as is customary for distributional
meta-analyses (Hunter & Schmidt, 1990). The primary issue in
such meta-analyses is that effects measured at the individual level
necessarily will be weighted more heavily than effects measured at
a group level of analysis. For example, let us say that a study
assessing the cohesionperformance relation was conducted, and
measurements were taken both at the individual and group level of
analysis. This hypothetical study used 25 groups with 5 people in
each group. If both the individual-level and group-level correla-
tions were later included in a meta-analysis, the effect obtained at
the individual level of analysis, with an N 125, would be
weighted five times greater than the effect obtained at the group
level of analysis, with an N 25. Thus, not only do individual-
level assessments of group cohesion and performance suffer from
conceptual ambiguity, the ambiguities are amplified when correct-
ing for sample size.
Stochastically Dependent Effects
Beyond the problems of combining different levels of analysis,
there are methodological and statistical issues that add uncertainty
to Mullen and Coppers (1994) analysis of components of cohe-
sion. Advances in the understanding of the methods of meta-
analysis have revealed that the use of stochastically dependent
effect sizes can lead to large errors in parameter estimation (Gleser
& Olkin, 1994). Stochastically dependent describes a situation in
meta-analysis in which several estimates of an effect from the
same sample are treated as separate, independent estimates. This
inclusion of nonindependent effects often occurs when there are
too few independent effects to conduct analyses with adequate
power. For example, in Mullen and Coppers meta-analysis of
components of cohesion, reasonable estimates of cohesionperfor-
mance relationships for each component of cohesion would have
been difficult given the small number of independent effects for
the respective components. Therefore, Mullen and Cooper took
advantage of the fact that several studies had multiple estimates of
the cohesionperformance relation, thus rendering an overall more
stable effect size estimate. This method of analysis, however,
unduly weighted the importance of studies with multiple estimates.
Multiple measurements of the same cohesionperformance rela-
tion may make a more reliable estimate, but this stability is
artificial because it does not take into account the nonindepen-
dence of the data. We sought to include only stochastically inde-
pendent effects within each of our effect size distributions in order
to obtain a more accurate estimate of relations between compo-
nents of cohesion and performance.
Inappropriate Regression Weighting
A final issue with Mullen and Coppers (1994) analysis con-
cerns the use of regression weights to examine the independent
contribution of each component of cohesion in predicting perfor-
mance. These regression weights were based on the relative pro-
portion of items measuring a particular component of cohesion in
a questionnaire.
Unfortunately, the number or proportion of items
measuring each component in a questionnaire cannot provide
information concerning the independence of each components
relation to performance. For example, a hypothetical cohesion
performance study examines 10 groups, each having rated cohe-
siveness with a four-item scale. Three of the items assess task
commitment, and one item assesses interpersonal attraction. In-
cluded also is a single-item criterion measure of group perfor-
mance. In this hypothetical study, the correlation between the
interpersonal attraction item and performance was .80, and the
correlation between the average task commitment score (i.e., the
average of all three items) and performance was .20. The correla-
tion between the overall average cohesion score (i.e., all four
cohesion items) and performance was .35an effect that is larger
than most in the cohesionperformance literature. Clearly, the
effect size of the overall scale is due to the one interpersonal
attraction item. In Mullen and Coppers analysis, however, this
effect would have been attributed mostly to task commitment
because the overall scale has three times as many task commitment
items. This possibility clouds Mullen and Coppers interpretation
that the task commitment component of cohesion is the only
independent predictor of performance.
In the current study, we circumvented this problem by including
only independent estimates of each components effect in any
particular meta-analytic distribution of effects. That is, if an effect
size estimate included items assessing more than one component
simultaneously, we did not include it in the analysis. This strategy
may not have been possible at the time of Mullen and Coppers
(1994) meta-analysis because there were not sufficient numbers of
studies to examine the effect size estimates separately. However,
the substantial increase in the recent cohesionperformance liter-
ature since Mullen and Coppers work afforded us the luxury of
such an analysis.
Patterns of Team Workflow
Throughout the history of organizational research on groups,
one sentiment has been pervasive: task type matters. Whether
researchers are concerned with leadership style (Weed, Mitchell,
& Moffitt, 1976), group member status (Kirchler & Davis, 1986),
group structure (Stewart & Barrick, 2000), or group cooperation
(Kabanoff & OBrien, 1979), task type has attained an important
role. As a result, many taxonomies of task type have been pro-
posed, emphasizing a variety of specific characteristics (e.g.,
Hackman, 1968; Hackman & Morris, 1975; McGrath, 1984;
Steiner, 1972). Although these taxonomies have been useful for
many areas of group research, few have exhibited any relevance
for group cohesion. It appears, however, that the interdependence
of the task may be important. In particular, Gully et al. (1995)
found task interdependence to moderate cohesionperformance
relations at the group level.
In the current meta-analysis, we followed up on the finding by
Gully et al. (1995) by examining how specific aspects of task
interdependence might interact with group cohesion and perfor-
mance. Specifically, we examined how the pattern of a teams
workflow can enhance the beneficial aspects of group cohesion.
Thompson (1967) initially discussed the notion of internal inter-
dependence at a more macro-organizational level. In this dis-
cussion, he defined four forms of interdependence describing
how different branches of an organization exchange informa-
tion and work. Tesluk, Mathieu, Zaccaro, and Marks (1997)
provided a similar taxonomy that described how work flows
between members of a team. Because our interest was at the
team or group level, we adopted the specifics of Tesluk et al.s
Conceptually, differences in the exchange of work between
members of a team can vary in several ways, including the direc-
tion of workflow and the number of exchanges. For example,
pooled workflow involves tasks that aggregate individual perfor-
mances to the group level. No interactions or exchanges between
group members are required in this pattern of teamwork. Work
does not flow through multiple members of the group, and per-
formance simply is the sum or some other aggregation of the group
membersperformances. Sequential workflow describes tasks that
move from one member of the team to another but not in a
back-and-forth manner. Group performance is not simply the pool-
ing of each members performance but is a function of how the
work progresses through each member of the group. For example,
in an assembly line, each member of a group would be responsible
for a particular portion of the final product. Line assembly, how-
ever, implies that after one portion of the product is assembled, it
is passed along to the next person until it reaches the end of the
line. After the last person has completed his or her part, the product
is complete. Thus, work flows sequentially from the first team
member to the last team member.
The final two patterns of team workflow involve considerably
more workflow between team members. Reciprocal workflow is
similar to sequential in that work flows only from one member to
another, but the flow is now bidirectional; team members can
exchange work with one another multiple times. The team perfor-
mance, however, is accomplished when the last person in the
group has completed his or her performance. Finally, intensive
For a few studies, weights were assigned on the basis of the extent to
which a particular component was manipulated in experimental conditions.
Our argument, however, still holds for this procedure. That is, a greater
emphasis on one component in a manipulation of cohesion does not
necessarily mean it held the responsibility for the relation between overall
cohesion and performance.
patterns of workflow occur when the work has the opportunity to
flow between all members of the group, and the entire group must
collaborate to accomplish the task.
As can be seen, both the directional changes and the amount of
workflow between team members increases as the patterns
progress from pooled to sequential to reciprocal to intensive. To
the extent that group members exchange greater amounts of work
between members, group processes such as cohesion should gain
an important function in contributing to performance. In discussing
this notion, Tesluk et al. (1997) recommended a variety of human
resource programs to improve productivity for each pattern of
team workflow. As the amount of between-member workflow
increased, these authors recommended a greater emphasis on team-
level (as opposed to individual-level) programs. Consistent with
these observations, we expected cohesion to bear stronger relations
to performance as the level of team workflow increased from
pooled through intensive. That is, factors such as attraction to
group members, a shared commitment to the task, and a sense of
pride in belonging to the group should have a greater impact on
performance as the workflow between members increases in each
pattern of teamwork.
Our goals for the current meta-analytic integration of cohesion
performance relations were (a) to gain a more complete under-
standing of cohesionperformance relations with respect to differ-
ent types of criteria employed in group cohesion studies, (b) to
constructively reexamine the independent contributions of inter-
personal attraction, group pride, and task commitment in relation
to performance, and (c) to identify what particular patterns of
workflow would have benefits or detriments for the link between
cohesion and performance. Our specific hypotheses regarding per-
formance criteria were that cohesionperformance relations would
be stronger when performance is conceptualized and measured as
a behavior as opposed to an outcome, and measures of perfor-
mance efficiency would have a stronger relation to cohesion than
measures of performance effectiveness. We further hypothesized
that all three components of cohesion would bear significant
independent relations to performance criteria. Finally, we hypoth-
esized that patterns of team workflow would moderate the cohe-
sionperformance relation such that greater amounts of workflow
would be associated with stronger correlations.
Literature Search
Several different approaches were used in locating relevant articles.
First, attempts were made to retrieve as many articles as possible from the
reference sections of Mullen and Copper (1994) and Gully et al. (1995).
Second, we conducted computer searches of the PsycINFO and Sociofile
databases using the search terms cohesion, cohesiveness, interpersonal
attraction, group attraction, task commitment, task attraction,orgroup
pride. This search then was combined with a search for productivity,
performance, effectiveness, or efficiency. In addition, reference sections of
the final articles were scanned for any citation that might contain an
estimate of the cohesionperformance relation. Finally, requests for un-
published studies were made of several experts in the field of group
research, as well as the electronic mailing list for the Society for Person-
ality and Social Psychology. This overall search returned several hundred
possible articles, which then were narrowed down to 145 after discarding
clearly irrelevant or unobtainable studies.
The articles then were examined individually for inclusion in the final
analyses. Fifty studies were removed because (a) only multivariate or
partialed effect size estimates were available, (b) the performance variable
was a self- or within-group rating (see Gully et al., 1995), or (c) the
operationalization of cohesion or performance did not conform to our
definitions (see below). This process left 64 separate articles with 71
independent estimates and 186 total estimates of cohesionperformance
relations. Of these 64 separate articles, 11 were unpublished articles or
dissertations. Publication dates ranged from 1951 to 2002, with a median
year of publication of 1989 (SD 14.03). Sample sizes for the studies
averaged 45.30 (SD 33.93), and group sizes averaged 6.22 (SD 3.66).
Articles used in the meta-analysis are listed along with their uncorrected
correlations in Table 1.
In comparison to the two largest of the previous broad meta-analyses,
Gully et al. (1995) obtained 51 independent effect size estimates and
Mullen and Copper (1994) obtained 52 independent effect size estimates.
Note, however, that both of these meta-analyses included group- and
individual-level effects. When considering only the group-level effects,
Gully et al. contained only 35 independent effects and Mullen and
Copper only 39 independent effects. Of these two meta-analyses, we
included 29 of Mullen and Coppers 39 effects (74%) and 25 of Gully
et al.s 35 effects (71%). There were several reasons why the overlap
was not 100%: Some effects were multivariate (i.e., included variables
other than cohesion in the same analysis), some were self-reports of
performance, one study was unobtainable, and the dubious quality of
one study removed it from consideration in our analysis. Given these
considerations, the overlap with previous meta-analytic samples seemed
high. In addition, the current meta-analysis contained approximately
85% more group-level studies than the largest previous effort, which is
a substantial increase.
Coding of Study Characteristics
Perhaps one of the largest sources of error in meta-analyses is the
multitude of judgment calls made at various stages of the research synthe-
sis process (Wanous, Sullivan, & Malinak, 1989). For example, Mullen and
Coppers (1994) definition and subsequent coding of group interaction led
them to conclude that the level of interactivity does not moderate the
cohesionperformance relation. Gully et al. (1995), however, defined in-
teraction differently (actually termed interdependence) and found that
highly interactive groups exhibited a stronger cohesionperformance rela-
tion at the group-level of analysis. It is likely that the use of a different
operational definition of group interaction is at least partly responsible for
the discrepant conclusions. An important early step, therefore, was to arrive
at thorough operational definitions for every relevant variable (Cooper,
1998). Provided here is a list of each variable along with how it was
operationally defined. In addition, we have included examples of each
variable in Appendix A.
Performance as Behavior
In this category we included effects whose criterion measure was an
evaluation of actions or behaviors relevant to the goals of the study as
indicated by the experimenter (i.e., in laboratory experiments) or the goals
of the organization (i.e., in field studies). Following Campbell et al. (1993),
this definition also included indices of unobservable cognitive behaviors
Table 1
Sample Sizes and Observed (Uncorrected) Correlations From Each Study and for Each Variable of Interest
Study N IA TC GP Beh. Outcm. Effct. Effic. P S R I
Arroyo, 1997 12 .349 .349 .349 .349
Bakeman & Helmreich, 1975 10 .430 .430 .430 .430
Barrick et al., 1998 51 .270
Bhatara, 1972
Study 1 40 .354 .354 .316 .392
Study 2 20 .478 .478 .478
Bird, 1977 8 .745 .745 .745
Blades, 1986 45 .110
Burchfield, 1997 72 .105 .105 .105 .105
Carpenter & Radhakrishnan, 2002 30 .174 .174 .174 .174
Carron & Ball, 1977 12 .031 .479 .116 .303 .303 .303
Cohen et al., 1960 16 .131 .131 .131 .131
Colarelli & Boos, 1992 86 .050 .050 .050 .050
Cotter, 1979 13 .048 .042 .052 .052 .052
Craig & Kelly, 1999 61 .095 .159 .032 .032
Darley et al., 1952 13 .331 .277 .307 .307 .307
Deep et al., 1967 9 .434 .368 .381 .355 .368
Duffy & Shaw, 2000 137 .200 .200
Eisenberg, 2001 48 .268 .268 .248
Elias et al., 1989 36 .417 .417 .417 .417
Fiedler, 1954 22 .335 .335 .150 .520 .335
Fodor & Smith, 1982 40 .077 .077 .077 .077
Fox, 1986 94 .080 .080
Gekoski, 1952 21 .100 .100 .100 .100 .100
George & Bettenhausen, 1990 33 .040 .040 .040 .040
Gonzalez et al. (in press) 71 .090 .260 .175 .175 .175
Goodacre, 1951 12 .725 .725 .725
Greene, 1989 54 .150 .150
Hemphill & Sechrest, 1952
85 .178 .151 .278 .162 .360
Hoegl & Gemuenden, 2001 145 .170 .215
Jaffe & Nebenzahl, 1990 20 .200 .389 .094 .094 .094
Jehn, 1994 88 .170 .170 .170 .170
Jehn & Shah, 1997 53 .555 .555 .460 .650 .555
Karau & Hart, 1998 30 .488 .488 .488 .488
Keller, 1986 30 .435 .510 .360
Keyton & Springston, 1990 35 .079 .112 .017 .017
Klein & Mulvey, 1995
Study 1 52 .230 .230 .230 .230
Study 2 89 .150 .150 .150 .150
Landers et al., 1982 10 .780 .610 .540 .640 .640 .640
Langfred, 1998 61 .410 .170 .290
Langfred, 2000
Sample 1 67 .280 .280 .280
Sample 2 61 .650 .650 .650
Lodahl & Porter, 1961 55 .190 .190 .190
Lorenz, 1985 21 .469 .469
Martens & Peterson, 1971 144 .064 .158 .130 .130 .130
Melnick & Chemers, 1974 21 .220 .130 .063 .063 .063
Miesing & Preble, 1985 6 .867 .867 .867
Mossholder & Bedeian, 1983 18 .120 .120
Mulvey & Klein, 1998
Study 1 59 .370 .370 .370 .370
Study 2 101 .350 .350 .350 .350
Neal, 1997 25 .520 .520 .520
Neubert, 1999 21 .490 .490
Norris & Niebuhr, 1980 18 .440 .440 .440 .440
Podsakoff et al., 1997
Study 1 40 .260 .260
Study 2 71 .025 .100 .025
Porter & Lilly, 1996 80 .190 .190 .190 .190
Seers et al., 1995 6 .700
Smith et al., 1994 53 .350 .350 .350 .350
Stinson & Hellebrandt, 1972
Sample 1 11 .110 .110 .110
Sample 2 14 .000 .000 .000
Tehan, 1983 16 .465 .465 .465
(i.e., “‘solutions,’‘statements,or answersproduced as a result of covert
cognitive behavior and totally under the control of the individual p. 40).
Performance as Outcome
Criteria were categorized as performance outcomes if they represented
the consequences or results of performance behaviors.
Measures of Effectiveness
Performance effectiveness was defined as an evaluation of the results of
performance with no consideration of the costs of achieving the results.
Measures of Efficiency
Performance efficiency was defined as the effectiveness of a group with
some consideration of the cost of achieving that level of effectiveness, that
is, a ratio or factoring in of inputs relative to outputs. If a measure took
inputs into account in any way, then it was considered an efficiency
measure. We interpreted inputs in a broad manner, including time, effort,
and other resources expended, as well as number of errors made and
relative size of the group (if size offered performance benefits). Because
time is always involved in performance, we considered time as an input if
it was explicitly mentioned as part of the task (e.g., participants had 15 min
to come up with solutions, workers were stopped after 12 min, and so
Team Workflow
Our definitions for teamwork patterns follow directly from Tesluk et al.
(1997, p. 201).
Pooled. Work and activities are performed separately by all team
members, and work does not flow between members of the team.
Sequential. Work and activities flow from one member to another in
the team but mostly in one direction.
Reciprocal. Work and activities flow between team members in a
back-and-forth manner, but only a single team member is worked with at
a given moment in time.
Intensive. Work and activities come into the team, and members must
collaborate as a team in order to accomplish the teams task.
Variables were considered to measure cohesion if they fell into one of
the following component categories:
Interpersonal attraction. A shared liking for or attachment to the
members of the group.
Task commitment. The extent to which the task allows the group to
attain important goals or the extent to which a shared commitment to the
groups task exists.
Group pride. The extent to which group members exhibit liking for the
status or the ideologies that the group supports or represents, or the shared
importance of being a member of the group.
Beyond the categorizations described above, there were several other
judgment calls that needed to be made. Frequently, there were multiple
estimates of the cohesionperformance relation, even within a particular
category (e.g., three reported correlations between interpersonal attraction
and effectiveness for one particular task). In general, one overall estimate
was obtained with a sample-weighted average. However, if the correlations
were computed at different time periods in the groups development, the
correlation between the latest occurring measurements of cohesion and
performance was assumed to be the best estimate. Our reasoning for using
the most recent cohesionperformance correlation concerns the nature of
the construct of cohesion. Cohesion is not something that occurs immedi-
ately on a groups formation. It develops after the group has had an
opportunity to work together or at least become acquainted with each other
(Gosenpud, 1989; Harrison, Price, & Bell, 1998; Matheson, Mathes, &
Murray, 1996). Therefore, we felt that the effects of cohesion on group
performance would most likely emerge later in the groups existence as
compared with earlier.
In addition, there was a question of whether a questionnaire mostly
measured one particular component over others. Because we wished to
examine the independent effects of different components of cohesion on
performance, some effects could not be coded for the particular type of
cohesion. For example, if a questionnaire mostly contained items measur-
ing interpersonal attraction but also had one or two items measuring task
commitment, the effect was not coded for component of cohesion. This
Group decision-making tasks represent a source of possible confusion
with our definitions. In several cases, the group decision-making process
was unobservable to the person who judged the performance (i.e., only the
end result was judged, not the process). Despite being unobservable we
coded these situations as outcomes. Our reasoning was that many observ-
able behaviors occurred in the context of the performance and the unob-
servable element had little to do with cognitive processes that could not be
measured. Therefore, this type of performance could not fall under the
unobservable cognitive behaviors portion of our performance behavior
definition. If, however, aspects of the decision-making process were in-
cluded as measures of performance, we coded these elements as behaviors.
Table 1 (continued)
Study N IA TC GP Beh. Outcm. Effct. Effic. P S R I
Terborg et al., 1976 42 .300 .300 .300 .300
Tesluk & Mathieu, 1999 88 .265 .229 .043
Tziner & Vardi, 1983 115 .320 .320
Wech et al., 1998 71 .200
Wekselberg et al., 1997
10 .476 .476 .476
Williams & Hacker, 1982 9 .605 .720 .770 .747 .747 .747
Wolfe & Box, 1988 36 .045 .045 .045 .045
Wong, 1992 40 .268 .268 .268 .268
Zaccaro & Lowe, 1986 54 .040 .439 .200 .200 .200
Zaccaro & McCoy, 1988 33 .161 .105 .133 .133 .133
Zaccaro et al., 1995 46 .301 .301 .301 .301
Note. IA interpersonal attraction; TC task commitment; GP group pride; Beh. performance as behavior; Outcm. performance as outcome;
Effct. effectiveness criteria; Effic. efficiency criteria; P pooled workflow; S sequential workflow; R reciprocal workflow; I intensive
Different effects within the study had different Ns. The overall mean N is provided.
ensured that the effects of the other components did not contaminate
assessments of a particular component of cohesion.
Data Analysis Procedure
Tests of the hypotheses were conducted using Raju, Burke, Normand,
and Langlois (RBNL, 1991; Finkelstein, Burke, & Raju, 1995) meta-
analytic procedures, with a random effects model (cf. Raju & Drasgow,
2003). The RBNL procedure uses sample statistics including available
information on sample-based artifacts (i.e., predictor and criterion reliabili-
ties) to estimate individually corrected effects with standard errors for these
corrected effects. Subsequently, this meta-analytic procedure computes
sample-size weighted estimates of the mean and variance of corrected
effects. In contrast, most other meta-analytic procedures (cf. Hunter &
Schmidt, 1990) rely on distributions of hypothetical artifact values (as
discussed in more detail in Paese & Switzer, 1988, and Raju, Pappas, &
Williams, 1989) for estimating the mean and variance of corrected effects.
It is noteworthy that the RBNL meta-analytic procedure permits the con-
struction of a confidence interval around the estimated mean corrected
effect (cf. Finkelstein et al., 1995). To use the RBNL procedure, all study
effect sizes were first converted to Pearson correlations. These correlations
were then corrected for sampling error, as well as unreliability, in the
predictor and criterion measures. In instances in which a study did not
report reliability information, the average reliability from available studies
was substituted for the missing values.
Our first step was to constructively replicate the meta-analysis by
Mullen and Copper (1994). This was done primarily to ensure that poten-
tially differing results concerning components of cohesion were not due
solely to differences in the particular studies included. To accomplish this,
we analyzed separately the set of studies that overlapped with Mullen and
Copper. After obtaining high levels of agreement, we proceeded to code
the remaining studies (see Appendix A for indices of interrater agreement).
The first author coded the variables for each study, and the fourth author
cross-coded all variables for each study. Where there were disagreements,
the two coders reached agreement through discussion. This double coding
procedure and the checks on interrater agreement for the coding of key
study characteristics are consistent with recent recommendations for cod-
ing studies in meta-analyses (see Burke & Landis, 2003). We note here that
the pattern of results obtained for the Mullen and Copper replication did
not differ substantially from the results of the final set of studies. There-
fore, the reported results reflect our complete set of studies.
Overview of Analyses
To examine our hypotheses, we focused on three main criteria:
mean corrected correlation (M
), estimated SE of M
, and confi
dence intervals around M
. We chose to focus on corrected coef
ficients because we were primarily interested in estimated relations
between well-developed, highly reliable measures of cohesion and
performance. Note that we interpreted our findings using confi-
dence intervals as opposed to credibility intervals. Confidence
intervals reflect the accuracy of the mean coefficient through the
use of the standard error of the corrected mean correlation (Finkel-
stein et al., 1995), whereas credibility intervals estimate the gen-
eralizability of the effect through the use of the standard deviation
of the observed or corrected correlations (Schmidt & Hunter, 1977;
Whitener, 1990).
We have not reported tests of generalizability for several rea-
sons. First, we were not concerned with the portability of the
effect. Previous meta-analyses have demonstrated significant het-
erogeneity in the cohesionperformance relation (Gully et al.,
1995; Mullen & Copper, 1994). As discussed earlier, several
moderating variables already have been identified, mostly con-
cerning the predictor side of the relation. Our goal was to examine
constructs that produce different mean correlations based on the-
oretical differences. Therefore, we examined corrected mean co-
efficients and the extent to which their confidence intervals over-
lap. A second reason why generalizability tests, such as the
credibility interval, were not used is because of their notoriously
inaccurate Type I error rates (Cornwell, 1993; Koslowsky & Sagie,
1993; Sackett, Harris, & Orr, 1986; Spector & Levine, 1987).
We first present results concerning different types of criteria
including behavior versus outcome and effectiveness versus effi-
ciency. Following this, we address the independent contributions
of interpersonal attraction, group pride, and task commitment in
predicting performance. We then examine, where possible, the
interactions between components of cohesion and the various
types of criteria. Finally, we address the moderating effects of
team workflow.
Different Types of Criteria
Table 2 presents summary information for each construct of
interest. Table 3 presents average reliabilities for each predictor
and criterion category where available. Below we consider the
results for each of these variables in more depth.
Behavior Versus Outcome
We considered performance behaviors to be more closely linked
to the process of cohesion than performance outcomes, which
often are determined by factors unrelated to the efforts of the
group. Therefore, we predicted that cohesionperformance rela-
tions, on average, would be stronger when performance was op-
erationalized as behavior than when performance was operation-
alized as an outcome. In support of this hypothesis, the mean
corrected correlation was greater for performance behaviors (M
.301) than for performance outcomes (M
.168). Note that the
confidence interval around the cohesionbehavior correlation does
not include the cohesionoutcome correlation and vice versa.
Indeed, these correlations were significantly different from each
other (Z 2.573, p .05). Also of note in this analysis is that the
confidence intervals of these correlations did not include zero.
Thus, cohesion is related to both conceptualizations of perfor-
mance but accounts for more variability in performance behavior.
Although our meta-analytic estimates have eliminated stochastically
dependent effects within a particular analysis, a small amount of sample
dependency exists when comparing meta-analytic correlations with each
other. Specifically, many studies included several effect size estimates that
could be included in more than one of our moderator categories (e.g., if a
study included one questionnaire that was interpersonal attraction and
another that was task commitment). Thus, when comparing these two
average correlations, the same effect size does not appear in both estimates,
but a small percentage of the same individuals may have contributed to
both estimates. To examine this issue, we reanalyzed our data after remov-
ing all studies with overlapping samples. Because the results of these
analyses did not differ substantially from the original estimates, and be-
cause there were no changes in the pattern or significance of our results, we
reported the original estimates.
Effectiveness Versus Efficiency
We also posited that measures of performance efficiency would
be particularly adept at capturing the process benefits of group
cohesion. In contrast, measures of performance effectiveness,
which only take outputs into account, should be less able to exhibit
the beneficial effects of a cohesive group. In line with this hypoth-
esis, efficiency measures (M
.310) possessed a stronger aver
age correlation than did effectiveness measures (M
These mean coefficients fell outside of the opposing correlations
confidence intervals and were significantly different from each
other (Z 2.787, p .05). Much like the results for behavior
versus outcome, note that both confidence intervals were above
zero; efficiency measures better reflected the benefits of cohesion,
but cohesive groups also were more effective.
Components of Cohesion
Most importantly for the analysis of components of cohesion, all
three mean correlations were significantly greater than zero. There
were differences in the magnitude between the mean effects for
each component of cohesion, but none of the mean corrected
effects were significantly different from each other (all Zs 1.30,
ps .18). The ascending order of effect sizes was interpersonal
attraction (M
.199), group pride (M
.261), and task com
mitment (M
.278). Thus, in contrast to the findings of Mullen
and Copper (1994), our analysis indicates that the three compo-
nents of cohesion each correlate meaningfully with performance
criteria. Also of interest is that group pride, despite being the most
frequently ignored component of cohesion and having only six
effect size estimates, had an average correlation that was signifi-
cantly greater than zero and as strong as the other components.
Interactions Between Components of Cohesion and Types
of Criteria
Although there were too few studies to break down each com-
ponent of cohesion with each category of criteria, several analyses
were possible. Interpersonal attraction had, by far, the largest
number of studies, and we were able to examine how this com-
ponent related to behavior versus outcome, as well as to effective-
ness versus efficiency. Task commitment, although having sub-
stantially fewer effects than interpersonal attraction, still allowed
for tests of behavior versus outcome and effectiveness versus
efficiency. The handful of group pride effects, however, all used
criteria of effectiveness (i.e., no measures of efficiency) and out-
come (i.e., no measures of performance behavior).
Interpersonal attraction displayed a pattern similar to the overall
findings for behavior versus outcome. Specifically, the interper-
sonal attractionoutcome relation (M
.139) was significantly
smaller than the interpersonal attractionbehavior relation (M
Table 2
Mean Uncorrected Correlations (r ), Mean Corrected Correlations (M
), Variances of
Standard Errors of the Mean of
ˆ (SE
), 95% Confidence Intervals (CI), Number of Effect Sizes
(K), and Number of Groups (N) for Criterion and Component Moderators
Moderator r M
95% CI KN
Behavior .267 .301 .008 .041 .220, .383 19 778
Outcome .147 .168 .037 .036 .096, .239 47 2,125
Effectiveness .155 .175 .045 .041 .095, .256 40 1,899
Efficiency .272 .310 .006 .031 .249, .370 31 1,337
Component of cohesion
Interpersonal attraction (IA) .171 .199 .052 .042 .117, .281 43 2,049
Behavior .279 .315 .028 .069 .179, .451 10 482
Outcome .124 .139 .053 .050 .041, .237 31 1,446
Effectiveness .132 .148 .074 .062 .026, .270 25 1,187
Efficiency .240 .284 .032 .055 .176, .393 19 792
Task commitment (TC) .246 .278 .000 .043 .194, .361 16 579
Behavior .281 .302 .026 .074 .156, .448 4 176
Outcome .242 .273 .000 .055 .166, .381 11 342
Effectiveness .205 .232 .002 .055 .124, .340 10 341
Efficiency .306 .343 .000 .068 .210, .475 6 238
Group pride .242 .261 .000 .065 .133, .389 6 209
Table 3
Sample-Weighted Average Reliabilities for Each Predictor and
Criterion Category (r
and r
Predictor r
Interpersonal attraction .845 24 1,422
Task commitment .866 5 245
Group pride ——
Criterion r
Performance behavior .874 16 719
Performance outcome .888 34 1,525
Effectiveness .911 27 1,283
Efficiency .894 26 1,218
Note. Dashes are indicative of the fact that no reliability coefficients were
available for studies examining group pride; therefore, the analysis used the
average reliability for all cohesion measures. K number of effects
reporting reliability; N total number of groups for each average reliabil-
.315; Z 2.173, p .05). Task commitment, in contrast, did not
exhibit a clear difference between behavior and outcome (behav-
ior, M
.302; outcome, M
.273; Z .313, p .754).
Interpersonal attraction displayed the same pattern for effective-
ness versus efficiency as observed in the overall analyses, although
the difference now reached only marginal levels of significance
(effectiveness, M
.148; efficiency, M
.284; Z 1.722, p
.085). Although the same pattern was noticeable for effectiveness
versus efficiency in the task commitment average correlations, the
difference did not attain conventional levels of significance (ef-
fectiveness, M
.232; efficiency, M
.343; Z 1.386, p
Team Workflow
The final hypothesis we examined held that cohesion would be
beneficial to groups whose patterns of team workflow required
greater workflow between members. Following the typology of
Tesluk et al. (1997), we compared four progressively increasing
patterns of team workflow: pooled, sequential, reciprocal, and
intensive. Because the underlying moderator construct is probably
best expressed as a continuum ranging from low levels of work-
flow to high levels of workflow, we placed the four categories on
a 4-point continuum and examined how well the continuous mod-
erator predicted the corrected correlations using weighted least
squares regression (see Steel & Kammeyer-Mueller, 2002, for
more details on the analysis of continuous moderators in meta-
The results of this analysis found that the teamwork
moderator accounted for a significant amount of variance in the
corrected correlations (R
.096), F(1, 52) 5.430, p .05. As
predicted, as team workflow increased, the cohesionperformance
relation became stronger.
Our hypotheses for this meta-analysis mainly were concerned
with three issues: the independent contribution of each component
of cohesion, an examination of the criterion domain, and the role
of team workflow. With respect to the components of cohesion,
interpersonal attraction, task commitment, and group pride all
displayed independent relations to group performance. As we
noted in the Results section, however, our assessment of group
prideperformance relations is somewhat limited by the small
number of effects available. Moreover, the six effects that mea-
sured group pride all came from a rather homogeneous set of
studies. All were studies of sports teams that used only effective-
ness measures of outcomes (usually a winloss ratio). Clearly,
more research is needed to determine the importance of this
component of cohesion.
Our analyses of the criterion domain revealed two noteworthy
findings. First, in line with Campbell et al. (1993; Campbell,
Gasser, & Oswald, 1996), performance behaviors exhibited stron-
ger relations with cohesion than did performance outcomes. In
addition, measures of efficiency reflected the beneficial effects of
group cohesion better than measures of effectiveness. Finally,
patterns of team workflow moderated the cohesionperformance
relation; tasks exhibiting greater amounts of workflow also held
stronger cohesionperformance relations.
Relating Components of Cohesion to Various
Performance Criteria
We attempted to assess the relations between each component of
cohesion and multiple criterion categories. Unfortunately, we
could not examine each criterion variable with each component of
cohesion because of a lack of studies in one or more of the
categories. Interpersonal attraction exhibited correlations that
largely were consistent with the overall findings. In contrast, task
commitment displayed weaker differences between behavior and
outcome, as well as effectiveness and efficiency. One potential
explanation for the component differences is that task commit-
ment, in contrast to interpersonal attraction, may not be as likely to
reflect the fluidity of group behavior as it is to reflect an overall
stronger motivation to perform well (Festinger, Schachter, & Back,
1950). As such, the benefits of task commitment apply despite the
particular choices of performance measurement.
It is clear, however, that these propositions are in need of more
focused research attention. To our knowledge, no researcher has
attempted a direct test of components of cohesion with any of the
criteria examined in this meta-analysis. Indeed, this dearth of
research contributed to our inability to examine the relations
between performance criteria and group pride, as well as the
relations between any component and patterns of teamwork.
When Are Cohesive Groups Advantageous?
A consideration of the relations between components of cohe-
sion and various criteria reveals several situations in which cohe-
sive groups are likely to perform better. Certainly, when efficiency
is an important goal in the organization (i.e., as opposed to situa-
tions in which successful completion of a task is the main require-
ment, e.g., winning a game, successful surgical performance, ob-
taining a high grade on a project, etc.), cohesive groups gain an
advantage. Of particular interest was the finding that the distinc-
tion between efficiency and effectiveness was not as strong when
cohesion was conceived of as task commitment. This finding is not
to say, however, that task commitment does not bear strong rela-
tions to measures of effectiveness or efficiency. In fact, our results
suggest that the reverse is true; task commitment held moderate
relations across all of the examined criterion domains. Whether
this finding holds for patterns of team workflow, however, must
await future studies that compare this moderator with the various
components of cohesion.
Cohesive groups also achieve performance benefits when group
performance is conceptualized as a behavior instead of an out-
come. We argued that the stronger relation for performance be-
havior was due to external impediments inherent in outcome
measures; however, we also point out that because performance
behaviors are causally antecedent to performance outcomes, they
may be more closely tied to cohesion and the group processes that
result from cohesion. Indeed, one would have difficulty imagining
any part of a cohesionoutcome relationship that is not mediated
Consistent with research examining continuous moderator tests (Steel
& Kammeyer-Mueller, 2002), we used inverse sampling error weighting
for this analysis. However, we relied on the RBNL estimates of sampling
error instead of more simplistic variance formulas that do not take artifact
information into account.
by performance behavior. This finding illustrates the necessity of
defining performance as behavior in order to appropriately identify
those constructs that are predictive. Campbell et al. (1996) la-
mented that applied psychology would be hindered to the extent
that we conceive of performance as an outcome, and this meta-
analysis provides justification for their disapprobation.
Perhaps the most interesting point concerning the utility of
group cohesion is that every way in which we examined compo-
nents of cohesion and domains of criteria resulted in a positive
mean correlation, and all of these mean correlations were signifi-
cantly different from zero. Certainly, previous meta-analyses on
the overall cohesionperformance effect left little doubt that co-
hesion benefits performance, but our analysis suggests that this
benefit cuts across many different conceptualizations of the cohe-
sion and performance constructs. In addition, although differences
in magnitude between the mean corrected correlations were often
not large (e.g., effect size difference of .13 between the respective
mean corrected correlations for behavior versus outcome criteria),
these small differences in correlations at the group level can have
substantial practical utility when considered within a decision
theoretic utility sense.
As a final note, we acknowledge the potential relation of cohe-
sion to another criterion domain, broadly termed contextual per-
formance. Although there is some debate concerning what actions
or behaviors should be considered indicators of contextual perfor-
mance, as well as the appropriate nomenclature for this construct
domain, generally contextual performance includes behaviors that
do not support the technical core itself so much as they support the
broader organizational, social, and psychological environment in
which the technical core must function (Borman & Motowidlo,
1993, p. 73). Because contextual performance often includes ac-
tions that are helpful to other members of a group, it seems likely
that cohesive groups would experience higher levels of contextual
performance (LePine, Hanson, Borman, & Motowidlo, 2000).
Indeed, we attempted to assess this relation, and from the few
studies that contained appropriate effects, we found a positive
correlation between cohesion and contextual performance. Be-
cause contextual performance usually occurs at the individual
level, however, most of the studies examining this relation did not
meet our inclusion criterion of group-level effects. Nevertheless,
we feel it worthwhile to mention our cursory examination as an
encouragement to other researchers pursuing this topic.
When Are Cohesive Groups Less Advantageous?
Despite the overall rosy picture presented above for the utility of
group cohesion, there definitely were circumstances in which
cohesive groups provided little help for performance. Obviously,
the flip side of the domains mentioned above require acknowledg-
ment: Compared with performance behaviors, performance out-
comes do not reflect the advantages of cohesion, and compared
with efficiency measures, effectiveness measures ironically are
less effective in capturing the benefits of being in a cohesive
group. Finally, as our analysis of team workflow demonstrated,
groups who engage in fewer exchanges of work (e.g., pooled
workflow) do not benefit from cohesion as much as those groups
whose workflow is intensive.
As we noted above, the difference between behavior versus
outcome and effectiveness versus efficiency held mostly for stud-
ies that conceived of cohesion as interpersonal attraction. Although
our meta-analysis was unable to identify the reasons why inter-
personal attraction held these stronger relations, there are specu-
lative reasons to expect this result. Efficiency often relies on the
communication and cooperation of group members. To the extent
that interpersonal attraction facilitates these group processes, it
will likely lead to more efficient group performance. If the quality
of group inputs is not taken into account, as is the case with
measures of effectiveness, then interpersonal attraction may be
somewhat insensitive as a predictor. Unfortunately, research di-
rectly examining these assertions is largely absent in the organi-
zational literature.
The fact that correlations involving pooled team workflow were
weak in comparison with more intensive patterns makes intuitive
sense when considering the types of tasks involved in these stud-
ies. If the members of a team focus on their own individual
performances, then many of the benefits of cohesion would have
no bearing on the teams performance. Interestingly, it may be the
case that task commitment would boost performance in tasks
requiring pooled teamwork whereas interpersonal attraction would
not. The benefits of task commitment revolve around a shared
motivation to do well on a task. Because this motivation does not
necessarily require members to work together, tasks requiring
pooled teamwork might still benefit from this shared commitment.
The same logic applies to group pride: If members of a team are
motivated to maintain the esteem of the group, then their motiva-
tion may carry them through individual tasks that are then pooled
to the team level. Unfortunately, there were not sufficient numbers
of studies in our sample to test these hypotheses, but certainly our
thoughts merit consideration in future studies of cohesion and
There are several possible limitations present in our meta-
analysis that warrant discussion. First, there are issues regarding
correlations or confounds amongst moderators in other meta-
analyses and within our own. A case could be made, for example,
that group size is theoretically related to one or more of our
included variables (e.g., team size occasionally is considered an
input for efficiency measures). If this were the case, it would be
unclear which variable was responsible for our observed results.
Obviously, we cannot include all previously examined moderators,
and to that extent, we must remain cautious with the interpretation
of our results. It is possible, for example, that our ideas and results
concerning particular patterns of workflow add little beyond Gully
et al.s (1995) notions of task interdependence. We would argue,
however, that identifying the manner in which work is exchanged
in a group provides a more precise understanding of the actual
process involved. Indeed, team workflow describes one of the
potential mechanisms that make groups interdependent. At a min-
imum, conceptualizing workflow as a continuous variable allows
for a more precise description of the actual process of teamwork
Because group size was fairly easy to code, we did examine this
possibility. We recomputed our average effects for each variable control-
ling for group size. No meta-analytic effect was significantly different from
its original estimate. Furthermore, the deviation of greatest magnitude was
rather small (.019).
than do categories of task interdependence, and results of our
moderator analyses are consistent with expectations for workflow
as a continuous variable.
Aside from overlap with moderators of other meta-analyses, it is
possible that variables within our own meta-analyses were con-
founded. For example, if a significant proportion of the same set of
studies included both outcome criteria and measures of effective-
ness, then it is difficult to say which variable is responsible for the
average effect. Indeed, of the six possible relations between our
four variables, which, for the purpose of breaking down distribu-
tions of correlations, were treated as moderators (excluding group
pride because of the small number of effects), two correlations
were significant. Behavioroutcome was related to effectiveness
efficiency (r .456) such that studies with behavioral measures of
performance tended also to use efficiency criteria (or, conversely,
that studies with measures of outcomes tended also to use effec-
tiveness criteria; see Appendix B, however, for examples of stud-
ies that were coded in the four possible combinations of catego-
ries). In addition, pattern of team workflow was related to
behavioroutcome (r .389) such that studies with more work-
flow between members tended to use outcome measures (or,
conversely, studies with less workflow between members tended
to use behavioral measures).
Appropriate evaluation of these relations, however, is a difficult
task. Using a weighted least squares regression approach (Steel &
Kammeyer-Mueller, 2002), as we did in some other analyses,
proved difficult to conduct for reasons of study exclusion. In
particular, many studies had to be eliminated because they could
not be coded for all relevant variables. Further still, some studies,
despite having an effect size estimate for all of the relevant
categories, did not have the same effect size estimate. Thus, the
remaining set of studies was small in comparison to the total
sample of studies, and more importantly, the criteria for elimina-
tion was potentially biased. Our solution, therefore, is to encourage
future researchers to tackle the question of whether stronger cohe-
sionperformance relations surface as a result of assessing behav-
iors (as opposed to outcomes) by using measures of efficiency (as
opposed to effectiveness) or if these relations benefit from both
factors in an additive fashion or reflect an unexamined causal
ordering of variables (e.g., see Gonzalez, Burke, Santuzzi, &
Bradley, in press; Kirkman & Rosen, 1999).
In summary, we tested several hypotheses in the cohesionper-
formance literature using meta-analytic techniques. Our results
provide compelling evidence for expected differential relations
between group cohesion and different types of criteria. Similarly,
our sample of studies found that groups who take the most advan-
tage of cohesion typically engage in intensive patterns of work-
flow. Finally, these results suggest that all three of Festingers
(1950) original components of cohesioninterpersonal attraction,
task commitment, and group prideeach bear significant inde-
pendent relations to performance across many criterion categories.
These findings not only enhance our understanding of the con-
struct domains of group cohesion and group performance, but also
add to our knowledge of the magnitudes of effects between con-
structs in the respective domains. A final benefit of our meta-
analyses was in identifying important gaps in our understanding of
relations between group cohesion and group performance, hope-
fully providing guidance for future primary studies.
References marked with an asterisk indicate studies included in the
*Arroyo, D. G. (1997). Cohesion, performance, and satisfaction in the
co-acting sport of collegiate wrestling. Unpublished doctoral disserta-
tion, University of Connecticut, Storrs.
Austin, J. T., & Villanova, P. (1992). The criterion problem: 19171992.
Journal of Applied Psychology, 77, 836874.
*Bakeman, R., & Helmreich, R. (1975). Cohesiveness and performance:
Covariation and causality in an undersea environment. Journal of Ex-
perimental Social Psychology, 11, 478489.
*Barrick, M. R., Stewart, G. L., Neubert, M. J., & Mount, M. K. (1998).
Relating member ability and personality to work-team processes and
team effectiveness. Journal of Applied Psychology, 83, 377391.
*Bhatara, K. (1972). Effect of cohesiveness on goal setting behavior in the
group. Jabalpur, India: Banarsidas Bhanot.
*Bird, A. M. (1977). Team structure and success as related to cohesiveness
and leadership. Journal of Social Psychology, 103, 217223.
*Blades, J. W. (1986). Rules for leadership: Improving unit performance.
Washington, DC: National Defense University Press.
Borman, W. C., & Motowidlo, S. J. (1993). Expanding the criterion
domain to include elements of contextual performance. In N. Schmitt &
W. C. Borman (Eds.), Personnel selection in organizations (pp. 7198).
San Francisco: Jossey-Bass.
Borucki, C. C., & Burke, M. J. (1999). An examination of service-related
antecedents to retail store performance. Journal of Organizational Be-
havior, 20, 943962.
*Burchfield, M. A. (1997). Personality composition as it relates to team
performance. Unpublished doctoral dissertation, Stevens Institute of
Technology, Hoboken, New Jersey.
Burke, M. J., & Landis, R. S. (2003). Methodological and conceptual
challenges in conducting and interpreting meta-analyses. In K. R. Mur-
phy (Ed.), Validity generalization: A critical review (pp. 287309).
Mahwah, NJ: Erlbaum.
Campbell, J. P. (1990). Modeling the performance prediction problem in
industrial and organizational psychology. In M. D. Dunnette & L. M.
Hough (Eds.), Handbook of industrial and organizational psychology
(2nd ed., Vol. 1, pp. 687732). Palo Alto, CA: Consulting Psychologists
Campbell, J. P., Gasser, M. B., & Oswald, F. L. (1996). The substantive
nature of performance variability. In K. R. Murphy (Ed.), Individual
differences and behavior in organizations (pp. 258299). San Francisco:
Campbell, J. P., McCloy, R. A., Oppler, S. H., & Sager, C. E. (1993). A
theory of performance. In N. Schmitt & W. C. Borman (Eds.), Personnel
selection in organizations (pp. 3570). San Francisco: Jossey-Bass.
Carless, S. A, & De Paola, C. (2000). The measurement of cohesion in
work teams. Small Group Research, 31, 7188.
*Carpenter, S., & Radhakrishnan, P. (2002). The relation between allo-
centrism and perceptions of in-groups. Manuscript submitted for
Carron, A. V. (1982). Cohesiveness in sport groups: Interpretations and
considerations. Journal of Sport Psychology, 4, 123138.
*Carron, A. V., & Ball, J. R. (1977). An analysis of the causeeffect
characteristics of cohesiveness and participation motivation in intercol-
legiate hockey. International Review of Sport Sociology, 12, 4960.
Carron, A. V., Colman, M. M., Wheeler, J., & Stevens, D. (2002). Cohe-
sion and performance in sport: A meta-analysis. Journal of Sport and
Exercise Psychology, 24, 168188.
Carron, A. V., Widmeyer, W., & Brawley, L. (1985). The development of
an instrument to assess cohesion in sport teams: The Group Environment
Questionnaire. Journal of Sport Psychology, 7, 244266.
Cartwright, D. (1968). The nature of group cohesiveness. In D. Cartwright
& A. Zander (Eds.), Group dynamics: Research and theory (3rd ed., pp.
91109). New York: Harper & Row.
*Cohen, D., Whitmyre, J. W., & Funk, W. H. (1960). Effect of group
cohesiveness and training upon creative thinking. Journal of Applied
Psychology, 44, 319322.
*Colarelli, S. M., & Boos, A. L. (1992). Sociometric and ability-based
assignment to work groups: Some implications for personnel selection.
Journal of Organizational Behavior, 13, 187196.
Cooper, H. (1998). Synthesizing research: A guide for literature reviews
(3rd ed.). Thousand Oaks, CA: Sage.
Cornwell, J. M. (1993). Monte Carlo comparisons of three tests for homo-
geneity of independent correlations. Educational and Psychological
Measurement, 53, 605618.
Cota, A. A., Evans, C. R., Dion, K. L., Kilik, L., & Longman, R. S. (1995).
The structure of group cohesion. Personality & Social Psychology
Bulletin, 21, 572580.
*Cotter, L. L. (1979). Group cohesiveness and team success among wom-
ens intercollegiate basketball teams. Unpublished doctoral dissertation,
Indiana University, Bloomington.
*Craig, T. Y., & Kelly, J. R. (1999). Group cohesiveness and creative
performance. Group Dynamics: Theory, Research, & Practice, 3, 243
*Darley, J. G., Gross, N., & Martin, W. C. (1952). Studies of group
behavior: Factors associated with the productivity of groups. Journal of
Applied Psychology, 36, 396403.
Davis, J. (1969). Group performance. Reading, MA: Addison-Wesley.
*Deep, S. D., Bass, B. M., & Vaughan, J. A. (1967). Some effects on
business gaming of previous quasi-T group affiliations. Journal of Ap-
plied Psychology, 51, 426431.
*Duffy, M. K., & Shaw, J. D. (2000). The Salieri Syndrome: Consequences
of envy in groups. Small Group Research, 31, 323.
*Eisenberg, J. (2001). The effects of reward schemes, individualism-
collectivism, and intrinsic motivation on teams creative performance.
Unpublished doctoral dissertation, Colorado State University, Fort
*Elias, F. G., Johnson, M. E., & Fortman, J. B. (1989). Task-focused
self-disclosure: Effects on group cohesiveness, commitment to task, and
productivity. Small Group Behavior, 20, 8796.
Evans C. R., & Dion, K. L. (1989). Group cohesion and performance: A
meta-analysis. Small Group Research, 22, 175186.
Festinger, L. (1950). Informal social communication. Psychological Re-
view, 57, 271282.
Festinger, L., Schachter, S., & Back, K. (1950). Social pressures in
informal groups: A study of human factors in housing. New York:
Harper & Brothers.
*Fiedler, F. E. (1954). Assumed similarity measures as predictors of team
effectiveness. Journal of Abnormal & Social Psychology, 49, 381388.
Finkelstein, L. M., Burke, M. J., & Raju, N. S. (1995). Age discrimination
in simulated employment contexts: An integrative analysis. Journal of
Applied Psychology, 80, 652663.
Flanagan, J. C. (1956). The evaluation of methods in applied psychology
and the problem of criteria. Occupational Psychology, 30, 19.
*Fodor, E. M., & Smith, T. (1982). The power motive as an influence on
group decision making. Journal of Personality and Social Psychology,
42, 178185.
*Fox, E. C. (1986). Team cohesion, ability, and coaches leadership
effectiveness as predictors of success in womens intercollegiate soft-
ball. Unpublished doctoral dissertation, University of Georgia, Athens.
*Gekoski, N. (1952). Predicting group productivity. Personnel Psychol-
ogy, 5, 281292.
*George, J. M., & Bettenhausen, K. (1990). Understanding prosocial
behavior, sales performance, and turnover. Journal of Applied Psychol-
ogy, 75, 698709.
Gleser L. J., & Olkin, I. (1994). Stochastically dependent effect sizes. In H.
Cooper & L. Hedges (Eds.), The handbook of research synthesis (pp.
339355). New York: Russell Sage Foundation.
*Gonzalez, M. G., Burke, M. J., Santuzzi, A. M., & Bradley, J. (in press).
The impact of group process variables on the effectiveness of distance
collaboration groups. Computers in Human Behavior.
*Goodacre, D. M. (1951). The use of a sociometric test as a predictor of
combat unit effectiveness. Sociometry, 14, 148152.
Gosenpud, J. (1989). The prediction of simulation performance as it is
affected by time. Simulation & Games, 20, 319350.
Gowda, R. G. (1988). Influence of individual characteristics and group
cohesiveness on programmer productivity. Unpublished doctoral disser-
tation, Georgia State University.
*Greene, C. N. (1989). Cohesion and productivity in work groups. Small
Group Behavior, 20, 7086.
Gully, S. M., Devine, D. J., & Whitney, D. J. (1995). A meta-analysis of
cohesion and performance: Effects of level of analysis and task inter-
dependence. Small Group Research, 26, 497520.
Hackman, J. R. (1968). Effects of task characteristics on group products.
Journal of Experimental Social Psychology, 4, 162187.
Hackman, J. R., & Morris, C. G. (1975). Group tasks, group interaction
process, and group performance effectiveness: A review and proposed
integration. In L. Berkowitz (Ed.), Advances in experimental social
psychology (Vol. 8, pp. 4599). New York: Academic Press.
Harrison, D. A., Price, K. H., & Bell, M. P. (1998). Beyond relational
demography: Time and the effects of surface- and deep-level diversity
on work group cohesion. Academy of Management Journal, 41, 96107.
*Hemphill, J. K., & Sechrest, L. B. (1952). A comparison of three criteria
of aircrew effectiveness in combat over Korea. Journal of Applied
Psychology, 36, 323327.
Hollingshead, A. B. (1998). Communication, learning and retrieval in
transactive memory systems. Journal of Experimental Social Psychol-
ogy, 34, 423442.
Hollingshead, A. B. (2000). Perceptions of expertise and transactive mem-
ory in work relationships. Group Processes and Intergroup Relations, 3,
*Hoegl, M., & Gemuenden, H. G. (2001). Teamwork quality and the
success of innovative projects: A theoretical concept and empirical
evidence. Organization Science, 12, 435449.
Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Cor-
recting for sources of error and bias in research findings. Newbury
Park, CA: Sage.
*Jaffe, E. D., & Nebenzahl, I. D. (1990). Group interaction and business
game performance. Simulation and Gaming, 21, 133146.
*Jehn, K. A. (1994). Enhancing effectiveness: An investigation of advan-
tages and disadvantages of value-based intragroup conflict. International
Journal of Conflict Management, 5, 223238.
*Jehn, K. A., & Shah, P. P. (1997). Interpersonal relationships and task
performance: An examination of mediation processes in friendship and
acquaintance groups. Journal of Personality and Social Psychology, 72,
Kabanoff, B., & OBrien, G. E. (1979). The effects of task type and
cooperation upon group products and performance. Organizational Be-
havior & Human Decision Processes, 23, 163181.
*Karau, S. J., & Hart, J. W. (1998). Group cohesiveness and social loafing:
Effects of a social interaction manipulation on individual motivation
within groups. Group Dynamics, 2, 185191.
*Keller, R. T. (1986). Predictors of the performance of project groups in R
& D organizations. Academy of Management Journal, 29, 715726.
*Keyton, J., & Springston, J. (1990). Redefining cohesiveness in groups.
Small Group Research, 21, 234254.
Kirchler, E., & Davis, J. H. (1986). The influence of member status
differences and task type on group consensus and member position
change. Journal of Personality and Social Psychology, 51, 8391.
Kirkman, B. L., & Rosen, B. (1999). Beyond self-management: Anteced-
ents and consequences of team empowerment. Academy of Management
Journal, 42, 5874.
*Klein, H. J., & Mulvey, P. W. (1995). Two investigations of the relation-
ships among group goals, goal commitment, cohesion, and performance.
Organizational Behavior and Human Decision Processes, 61, 4453.
Klein, K. J., Dansereau, F. & Hall, R. J. (1994). Levels issues in theory
development, data collection, and analysis. Academy of Management
Review, 19, 195229.
Koslowsky, M., & Sagie, A. (1993). On the efficacy of credibility intervals
as indicators of moderator effects in meta-analytic research. Journal of
Organizational Behavior, 14, 695699.
*Landers, D. M., Wilkinson, M. O., Hatfield, B. D., & Barber, H. (1982).
Causality and the cohesionperformance relationship. Journal of Sport
Psychology, 4, 170183.
*Langfred, C. W. (1998). Is group cohesiveness a double-edged sword? An
investigation of the effects of cohesiveness on performance. Small
Group Research, 29, 124143.
*Langfred, C. W. (2000). The paradox of self-management: Individual and
group autonomy in work groups. Journal of Organizational Behavior,
21, 563585.
LePine, J. A., Hanson, M., Borman, W., & Motowidlo, S. J. (2000).
Contextual performance and teamwork: Implications for staffing. In
G. R. Ferris & K. M. Rowland (Eds.), Research in personnel and human
resources management (Vol. 19, pp. 5390). Stamford, CT: JAI Press.
*Lodahl, T. M., & Porter, L. W. (1961). Psychometric score patterns, social
characteristics, and productivity of small industrial workgroups. Journal
of Applied Psychology, 45, 7379.
*Lorenz, C. E. (1985). The influence of expectations and cohesiveness on
the performance of small groups: A study of quality circles. Unpublished
doctoral dissertation, North Carolina State University, Raleigh.
Lott, A. J., & Lott, B. E. (1965). Group cohesiveness as interpersonal
attraction: A review of relationships with antecedent and consequent
variables. Psychological Bulletin, 64, 259309.
*Martens, R., & Peterson, J. A. (1971). Group cohesiveness as a determi-
nant of success and member satisfaction in team performance. Interna-
tional Review of Sport Sociology, 6, 4959.
Mathieu, J. E., Heffner, T. S., Goodwin, G. F., Salas, E., & Cannon-
Bowers, J. A. (2000). The influence of shared mental models on team
process and performance. Journal of Applied Psychology, 85, 273283.
Matheson, H., Mathes, S., & Murray, M. (1996). Group cohesion of female
intercollegiate coacting and interacting teams across a competitive sea-
son. International Journal of Sport Psychology, 27, 3749.
McGrath, J. E. (1984). Groups, interaction and performance. Englewood
Cliffs, NJ: Prentice-Hall.
*Melnick, M. J., & Chemers, M. M. (1974). Effects of group social
structure on the success of basketball teams. Research Quarterly, 45,
Mickelson, J. S., & Campbell, J. H. (1975). Information behavior: Groups
with varying levels of interpersonal acquaintance. Organizational Be-
havior and Human Decision Processes 13, 193205.
*Miesing, P., & Preble, J. F. (1985). Group process and performance in a
complex business simulation. Small Group Behavior, 16, 325338.
*Mossholder, K. W., & Bedeian, A. G. (1983). Group interaction pro-
cesses: Individual and group-level effects. Group & Organization Stud-
ies, 8, 187202.
Mudrack, P. E. (1989). Group cohesiveness and productivity: A closer
look. Human Relations, 9, 771785.
Mullen, B., & Copper, C. (1994). The relation between group cohesiveness
and performance: An integration. Psychological Bulletin, 115, 210227.
*Mulvey, P. W., & Klein, H. J. (1998). The impact of perceived loafing
and collective efficacy in group goal processes and group performance.
Organizational Behavior and Human Decision Processes, 74, 6287.
*Neal, D. J. (1997). Group competitiveness and cohesion in a business
simulation. Simulation & Gaming, 28, 460476.
*Neubert, M. J. (1999). Too much of a good thing or the more the merrier?
Exploring the dispersion and gender composition of informal leadership
in manufacturing teams. Small Group Research, 30, 635646.
*Norris, D. R., & Niebuhr, R. E. (1980). Group variables and gaming
success. Simulation and Games, 11, 301312.
Oliver, L. W., Harman, J., Hoover, E., Hayes, S. M., & Pandhi, N. A.
(1999). A quantitative integration of the military cohesion literature.
Military Psychology, 11, 5783.
Ostroff, C. (1993). Comparing correlations based on individual-level and
aggregated data. Journal of Applied Psychology, 78, 569582.
Paese, P. W., & Switzer, F. S., III. (1988). Validity generalization and
hypothetical reliability distributions. Journal of Applied Psychology, 73,
*Podsakoff, P. M., MacKenzie, S. B., & Ahearne, M. (1997). Moderating
effects of goal acceptance on the relationship between group cohesive-
ness and productivity. Journal of Applied Psychology, 82, 974983.
*Porter, T. W., & Lilly, B. S. (1996). The effects of conflict, trust, and task
commitment on project team performance. International Journal of
Conflict Management, 7, 361376.
Raju, N. S., Burke, M. J., Normand, J., & Langlois, G. M. (1991). A new
meta-analytic approach. Journal of Applied Psychology, 76, 432446.
Raju, N. S., & Drasgow, F. (2003). Maximum likelihood estimation in
validity generalization. In K. R. Murphy (Ed.), Validity generalization:
A critical review (pp. 263286). Hillsdale, NJ: Erlbaum.
Raju, N. S., Pappas, S., & Williams, C. P. (1989). An empirical Monte
Carlo test of the accuracy of the correlation, covariance, and regression
slope models for assessing validity generalization. Journal of Applied
Psychology, 74, 901911.
Sackett, P. R., Harris, M. M., & Orr, J. M. (1986). On seeking moderator
variables in the meta-analysis of correlational data: A Monte Carlo
investigation of statistical power and resistance to Type-1 error. Journal
of Applied Psychology, 71, 302310.
Schmidt, F. L., & Hunter, J. E. (1977). Development of a general solution
to the problem of validity generalization. Journal of Applied Psychology,
62, 529540.
Scullen, S. E. (1997). When ratings from one source have been averaged,
but ratings from another source have not: Problems and solutions.
Journal of Applied Psychology, 82, 880888.
Seashore, S. E. (1954). Group cohesiveness in the industrial work group.
Ann Arbor: Institute for Social Research, University of Michigan.
*Seers, A., Petty, M. M., & Cashman, J. F. (1995). Team-member ex-
change under team and traditional management: A naturally occurring
quasi-experiment. Group & Organization Management, 20, 1838.
*Smith, K. G., Smith, K. A., Olian, J. D., Sims, H. P., OBanon, D. P., &
Scully, J. A. (1994). Top management team demography and process:
The role of social integration and communication. Administrative Sci-
ence Quarterly, 39, 412438.
Smith, P. C. (1976). Behavior, results, and organizational effectiveness:
The problem of criteria. In M. D. Dunnette (Ed.), Handbook of industrial
and organizational psychology (pp. 745775). Chicago: Rand McNally.
Spector, P. E., & Levine, E. L. (1987). Meta-analysis for integrating study
outcomes: A Monte Carlo study of its susceptibility to Type I and Type
II errors. Journal of Applied Psychology, 72, 39.
Steel, P. D., & Kammeyer-Mueller, J. D. (2002). Comparing meta-analytic
moderator estimation techniques under realistic conditions. Journal of
Applied Psychology, 87, 96111.
Steiner, I. (1972). Group processes and productivity. San Diego, CA:
Academic Press.
Stewart, G. L., & Barrick, M. R. (2000). Team structure and performance:
Assessing the mediating role of intrateam process and the moderating
role of task type. Academy of Management Journal, 43, 135148.
*Stinson, J. E., & Hellebrandt, E. T. (1972). Group cohesiveness, produc-
tivity, and strength of formal leadership. Journal of Social Psychology,
87, 99105.
Stogdill, R. M. (1972). Group productivity, drive, and cohesiveness. Or-
ganizational Behavior and Human Performance, 8, 2643.
*Tehan, T. A. (1983). The development and evaluation of effectiveness of
cohesiveness training with high school basketball teams. Unpublished
doctoral dissertation, Virginia Commonwealth University, Richmond.
*Terborg, J. R., Castore, C., & DeNinno, J. A. (1976). A longitudinal field
investigation of the impact of group composition on group performance
and cohesion. Journal of Personality and Social Psychology, 34, 782
*Tesluk, P. E., & Mathieu, J. E. (1999). Overcoming roadblocks to effec-
tiveness: Incorporating management of performance barriers into models
of work group effectiveness. Journal of Applied Psychology, 84, 200
Tesluk, P. E., Mathieu, J. E., Zaccaro, S. J., & Marks, M. A. (1997). Task
and aggregation issues in the analysis and assessment of team perfor-
mance. In M. T. Brannick, E. Salas, & C. Prince (Eds.), Team perfor-
mance and measurement: Theory, methods, and applications (pp. 197
224). Mahwah, NJ: Erlbaum.
Thompson, J. D. (1967) Organizations in action. New York: McGraw-Hill.
Tziner, A. (1982). Differential effects of group cohesiveness types: A
clarifying overview. Social Behavior and Personality, 10, 227239.
*Tziner, A., & Vardi, Y. (1983). Ability as a moderator between cohe-
siveness and tank crews performance. Journal of Occupational Behav-
ior, 4, 137143.
Wanous, J. P., Sullivan, S. E., & Malinak, J. (1989). The role of judgment
calls in meta-analysis. Journal of Applied Psychology, 74, 259264.
*Wech, B. A., Mossholder, K. W., Steel, R. P., & Bennett, N. (1998). Does
work group cohesiveness affect individuals performance and organiza-
tional commitment? A cross-level examination. Small Group Research,
29, 472494.
Weed, S. E., Mitchell, T. R., & Moffitt, W. (1976). Leadership style,
subordinate personality, and task type as predictors of performance and
satisfaction with supervision. Journal of Applied Psychology, 61, 58
Wegner, D. M., Erber, R., & Raymond, P. (1991). Transactive memory in
close relationships. Journal of Personality and Social Psychology, 61,
*Wekselberg, V., Goggin, W. C., & Collings, T. J. (1997). A multifaceted
concept of group maturity and its measurement and relationship to group
performance. Small Group Research, 28, 328.
Whitener, E. M. (1990). Confusion of confidence intervals and credibility
intervals in meta-analysis. Journal of Applied Psychology, 75, 315321.
Wilderom, P. M., Glunk, U., & Maslowski, R. (2000). Organizational
culture as a predictor of organizational performance. In N. M. Ashka-
nasy, C. P. M. Wilderom, & M. F. Peterson (Eds.), Handbook of
organizational culture and climate (pp. 193209). Thousand Oaks, CA:
*Williams, J. M., & Hacker, C. M. (1982). Causal relationships among
cohesion, satisfaction, and performance in womens intercollegiate field
hockey teams. Journal of Sport Psychology, 4, 324337.
*Wolfe, J., & Box, T. M. (1988). Team cohesion effects on business game
performance. Simulation and Games, 19, 8298.
*Wong, L. (1992). The effects of cohesion on organizational performance:
A test of two models. Unpublished doctoral dissertation, Texas Techno-
logical University, Lubbock.
*Zaccaro, S. J., Gualtieri, J., & Minionis, D. (1995). Task cohesion as a
facilitator of team decision making under temporal urgency. Military
Psychology, 7, 7793.
*Zaccaro, S. J., & Lowe, C. A. (1986). Cohesiveness and performance on
an additive task: Evidence for multidimensionality. Journal of Social
Psychology, 128, 547558.
*Zaccaro, S. J., & McCoy, C. (1988). The effects of task and interpersonal
cohesiveness on performance of a disjunctive group task. Journal of
Applied Social Psychology, 18, 837851.
(Appendixes follow)
Received January 7, 2002
Revision received January 28, 2003
Accepted March 24, 2003
Appendix B
Examples of Studies That Were Multiply Coded Under the BehaviorOutcome and EffectivenessEfficiency Categories
Effectiveness Efficiency
Behavior Total time spent engaging in work-related activities (Bakeman
& Helmreich, 1975); the number of action proposals in a
group decision-making task (Fodor & Smith, 1982); overall
evaluation of behaviors exhibited in an oral presentation
(Keyton & Springston, 1990)
Number of ideas generated in twelve minutes for
brainstorming-task (Cohen et al., 1960); summing and
extending invoices in a seven minute period (Wong, 1992);
number of product solution ideas, weighted by ratings of
creativity (Eisenberg, 2001)
Outcome Winloss ratio for athletic teams (Bird, 1977; Carron & Ball,
1977); overall score on Moon Survival Task (Carpenter &
Radhakrishnan, 2002); rank or overall grade on a class
project or business simulation (Colarelli & Boos, 1992;
Neal, 1997)
Points gained in a fifteen minute period on a cargo-loading
simulation (Burchfield, 1997); supervisory rating of units
produced and errors made (Gekoski, 1952); total monthly
sales, divided by number of employees (George &
Bettenhausen, 1990)
Appendix A
Examples and Coding Agreement for Each Moderator Variable Included in the Study
Characteristic Examples Agreement (
Type of performance .84
Behavior Ratings of specific combat behaviors during tactical field problems, supervisory ratings of specific
work behaviors (e.g., knowledge of tasks, planning, and so forth), cognitive performance on
puzzle tasks
Outcome Final grades on class projects, supervisor ratings of work group outcomes, and winloss ratio in
sports teams
Type of performance measure .89
Effectiveness Ratings or counts of total group output, team ranking in a business simulation, and quality of
group decisions
Efficiency Group output over a specific period of time, return on investment in simulation games and actual
organizations, brainstorming results over a specific period of time, and supervisor ratings of
Pattern of team workflow .96
Pooled Individual sales performance (that was then aggregated to the group level), solving of individual
puzzles, and collegiate wrestling competitions
Sequential Clerical work that proceeded in stages and group card-sorting tasks (in which one member
follows the same person each time)
Reciprocal Surveying teams, group puzzle tasks (in which members interact with other group members one at
a time to complete the task), and class projects conducted over e-mail (i.e., members interact
with other members, but there is no simultaneous interaction)
Intensive Decision-making tasks, group puzzle tasks (in which members must interact with the other team
members simultaneously to complete the puzzle), business simulations, and class group
Component of cohesion
Interpersonal attraction Manipulations of attitude similarity, sociometric nominations, measures of attitude similarity, and
any items indicating preference for the members of the group
Task commitment Measures and manipulations of task enjoyment, importance, or attraction (personal or group)
Group pride Items assessing the value of group membership, importance of belonging to the group, and
measures of attraction to the group itself (i.e., apart from its members)
Two coders rated each category of moderator for all studies. For example, each correlation was examined for whether performance was a behavior, an
outcome, or ambiguous with respect to this category.
Cohesion and Performance in Groups:
A Meta-Analytic Clarification of Construct Relations
Daniel J. Beal, Robin R. Cohen, Michael J. Burke, and Christy L. McLendon
Published in the Journal of Applied Psychology (2003), Volume 88, Issue 6, pp. 989-1004
'This article may not exactly replicate the final version published in the APA journal. It is not the
copy of record.'
... Cohesion refers to the strength and extent of interpersonal connection between team members. Cohesion is a multidimensional team attribute that includes task cohesion, social cohesion, and group pride, each linked with team performance in multiple studies [31,32]. Several comprehensive meta-analyses have linked social and task cohesion to performance in teams whose taskwork requires intensive between-member interactions, characteristic of TTs [32]. ...
... Cohesion is a multidimensional team attribute that includes task cohesion, social cohesion, and group pride, each linked with team performance in multiple studies [31,32]. Several comprehensive meta-analyses have linked social and task cohesion to performance in teams whose taskwork requires intensive between-member interactions, characteristic of TTs [32]. The relationship between cohesion and performance has been consistently positive across many team types, but found to be highly dependent on intra-group processes. ...
Full-text available
A translational team (TT) is a specific type of interdisciplinary team that seeks to improve human health. Because high-performing TTs are critical to accomplishing CTSA goals, a greater understanding of how to promote TT performance is needed. Previous work by a CTSA Workgroup formulated a taxonomy of 5 interrelated team-emergent competency “domains” for successful translation: 1). affect, 2). communication, 3). management, 4). collaborative problem-solving, and 5). leadership. These Knowledge Skills and Attitudes (KSAs) develop within teams from the team’s interactions. However, understanding how practice in these domains enhance team performance was unaddressed. To fill this gap, we conducted a scoping literature review of empirical team studies from the broader Science of Team Science literature domains. We identified specific team-emergent KSAs that enhance TT performance, mapped these to the earlier “domain” taxonomy, and developed a rubric for their assessment. This work identifies important areas of intersection of practices in specific competencies across other competency domains. We find that inclusive environment, openness to transdisciplinary knowledge sharing, and situational leadership are a core triad of team-emergent competencies that reinforce each other and are highly linked to team performance. Finally, we identify strategies for enhancing these competencies. This work provides a grounded approach for training interventions in the CTSA context.
... We expect that the effects of pathos appeals will be strongest in teams with high interdependence because this teamwork strengthens shared mental models (Mathieu, Heffner, Goodwin, Salas, & Cannon-Bowers, 2000) and transactive memory systems (Hollingshead, 2000) that facilitate communication. Likewise, interdependent teams are more cohesive, implying they are conducive to the receipt of information conveyed using shared meanings and shared emotions (cf., Beal, Cohen, Burke, & McLendon, 2003). Thus, empirical research is needed to explore how these boundary conditions affect the operation of pathos appeals. ...
Contemporary models of team adaptation are highly cognitive in nature and operate under the assumption that teams possess the capacity to adapt. However, teams also face contingencies that erode their adaptability, threatening to stall or undermine the adaptation process. Anecdote points to the ability of some teams to employ affective communication to adapt in the face of these internal or external contingencies. Drawing from cinematic examples, we consider how pathos appeals influence team adaptation. Specifically, we examine three cinematic stories in which pathos appeals were used to leverage emotion by adapting teams. These stories enhance our understanding of team adaptation in the face of difficult and often unforeseen challenges, laying the foundation for new theory and an expanded understanding of how emotion and persuasion affect team change. We discuss these implications and propose new directions for future research.
... Team bonding capital is important in organisations using project-based teams [68] because their services are complex [15,55,[69][70][71]. Team bonding capital has been found to motivate cooperative interactions in project-based teams [72] and along these lines, we posit that high team bonding capital promotes team motivation to acquire, share and integrate knowledge from different sources. ...
Full-text available
Professional service firms apply specialist knowledge to create customised solutions to client problems. In their work, teams of professionals undertake projects in which clients may be closely involved in co-creating solutions. However, we know little about the conditions under which client involvement contributes to better performance. We examine the direct and conditional contribution client involvement can make to project success and propose team bonding capital as a moderator. We conduct multi-level analysis of data from 58 project managers and 171 consultants nested in project teams. We find a positive impact of client involvement on both team performance and team member idea creativity. Team bonding capital moderates the relationships client involvement has with both team performance and individual member idea creativity, where the impact of client involvement is greater when team bonding capital is high. Implications for theory and practice are discussed.
... The design of game features for teamwork can be influenced by organizational psychology, where many conditions for excellent team performance have been suggested [26]. Interdependence and team cohesion, among other things, are cited in relation to team performance [27], therefore these variables are suitable beginning points for developing gamified components for 3D virtual teamwork. ...
Full-text available
Virtual teamwork has spread strongly in various fields in recent times. COVID lockdown measures radically boosted the use of distance-independent remote collaboration methods. Although there are many modern virtual 3D spaces available for teamwork today, they lack additional motivational factors other than the visual 3d experience. Based on this, we tried to research the possible benefits of using gamification technics in 3D virtual collaboration environments to increase virtual team members' motivation. This article attempts to explore and highlight the possibilities of this area, and propose a model framework for implementation of gamified elements in 3D virtual teamwork environments.
Full-text available
This study contributes a novel perspective on how new venture teams navigate task re-allocation during the new venture development phase. It highlights the relevance of task re-allocation conflict, shows how “negative affect expectations” shape the unfolding of such conflicts, and demonstrates why acting out conflict and its associated negative affect can enable team members to make substantial task re-allocations instead of symbolic ones. The analysis has implications for two bodies of research, which have previously not been considered in tandem: (1) research on task (re-)allocation, professionalization, and structural imprinting in new ventures, and (2) research on team conflict.
Auszug aus dem entsprechenden Buch Schlaile/Stöber (2022, S.13): Den Aufschlag machen in Kapitel 2 Sebastian Müller und Nils Christian Hoffmann, indem sie zunächst eine rollentheoretische Perspektive auf digitalen Konsum entwickeln und damit einen Beitrag zur konzeptionellen Begründung von Rollenverantwortung leisten. Gerade in Bezug auf nachhaltige, bewusstere Konsumentscheidungen stehen Konsumierende häufig vor einer Dilemmasituation, in der sie einem Rollenkonflikt ausgesetzt sind. Daran anknüpfend liefern Müller und Hoffmann methodische Einblicke in einen für die Konsumforschung adaptierten Forschungsansatz. Dieser moralpsychologische Ansatz verspricht, auch im digitalen Raum verschiedene rollenbedingte Konsummuster und normative Argumente zu erfassen, Rollenkonflikte aufzudecken und damit ggf. auch institutionelle (z.B. politische) Gestaltungsansätze zur Vereinfachung von Verantwortungsübernahme unterstützen zu können.
The Cambridge Handbook of Creativity and Emotions provides a state-of-the-art review of research on the role of emotions in creativity. This volume presents the insights and perspectives of sixty creativity scholars from thirteen countries who span multiple disciplines, including developmental, social, and personality psychology; industrial and organizational psychology; neuroscience; education; art therapy, and sociology. It discusses affective processes – emotion states, traits, and emotion abilities – in relation to the creative process, person, and product, as well as two major contexts for expression of creativity: school, and work. It is a go-to source for scholars who need to enhance their understanding of a specific topic relating to creativity and emotion, and it provides students and researchers with a comprehensive introduction to creativity and emotion broadly.
Full-text available
This study presents paradoxical tensions as the ‘missing link’ at the intersection of the entrepreneurial family and family firm management: a link that crucially affects innovation-related decisions such as ambidexterity. Specifically, the study argues that the relationship between family cohesion and organizational ambidexterity within entrepreneurial family firms is mediated by paradoxical tensions (latent and salient). Drawing on survey data from 206 German family firms, support is found for the hypotheses advanced. Building on cognitive resource theory, this study demonstrates that differently perceived paradoxical tensions in entrepreneurial family firms have different meaningful effects on organizational ambidexterity. Implications for theory and future research are discussed.
Selling is often associated with competitiveness, which has led many scholars to examine trait competitiveness as a driver of salesperson performance. However, it is also suggested that competitiveness contributes to less cohesive work environments which can have adverse effects on firms and their stakeholders. The authors explore this dichotomy by first conducting a field study utilizing multilevel, multisource data acquired from 358 salespeople working at 86 offices of a large company. Using Mplus, the authors employ a 1-2-1 multilevel model to test the effects of trait competitiveness on work group cohesion, along with subsequent effects on sales performance and turnover. Findings reveal a surprising positive relationship between trait competitiveness and work group cohesion, with the latter reducing turnover likelihood. Thus, trait competitiveness has indirect effects, through work group cohesion, in reducing turnover, while having direct and positive effects on sales performance. Results of a follow-up qualitative study reveal that the positive relationship between trait competitiveness and work group cohesion can be attributed to the competitive focus (i.e. interpersonal, intergroup, or intrapersonal) of the individuals within the sales organization. Overall, this work shows that organizations can have both competitive salespeople and cohesive work environments, contributing beneficially toward sales output and turnover.
Studies using the single-aggregate approach (L. R. James, 1982), where assessments made by individual respondents are correlated with other assessments that have been averaged across multiple respondents, can exhibit a systematic bias of 20 to 70% or more if they are used to estimate individual level relationships. Not only may results of such studies be erroneous, but theory development based on such studies may be misguided. A comprehensive solution (nested and crossed designs) to the single–aggregation problem is provided through generalizability theory. Results show that the aggregation bias is a function of both the generalizability (reliability) of individual responses and the number of individuals per group. Conceptual parallels to classical measurement theory are discussed. Factors are presented for converting single–aggregated correlations and standard deviations to estimates of the corresponding values using the individual as the level of analysis. (PsycINFO Database Record (c) 2016 APA, all rights reserved)
The purpose of this study was to examine whether, across a season of competition, women's intercollegiate coacting (gymnastics, swimming) and interacting (lacrosse, basketball) teams differed in cohesion. Seventy female athletes ranging in age from 18-22 years (gymnastics, n = 18; swimming, n = 22; lacrosse, n = 16; basketball, n = 13) participated in the study. The Group Environment Questionnaire (Widmeyer, Brawley, & Carron, 1985), which is composed of four subscales: Attraction to the Group-Task (ATG-T), Attraction to the Group-Social (ATG-S), Group Integration-Task (GI-T), Group Integration-Social (GI-S) was used to measure cohesion. Testing for cohesion occurred pre-season (3 days prior to first regular competition), mid-season, and post-season, A 2 × 2 ANCOVA was utilized to compare coacting and interacting team cohesion scores on each subscale at mid-season and post-season. Preseason measures were utilized as covariates. A significant sport by time interaction (p<.05) was found on only the ATG-T subscale at mid-season with coacting teams displaying significantly greater changes in cohesion than interacting teams. These findings are discussed in terms of gender, type of activity, overall season performance, and NCAA division status.
Organizations are increasingly relying on informal leaders within teams to positively affect team outcomes. Also, as the percentage of women in the workforce grows, women are likely to shoulder more of the informal leadership responsibility within teams. This study explores the relationship of team performance and team cohesion to informal leadership dispersion and the gender composition of the informal leadership core within 21 intact manufacturing teams. The results indicate that informal leadership dispersion (the number of perceived informal leaders divided by team size) is positively related to team cohesion. In addition, when females compose a larger portion of the informal leadership core, supervisors'ratings of team performance are higher.
This study extends previous work reviewing the cohesion-performance relationship by using meta-analytic techniques to assess the effects of level of analysis and task interdependence on the cohesion-performance relationship. A totel of 51 effect sizes from 46 empirical studies were obtained for the meta-analytic integration. Results suggest that level of analysis and task interdependence moderate the cohesion-performance relationship. Implications of the findings for future research on group cohesion and performance are discussed.