?

Academy of Management Journal

2001, Vol. 44, No. 2, 316-325.

SOCIAL NETWORKS AND THE PERFORMANCE

OF INDIVIDUALS AND GROUPS

RAYMOND T. SPARROWE
Cleveland State University

ROBERT C. LIDEN
SANDY J. WAYNE
University of Illinois at Chicago

MARIA L. KRAIMER
Cleveland State University

A field study involving 190 employees in 38 work groups representing five diverse
organizations provided evidence that social networks, as defined in terms of both
positive and negative relations, are related to both individual and group performance.
As hypothesized, individual job performance was positively related to centrality in
advice networks and negatively related to centrality in hindrance networks composed
of relationships tending to thwart task behaviors. Hindrance network density was
significantly and negatively related to group performance.

A growing body of management theory andgrades.

re- Brass (1981) found that the centrality of
search takes as its central premise the embeddedness employees' positions in a network representing the
(Granovetter, 1985) of individuals in social networks. flow of work was indirectly related to job perfor-
The distinctive characteristic of this stream of re- mance via job characteristics. Thus, one purpose of
search lies in how it draws on the structural proper- the current study was to replicate and extend pre-
ties of social networks in explaining outcomes. From vious research on the relationship between an in-
this perspective, individuals' positions within social dividual's network position within a work group
networks confer advantages, such as organizational and his or her job performance by examining the
assimilation (Sparrowe & Liden, 1997) and promo- role of informal network position in actual work
tions (Burt, 1992), or lead to disadvantages, such as settings.
organizational exit (Krackhardt & Porter, 1986). A related issue is whether group performance is a
Centrality, the extent to which a given individual is function of the structure of informal relationships
connected to others in a network, is the structural within groups. Although the relationship between
property most often associated with instrumental out- group interaction and performance has been the
comes, including power (Brass, 1984), influence in subject of considerable previous research, structure
decision making (Friedkin, 1993), and innovation has largely been viewed in terms of formal relation-
(Ibarra, 1993). ships rather than informal interaction patterns
Although previous research has demonstrated a (Guzzo & Shea, 1992). An important exception is
relationship between network structure and instru- group research conducted during the 1950s that
mental outcomes, relatively few studies have ex- examined the relationship between group commu-
plicitly examined the link between network cen- nication structure and performance (for a review,
trality and job performance. Baldwin, Bedell, and see Shaw [1964]). More recently, Baldwin and col-
Johnson (1997) found a positive relationship be- leagues' (1997) M.B.A. team study found that team
tween the network centrality of master of business interaction patterns consistent with cohesive work
administration (M.B.A.) team members and their groups were positively related to teams' final
grades. Thus, the second purpose of this study was
to extend this early and more recent group research
The authors gratefully acknowledge the funding of the
by examining group network structure and perfor-
Center for Human Resources Management (CHRM) at the
mance in work settings.
University of Illinois for this research. The interpreta-
tions, conclusions, and recommendations are those of the A third contribution of our study is its examina-
authors and do not necessarily represent those of CHRM. tion of the structure of informal relationships that
We also thank Timothy Franz and Noah Pinko for their potentially hinder individual and group perfor-
research assistance. mance. Although most research on informal net-
316

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
2001 Sparrowe, Liden, Wayne, and Kraimer 317

works has focused on positive or neutral relations, dent on any single individual (Cook & Emerson,
negative relations are a recognized possibility. 1978). Centrality also implies control over the re-
Drawing from research on "negative asymmetry" source acquisition of others because central indi-
(see Taylor [1991] for a review), Labianca, Brass, viduals can choose from a greater number of alter-
and Gray (1998) argued that negative relations in native individuals when exchanging beneficial
social networks are more important than positive resources. Moreover, the study's focus on centrality
relations for understanding attitudes and behaviors allows its results to be interpreted in the context of
because negative relations are more salient. Recent previous research that has demonstrated relation-
empirical research supports these authors' conten- ships between centrality and power (Brass, 1984),
tion that negative relationships may have a nega- influence in decision making (Friedkin, 1993), and
tive effect on attitudes and behaviors. Baldwin and innovation (Ibarra, 1993). However, an important
colleagues (1997) found that an individual M.B.A. difference between this study and recent research
team member's centrality in an "adversarial" net- is that we focus on centrality within work groups
work was negatively related to his or her satisfac- rather than within an organization as a whole.
tion. At the group level, the number of adversarial At the group level of analysis, we examined two
relations within a team was negatively related to structural properties of interaction patterns in rela-
perceptions of team effectiveness but positively re- tion to performance: network density and network
lated to the team grade. Similarly, Labianca and centralization. Density describes the overall level of
colleagues (1998) found that the number of nega- interaction of various kinds reported by network
tive (avoidance) relationships individuals had with members. Centralization reflects the extent to
out-group members was positively related to per- which interactions are concentrated in a small
ceptions of intergroup conflict, whereas the num- number of individuals rather than distributed
ber of friendship relations was not related to per- equally among all members. Density is analogous t
ceptions of intergroup conflict. Taken together, the mean number of ties per group member. T
these studies demonstrate the importance of nega- more ties each group member enjoys with oth
tive relationships in relation to attitudes and be- group members, the greater the density of the n
haviors. work. Group centralization, in contrast, is ana
gous to the variance in network ties per grou
member. When the variance in the number of net-
CONCE1PTUAL BACKGROUND AND
HYPOTHESES work ties per group member is low, no group mem-
ber enjoys substantially more ties than any other
Among the fundamental explanatory tenets groupofmember, and therefore no group member is
the social network perspective is the idea that
morethe
central than any other. Conversely, when the
structure of social interactions enhances or con- variance in the number of network ties per group
strains access to valued resources (Brass, 1984; member is high, some members have proportion-
Ibarra, 1993). Resources exchanged through infor- ately more ties and therefore are more central than
mal networks include work-related resources, such other group members. These attributes of network
as task advice and strategic information, but infor- structure are closely related to research conducted
mal networks also transmit social identity (norms) in the 1950s (Shaw, 1964) relating communication
and social support (Podolny & Baron, 1997). We patterns to group performance and so enable the
focus on the exchange of task advice and informa- results of our study to be interpreted in the context
tion, because these resources are likely to be posi- of previous work. As has previous small group
tively related to job performance, and on hindrance research, this study examined density and central-
relations, because of their potentially negative ef- ization in relation to work groups rather than whole
fects on job performance. organizations.
Our hypotheses regarding the relationship be-
tween network structure and performance are pre-
Advice Network Centrality and Individual
sented first at the individual level of analysis and Performance
then at the group level. At the individual level, we
focused on network centrality because it captures Advice networks are comprised of relations
the extent of an individual's access to resources, through which individuals share resources such as
such as task-specific knowledge and confidential information, assistance, and guidance that are re-
information about work-related issues. Central in- lated to the completion of their work. When the
dividuals, because of their more numerous connec- work performed by individuals is enhanced by task
tions to others, have more relationships to draw information available from others, an advice net-
upon in obtaining resources and so.are less depen- work is a means for obtaining resources that are

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
318 Academy of Management Journal April

instrumental in facilitating individual job perfor- Advice Network Structure and Group
mance. Centrality in the advice network reflects an Performance
individual's involvement in exchanging assistance
Just as one's position in social network structure
with coworkers and engaging in mutual problem
is expected to be related to individual performance,
solving. An individual who is central in the advice
network is, over time, able to accumulate knowl-
the social network structure of a group should be
associated with group performance. We expect that
edge about task-related problems and workable so-
the density of an advice network will be positively
lutions (Baldwin et al., 1997). This expertise not
related to group performance. Specifically, when
only enables the central individual to solve prob-
group members exchange advice with a larger pro-
lems readily, but also serves as a valued resource
portion of other group members, the group should
for future exchanges with coworkers. As others be-
benefit in terms of greater cooperation, greater in-
come dependent on a central individual for impor-
formation sharing, a stronger sense of accountabil-
tant advice, he or she gains an advantage that can
ity, greater agreement on expectations, and less ten-
be used in future exchanges for valued resources
dency to engage in social loafing. A relatively larger
(Cook & Emerson, 1978). Conversely, those who are
number of group members exchanging advice indi-
in peripheral positions in the advice network
cates greater mutual interdependence between
should find it much more difficult to develop ex-
members. As Molm (1994) suggested, mutual inter-
pertise about task-related problems and solutions
dependence fosters cooperation, which in turn en-
and are thus less likely to develop the competen-
hances group performance. A dense advice net-
cies and expertise necessary for high levels of per-
formance. work also benefits groups through the sharing of
information. The more members involved in the
Hypothesis 1. Centrality in a work group's ad- exchange of advice, the more pieces of nonredu
vice network will be positively related to an dant information that are likely to be share
individual's job performance. Amount of information sharing is related to t
quality of group decisions (Larson, Christense
Franz, & Abbott, 1998). The exchange of advic
Hindrance Network Centrality and Individual
Performance among a large proportion of a group's membe
should also make each member more aware of other
Negative exchange relations have been described group members' roles in the group. By advising one
in terms of such behaviors as interference, threats, another, members learn about the responsibilities
sabotage, and rejection (Sahlins, 1972) as well as inof each group member. Knowledge of each group
relation to affective responses to such behaviors,member's roles makes task behavior more visible
including annoyance, emotional upset, and anger and at the same time clarifies expectations and
(Pagel, Erdly, & Becker, 1987). Given the practical accountability. Increased visibility and account-
difficulties inherent in obtaining valid data aboutability counteract social loafing and thus enhance
negative relationships in field studies, researchersgroup performance (Wagner, 1995).
have used proxies in identifying network relations
characterized by negative exchanges such as avoid- Hypothesis 3a. The density of a work group's
ance (Labianca et al., 1998) or by adversarial rela- advice network will be positively related to
tionships (Baldwin et al., 1997). Because we are group performance.
interested in performance, we focused on relation-
Although we expect a positive relationship be-
ships with coworkers who thwart task behaviors,
tween advice network density and group perfor-
termed a hindrance network. Centrality in a hin-
drance network reflects the extent to which co-
mance, we contend that the relationship between
advice network centralization and group perfor-
workers describe a focal individual as a person who
mance will be negative. Our hypothesis regarding
makes it difficult for them to complete their work
centralization and group performance is consistent
by withholding valuable information, resources,
with the findings of the experimental research
and opportunities. To the extent that an individu-
conducted in the 1950s in which groups with
al's performance depends upon access to valuable decentralized communication networks were
resources from coworkers, hindrance relations will
more productive at complex tasks than were gr
be detrimental to performing various aspects of
with centralized communication networks (Shaw,
the job.
1964). All of the organizations included in the cur-
Hypothesis 2. Centrality in a workgroup's hin- rent study had implemented empowerment, which
drance network will be negatively related to had the effect of making respondents' tasks com-
individual job performance. plex. Even in the manufacturing organization we

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
2001 Sparrowe, Liden, Wayne, and Kraimer 319

studied, employees operated complex computer- group behavior (Borman, Hanson, Oppler, Pulakos,
guided machinery that make their jobs relatively & White, 1993). Thus, the greater the proportion of
complex. hindrance relations among a group's members (that
Our theoretical foundation for hypothesizing a is, the higher the density of the hindrance net-
negative relation between advice network central- work), the lower the group's performance will be.
ization and group performance is drawn from
Hypothesis 4. The density of a work group's
Molm's (1994) distinctions among independence,
hindrance network will be negatively related to
dependence, and interdependence. These arguments
group performance.
suggest that decentralized networks foster inter-
dependence, which in turn encourages cooperation.
Cooperation is encouraged because, in inter- METHODS
dependent relations, exchange partners share con-
Sample and Procedures
trol over joint outcomes. And, unlike actors in de-
pendent relations, who can gain at the expense of Respondents were drawn from five organizat
others, actors in interdependent relations cannot whose employees participated in a larger rese
"receive benefits without contributing to their pro- project on work group processes: a large pu
duction" (Molm, 1994: 165). Because network cen- university, a small manufacturing firm in the
tralization captures the extent to which exchange struction industry, a large distributor of indus
relations are concentrated among a few individu- products, and two geographically separate plant
als, the greater the centralization in an advice a consumer products company. Forty-seven
network, the less interdependence, and the less groups voluntarily participated in the research;
cooperation. As cooperation in the task domain response rate as a percentage of all groups w
decreases, group performance will suffer. vited to participate was 96 percent. Network
general work attitudes surveys were administ
Hypothesis 3b. Centralization in a work group's
on site in work groups to 269 individuals du
advice networks will be negatively related to
work hours. Group leaders (formal superviso
group performance.
individuals designated as team leaders) compl
questionnaires to assess individual and group-
Hindrance Network Structure and Group performance.
Performance Because network analysis requires a high
Our rationale for the negative effects of hin-
sponse rate (Wasserman & Faust, 1994), gr
with less than 80 percent participation wer
drance networks on individual performance may
cluded. Analyses were based on the remainin
be extended to the group level.1 Although one rel-
groups: 9 from the university, 5 from the manu
atively isolated hindering group member may not
turing firm, 8 from the distributor of indus
have much impact on a group, several such people
products, and 16 from the consumer products m
can adversely affect group performance. Specifi-
ufacturer. Because group leaders provided per
cally, the more group members whose job-related
mance data, they were not included in the netw
behaviors are thwarted by a hindering member or
data. The total sample for testing hypotheses at
members, the greater the harm to group perfor-
individual level of analysis was 190. The ave
mance. Our logic follows the general finding in
group size, excluding leaders, was 5.0 memb
group research that although the whole is more
The average length of time since group form
than the sum of the parts, the average of individual
was 20.2 months. Within each group, we rep
group member behaviors is consistently related to
missing data with the median value for the netw
to preserve the maximum amount of informati
(For instance, if four of five members of a grou
1 We did not frame a hypothesis regarding hindrance responded, we used median values for the netw
centralization and group performance paralleling Hy-for the fifth individual.) Median substitutions
pothesis 3b. Molm's (1994) concept of interdependence
resented 11.4 percent of the network data.
in group exchange does not justify such a prediction.
There is no comparable means of describing interdepen-
The average age of the respondents was
dence in the hindrance network. Negative exchanges do years. Men comprised 37.8 percent of the sam
not have the same potential for interdependence as pos- members. In terms of race, 63.6 percent were C
itive exchanges. That is, an interdependent network com- casian, 20.9 percent were African American,
prised of coordinated negative exchange relationships is percent were Hispanic, 2.7 percent were Asian,
difficult to conceptualize precisely because negative ex- the remainder were classified as "other." As for
changes preclude the possibility of mutual coordination. educational attainment, 11.1 percent had not com-

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
320 Academy of Management Journal April

pleted high school, 41.6 percent held a high school tion in the UCINET IV software package (Borgatti et
diploma, 9.3 percent had received technical train- al., 1992). First, the sum of the differences between
ing following high school graduation, 13.3 percent the largest individual centrality score and the
held an associate's degree, 21.2 percent held a scores of all the other individuals in the network
bachelor's degree, and 3.5 percent held a graduate was computed. This sum of the observed differ
degree. The respondents' average organizational ences in individual centrality scores was then d
tenure was 7.6 years. They performed a variety of vided by the maximum possible sum of difference
jobs, including clerical and production jobs and (Wasserman & Faust, 1994).
managerial jobs in marketing and customer sup- Individual job performance. We examined tw
port. aspects of job performance: in-role (performance o
required duties and responsibilities) and extra-role
Measures
(performance on discretionary behaviors that g
beyond the formal job description). Group leader
Networks and network centrality. The network rated each member's in-role and extra-role perfor-
surveys listed the names of each individualmance in a using 14 items scaled "strongly disagree"(
respondent's work group. Following the work to "strongly
of agree"(7). We used the 7-item scal
Burt (1992) and Ibarra (1993), we assessed advice developed by Williams and Anderson (1991) t
relations by asking the respondents two questions: measure in-role performance. Extra-role perfor
"Do you go to [name] for help or advice on work- mance was measured with the 6-item altruism scale
related matters?" and "Do you talk to [name] about developed by Smith, Organ, and Near (1983) and 1
confidential work-related matters?" A hindrance item ("This employee willingly gives his/her time
relation was elicited by asking, "Does [name] to help others who have been absent") from the
make it difficult for you to carry out your job re-Podsakoff, MacKenzie, Moorman, and Fetter (1990)
sponsibilities?" Because we were interested in thescale.
strength of the relationships among individuals On the basis of the results of a factor analysis
who knew each other, we elicited valued responses using varimax rotation, one of the in-role perfor-
to each network question using a seven-point scale, mance items ("engaged in activities that will di-
anchored by "not at all"(l), "some"(4), and "veryrectly affect his/her performance evaluation") and
much"(7). one of the extra-role performance items ("helps oth-
We computed normed in-degree centrality scores ers who have been absent") were deleted because of
for each individual (Borgatti, Everett, & Freeman, low or mixed loadings. After omission of these
1992) to allow for comparisons across groups ofitems, a second analysis resulted in two clean fac-
different sizes. In-degree centrality is a form of de-tors. The remaining 12 items comprised the in-role
gree centrality that counts only those relations with(a = .94) and extra-role (a = .90) performance
a focal individual reported by other group mem- scales, each consisting of 6 items each.
bers, and it thus does not suffer from the limitations Group performance. Group performance was as-
of self-reports, as does out-degree centrality. sessed by leaders using seven items developed for
We conducted a factor analysis using centrality this study on a scale ranging from "very poor"(l) to
scores for our three network measures. Two factors "outstanding"(7) (a = .99). Items addressed the
explaining 89 percent of the variance in the net- quality and quantity of work and the initiative,
work measures had eigenvalues greater than 1.0. cooperation, timeliness, and overall performance of
The two advice network centrality items showed groups.
high (greater than .90) loadings on the first factor,
with hindrance centrality loading on the second RESULTS
factor, thus demonstrating convergent and dis-
criminant validity. Descriptive statistics and correlations am
Network density. In binary network data, den- study variables are reported in Table 1 (individ
sity is the proportion of actual nominations among level variables) and Table 2 (group-level varia
the total possible number of nominations (Wasser- Table 3 presents the results of the regres
man & Faust, 1994). Because the relations in our analyses testing the hypothesized relationships
data were valued (measured on a scale from 1 to 7), tween network centrality and individual pe
we computed density as the sum of the actual re- mance. We controlled for organizational di
sponses divided by the total possible sum of re- ences by entering the organization dum
sponses. variables in step 1, followed by the centrality
Network centralization. Network centralization sures in step 2. In support of Hypothesis 1, ce
was computed following Freeman's (1979) defini- ity in the advice network was positively relate

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
2001 Sparrowe, Liden, Wayne, and Kraimer 321

TABLE 1
Descriptive Statistics and Correlations for Individual-Level Variablesa
Variable Mean s.d. 1 2 3 4 5 6 7 8

1. Organization 1 0.20 0.40

2. Organization 2 0.17 0.38 -.23**
3. Organization 3 0.09 0.29 -.16* -.11*
4. Organization 4 0.34 0.48 -.36** -.33** -.23**
5. Organization 5 0.19 0.40 -.25** -.23** -.15* -.36**
6. In-role performance 5.83 0.79 .17* -.10 .24** -.17* -.04
7. Extra-role performance 5.53 1.27 .31** -.21** .37** -.19** -.14* .65**
8. Advice centrality 3.52 1.04 .10 .19** .00 -.27** .03 .26** .22**
9. Hindrance centrality 1.49 0.66 .02 .13* -.06 -.01 -.09 -.30** -.27** -.09

an= 190.
* p < .05
** < .01

TABLE 2

Descriptive Statistics and Correlations for Group-Level Variablesa

Variable Mean s.d. 1 2 3 4 5 6 7 8

1. Organization 1 0.24 0.42

2. Organization 2 0.16 0.42 -.24
3. Organization 3 0.13 0.33 -.22 -.17
4. Organization 4 0.21 0.30 -.29* -.22 -.20
5. Organization 5 0.26 0.43 -.33* -.26 -.23 -.31*
6. Group performance 4.66 0.90 .26 -.21 .45** -.09 -.34*
7. Advice network density 1.87 0.09 .12 .48** .16 -.32* -.09 -.05
8. Advice network centralization 218.17 101.69 .32* .10 .05 -.11 -.25 -.15 .12
9. Hindrance network density 0.21 0.06 .09 .32* -.10 -.07 -.21 -.36* .28* .16

an = 38.
* p < .05
** P < .01

plained 20 percent of the variance in group per-

individual in-role and extra-role performance. Cen-
formance.
trality in the hindrance network was negatively
related to individual in-role and extra-role perfor-
mance, supporting Hypothesis 2. Taken together,
network centrality variables explained 13 percent DISCUSSION AND CONCLUSION
of the variance in in-role performance and 10 per-
Social Networks and Individual Performance
cent of the variance in extra-role performance.
Table 4 presents the results of regression analy-The results of this study offer support for the
hypothesized relationships between social net
ses testing the hypothesized relationships between
group network structure and group performance. work centrality and individual performance. In-
The organization dummy variables were entereddividuals
in who were central in their work groups
step 1, followed by the network measures in step advice
2. networks had higher levels of in-role and
Contrary to Hypothesis 3a, the parameter estimateextra-role performance than did individuals who
were not central players in such a network. Indi-
for density in the advice network was not statisti-
viduals who were central in a hindrance network
cally significant. Although the parameter estimate
for centralization in the advice network was in the had lower levels of both in-role and extra-role
hypothesized direction, its statistical significance performance. These results are important becau
was marginal (p = .06), and thus Hypothesis 3b they demonstrate that group members who a
found weak support. Supporting Hypothesis 4, the central to group advice-sharing are rated mor
parameter estimate for hindrance network density positively on individual performance. That th
was negative and statistically significant. Taken relationship was found with respect to both i
together, network density and centralization ex- role and extra-role performance contributed t

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
322 Academy of Management Journal April

TABLE 3
Results of Regression Analysis for Network Centrality and Individual Performancea
In-Role Performance Extra-Role Performance

Variable b s.e. b s.e. b s.e. b s.e.

Step 1: Control variables

Intercept 5.64 0.09 5.58 0.22 5.19 0.14 5.14 0.33
Organization 1 .80** 0.21 0.69** 0.19 1.12** 0.22 0.99** 0.22
Organization 2 0.45** 0.16 0.36* 0.15 -0.23 0.23 -0.33 0.23
Organization 3 0.01 0.16 -0.06 0.16 1.81** 0.30 1.67** 0.28
Organization 4 0.12 0.16 0.01 0.15 -0.02 0.23 -0.18 0.21

R2 .10 .27
F 5.33** 17.09**

Step 2: Centrality
Advice network 0.18** 0.05 0.23** 0.08
Hindrance network -0.33** 0.08 -0.46** 0.11

AR2 .13 .10

AF 15.33** 13.84**
Overall adjusted R2 .23 .35
Overall F 9.22** 17.60**

an = 190.
* p < .05
**p < .01

TABLE 4

Results of Regression Analysis for Group per job descriptions. Additionally, individua
Network Structure and Performancea performance traditionally has been evaluate
the basis of behaviors that individuals enga
Controls Network Structure
apart from coworkers. However, with the tr
toward greater use of teams in organizations
Variable b s.e. b s.e.
extent to which team members are involved in
behaviors that assist coworkers has become a sa-
Step 1: Control variables
Intercept 4.20 0.22 5.22 0.63 lient dimension of job performance. Indeed, o
Organization 1 0.85* 0.33 1.19* 0.31 results showed that individuals who were iden-
Organization 2 0.06 0.37 0.45 0.38 tified by coworkers as being active in providin
Organization 3 1.40** 0.39 1.51** 0.34
advice were rated more favorably by leaders th
Organization 4 0.32 0.34 0.46 0.30
individuals who were not mentioned as often b
R2 .35
F 4.48**
coworkers as providing such information.
Given the sparseness of research on negative
Step 2: Network
exchanges in structure
work settings, we found it notewor-
Advice network density 0.17 0.28
Hindrance network density -4.97** 1.74 thy that individuals who were identified by c
Advice network centralization -0.02+ 0.00 workers as hindering the work of others wer
AR2 .20 rated by managers as relatively lower on in-ro
AF 4.35*
and extra-role performance. It appears that, ju
Overall adjusted R2 .44
Overall F 5.21 *
as there are networks of individuals engaged
providing useful advice and support, there ar
an = 38. also networks of individuals who engage in be
tp = .06 haviors that hinder others from completing th
* p < .05 tasks. However, neither in the current study n
**p < .01
in the handful of other investigations examinin
negative exchange relationships have specifi
hindering behaviors been identified (Labianca
the research on organizational citizenship. It ap-al., 1998). Future research should examine the
pears that some of the advice provided by thosecauses of hindrance behavior and illuminate the
who are central to the flow of information in ways in which individuals in central positions
work groups extends beyond what is expected hindrance
as networks slow the work progress o

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
2001 Sparrowe, Liden, Wayne, and Kraimer 323

their coworkers. For example, negative affectivity hindrance network contain individuals who truly
could induce some individuals to intentionally restrict the progress of the group toward comple-
withhold effort or try to sabotage the work of tion of group tasks, or do these individuals present
coworkers (Robinson & Bennett, 1995). positions that conflict with those in the majority,
resulting in their being perceived by others as hin-
dering the group? A second area for future research
Social Networks and Group Performance
is to examine what causes negative exchanges to
The results did not support the hypothesis that develop. The results would have important impli-
advice network density is positively related to cations for how to reduce negative exchanges
group performance. This hypothesis might have within groups (that is, reduce hindrance network
been supported had the measure focused specifi- density), and thus increase group performance.
cally on advice in solving work-related problems
and assistance with work assignments rather than Limitations
general information about work, which may or may
not have been relevant to group performance. We This study has several potential limitations. The
found marginal support (p = .06) for the hypothe- first concerns the validity of our performance mea-
sized negative relation between advice network sures. Although common method variance was not
centralization and group performance. An interest- an issue because the social network and perfor-
ing paradox emerges when one compares these re- mance ratings data had different sources, aspects of
sults to those for individual performance. Specifi- the social context may have biased the subjective
cally, centrality in the advice network was evaluation of performance. Similarly, there may
positively associated with individual performance, have been a leniency bias in the group performance
whereas at the group level, centralization was neg- ratings, given that a group's effectiveness reflects
atively associated with group performance. These its leader's own performance. Thus, an alternative
results parallel those found in pioneering work on interpretation of our findings is that informal net-
group structure and process (Shaw, 1964). Studies work structure is related to supervisors' assess-
in which group structure was manipulated showed ments of individual and group performance, rather
that central positions in groups related positively to than to actual performance.
individual performance. On the other hand, cen- Second, our theoretical perspective implies that
tralization at the group level was positively related network structure precedes individual perfor-
to group performance only for simple tasks. On mance. However, it is possible that the relationship
complex tasks, centralization was not associated between individual performance and network
with group performance. In the organizations in- structure is reciprocal or that it is the reverse of
cluded in our sample, tasks were complex, suggest- what we have theorized. For example, it is possible
ing that our group-level results parallel those of the that coworkers seek out high performers as sources
early experimental studies. This pattern of findings of advice, thus enhancing high performers' central
indicates that group performance is more than the positions within informal networks.
sum of each member's individual performance Third, this study employed a single-item mea-
(Molm, 1994). sure of hindrance networks. Although use of single-
We found support for the negative relationship item measures of negative network relations is
between hindrance network density and group per- common in previous research involving large net-
formance, suggesting that uncooperative behaviors works (e.g., Baldwin et al., 1997; LaBianca et al.,
among group members are just as important as co- 1998), small group research does not face the same
operative behaviors in influencing group perfor- practical limitations as large group research. Al-
mance-perhaps even more so. Consistent with though sociometric techniques were used with
Brass and LaBianca's (1999: 324) argument in favor small groups in experimental research conducted
of examining both sides of the "social ledger," hin- during the 1950s (Shaw, 1964), it should be noted
drance density was negatively associated with that contemporary research typically studies net-
group performance, but the relationship between works larger in size than the groups studied here.
advice network density and performance was not
statistically significant. Group performance suffers Conclusion
to the extent that coworkers withhold resources or
avoid other group members. Future research in this In summary, this study contributes to severa
area is warranted, especially research employing streams of research. First, it adds to the social net-
designs that permit differentiation among types of work and individual performance literatures b
hindrance networks. A relevant question is, does a demonstrating a relationship between network

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
324 Academy of Management Journal April

structure and both in-role and extra-role perfor- Dunnette & L. M. Hough (Eds.), Handbook of indus-
mance in a field setting. These results also suggest trial and organizational psychology: 269-313. Palo
that the findings of laboratory experiments on Alto, CA: Consulting Psychologists Press.
group structure from the 1950s (Shaw, 1964) gen- Ibarra, H. 1993. Network centrality, power, and innova-
eralize to intact work groups in contemporary or- tion involvement: Determinants of technical and ad-

ganizational settings. In addition, the findings en- ministrative roles. Academy of Management Jour-
hance the social network literature by revealing nal, 36: 471-501.
that rarely studied hindrance networks are nega- Krackhardt, D. & Porter, L. W. 1986. The snowball effect:
tively related to individual in-role and extra-role Turnover embedded in communication networks.
performance as well as to group performance. Journal of Applied Psychology, 71: 50-55.
Taken together, the results of the current investiga- Labianca, G., Brass, D. J., & Gray, B. 1998. Social net-
tion provide a strong rationale for the integration of works and perceptions of intergroup conflict: The
modern social network analysis and the social psy- role of negative relationships and third parties.
chology of groups. Academy of Management Journal, 41: 55-67.
Larson, J. R., Christensen, C., Franz, T. M., & Abbott, A. S
1998. Diagnosing groups: The pooling, management,
REFERENCES and impact of shared and unshared case information
Baldwin, T. T., Bedell, M. D., & Johnson, J. L. 1997. Thein team-based medical decision making. Journal of
social fabric of a team-based M.B.A. program: Net-
Personality and Social Psychology, 75: 93-108.
Molm, L. D. 1994. Dependence and risk: Transforming
work effects on student satisfaction and perfor-
mance. Academy of Management Journal, 40: and structure of social exchange. Social Psychology
1369-1397. Quarterly, 57: 163-176.
Pagel, M. D., Erdly, W. W., & Becker, J. 1987. Social
Borgatti, S. P., Everett, M. G., & Freeman, L. C. 1992.
UCINET IV version 1.0 reference manual. Colum-networks: We get by with (and in spite of) a little
bia, SC: Analytic Technologies. help from our friends. Journal of Personality and
Borman, W. C., Hanson, M. A., Oppler, S. H., Pulakos,Social Psychology, 53: 794-804.
E. D., & White, L. A. 1993. Role of early supervisory
Podolny, J. M., & Baron, J. N. 1997. Resources and rela-
experience in supervisor performance. Journal of tionships: Social networks and mobility in the work-
Applied Psychology, 78: 443-449. place. American Sociological Review, 62: 673-693.
Podsakoff, P. M., MacKenzie, S. B., Moorman, R. H., &
Brass, D. J. 1981. Structural relationships, job character-
istics, and worker satisfaction and performance.Fetter, R. 1990. Transformational leader behaviors
Administrative Science Quarterly, 26: 331-348. and their effects on followers' trust in leader, satis-
Brass, D. J. 1984. Being in the right place: A structuralfaction, and organizational citizenship behaviors.
analysis of individual influence in an organization. Leadership Quarterly, 1: 107-142.
Administrative Science Quarterly, 29: 518-539. Robinson, S. L., & Bennett, R. J. 1995. A typology of
deviant workplace behaviors: A multidimensional
Brass, D. J., & Labianca, G. 1999, Social capital, social
liabilities, and management. In R. T. Leenders & S.scaling study. Academy of Management Journal,
Gabbay (Eds.), Corporate social capital and liabil-38: 555-572.
ity: 99-999. Boston: Kluwer Academic Publishers.
Sahlins, M. 1972. Stone age economics. New York:
Burt, R. S. 1992. Structural holes: The social structure Aldine de Gruyter.
of competition. Cambridge, MA: Harvard University Shaw, M. E. 1964. Communication networks. In L.
Press.
Berkowitz (Ed.), Advances in experimental social
Cook, K. S., & Emerson, R. M. 1978. Power, equity and psychology: 111-147. New York: Academic Press.
commitment in exchange networks. American So- Smith, C. A., Organ, D. P., & Near, J. P. 1983. Organiza-
ciological Review, 43: 721-739. tional citizenship behavior: Its nature and anteced-
Freeman, L. C. 1979. Centrality in social networks, Con- ents. Journal of Applied Psychology, 68: 653-663.
ceptual clarification. Social Networks, 1: 215-239. Sparrowe, R. T., & Liden, R. C. 1997. Process and struc-
Friedkin, N. E. 1993. Structural bases of interpersonal ture in leader-member exchange. Academy of Man-
influence in groups: A longitudinal case study. agement Review, 22: 522-552.
American Sociological Review, 58: 861-872. Taylor, S. E. 1991. Asymmetrical effects of positive and
Granovetter, M. 1985. Economic action and social struc- negative events: The mobilization-minimization
ture: The problem of embeddedness. American hypothesis. Psychological Bulletin, 110: 67-85.
Journal of Sociology, 91: 481-510. Wagner, J. A. 1995. Studies of individualism-collectiv-
Guzzo, R. A., & Shea, G. P. 1992. Group performance and ism: Effects on cooperation in groups. Academy of
intergroup relations in organizations. In M. D. Management Journal, 38: 152-172.

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms
2001 Sparrowe, Liden, Wayne, and Kraimer 325

Wasserman, S., & Faust, K. 1994. Social network anal- Chicago. His research focuses on interpersonal processes
ysis: Methods and applications. Cambridge, En- within the context of such topics as leadership, groups,
gland: Cambridge University Press. career progression, and employment interviews.
Williams, L. J., & Anderson, S. E. 1991. Job satisfaction Sandy J. Wayne (Ph.D., Texas A & M University) is an
and organizational commitment as predictors of or- associate professor of management at the University
ganizational citizenship behavior. Journal of Man- of Illinois at Chicago and the director of the Center
agement, 17: 601-617.
for Human Resource Management at the University of
Raymond T. Sparrowe (Ph.D., University of Illinois at Illinois. Her research interests include leadership,
Chicago) is an assistant professor of management and perceived organizational support, working relationships,
labor relations at the James J. Nance College of Business social influence processes, and groups.
Administration, Cleveland State University. In fall 2001, Maria L. Kraimer (Ph.D., University of Illinois at
he will be an assistant professor of organizational behav-Chicago) is an assistant professor of management and
ior at the Olin School of Business at Washington Univer-
labor relations at the James J. Nance College of Busi-
sity, Saint Louis. His research interests focus on the
ness Administration, Cleveland State University. She
interplay of interpersonal processes and social structure
in relation to work outcomes.
will join the University of Illinois at Chicago as an
assistant professor in fall 2001. Her research interests
Robert C. Liden (Ph.D., University of Cincinnati) is a include career-related issues, employee adjustment,
professor of management at the University of Illinois at and the employee-employer relationship.

This content downloaded from 149.157.1.168 on Thu, 26 Apr 2018 18:40:45 UTC
All use subject to http://about.jstor.org/terms