|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 TEAM NETWORK MULTIPLEXITY, SYNERGY, AND PERFORMANCE By EEAN R. CRAWFORD A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011
2 2011 Eean R. Crawford
3 To my mom and dad who made me what I am, to my wife who makes me even better, and to my kids who make me want to be the best
4 ACKNOWLEDGMENTS Above all, I thank my dear wife Astyn, for patiently persisting with me through every glorious high and every depressing low and my sweet children, for giving me hugs and kisses every day before I would leave and when I would come home. Second, I thank Jeff LePine for his continual collaboration, feedback, and encouragement. His mentoring has been the best I could have ever hoped for I am also grateful for my committee members Jason Colquitt, Timothy Judge, and Chris McCarty; how fortunate I am to have had such giants among men to guide me along the way. Third, Id like to thank my parents for teaching me from a young age to work and study hard. They instilled in me an intense curiosity for the world and the work ethi c to do something about it. Fourth, I want to thank every one of the MBA students who took the time to participate in my study. I held back tears of joy every time another completed survey was returned. Finally, I am incredibly indebted to my fellow students: Jessica Methot, who inspired me to learn about networks and remains my constant example; Jessica Rodell who has always cared, always listened, and always encouraged; Lauren Simon, who has commiserated with me since day one; Jen Knippen, who helped me both to focus and to laugh; and Brooke Buckman, invaluable officemate and PivotTable queen, who se second set of eyes and ears helped again and again. You are the best team I could ever have been on.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 8 ABSTRACT ..................................................................................................................... 9 CHAPTER 1 INTRODUCTION .................................................................................................... 11 2 LITERATU RE REVIEW .......................................................................................... 14 Teams and Team Member Interaction .................................................................... 14 Traditional Approaches to Team Member Interaction ....................................... 16 Social Network Analysis As a Useful Alternative .............................................. 19 Social Network Analysis .......................................................................................... 22 Social Network Terminology ............................................................................. 22 Types of Ties .................................................................................................... 24 Instrumental ties ......................................................................................... 25 Expressive ties ........................................................................................... 25 Multiplex ties .............................................................................................. 26 Network Density ............................................................................................... 28 Density of Multi plexity ....................................................................................... 30 Teams As Social Networks ..................................................................................... 30 Research on Instrumental Team Networks ...................................................... 31 Research on Positive Expressive Team Networks ........................................... 36 Research on Negative Expressive Team Networks ......................................... 40 Research including Instrumental and Expressive Team Networks ................... 42 Research on Multiplex Team Networks ............................................................ 44 Limitations of Existing Team and Network Literatures ...................................... 45 3 THEORY AND HYPOTHESES ............................................................................... 49 Team Synergy and Performance ............................................................................ 51 Team Process Gains ........................................................................................ 53 Team Process Losses ...................................................................................... 55 Team Motivat ion ............................................................................................... 57 The Role of Team Chemistry in Relation to Synergy and Performance .................. 61 Positive Multiplex Density (Structural Cohesion) .............................................. 68 Negative Multiplex Density (Structural Repulsion) ............................................ 72
6 4 METHODS .............................................................................................................. 78 Sample and Procedure ........................................................................................... 78 Measures ................................................................................................................ 81 Network Relations ............................................................................................ 81 Instrumental networks ................................................................................ 81 Positive e xpressive networks ..................................................................... 82 Negative expressive networks ................................................................... 82 Positive and negative multiplex networks .................................................. 83 Structural cohesion and repulsion .............................................................. 85 Team Perceptions ............................................................................................ 86 Teamwork processes ................................................................................. 86 Team engagement ..................................................................................... 87 Team maintenance .................................................................................... 88 Team Performance ........................................................................................... 89 Control Variables .............................................................................................. 89 5 RESULTS ............................................................................................................... 92 Network Characteristics .......................................................................................... 92 Network Measurement Model ................................................................................. 94 Mediator Measurement Model ................................................................................ 96 Descri ptive Statistics and Correlations .................................................................... 96 Analysis of Independence ....................................................................................... 98 Structural Models .................................................................................................... 98 Tests of Hypotheses for Team Synergy and Performance ............................. 1 00 Tests of Hypotheses for Structural Cohesion to Team Synergy ..................... 101 Tests of Hypotheses for Structural Repulsion t o Team Synergy .................... 102 Tests of Indirect Effects .................................................................................. 102 6 DISCUSSION ....................................................................................................... 112 Summary of Findings ............................................................................................ 112 Theoretical Contributions ...................................................................................... 113 Positive and Negative Multiplex Relationships ............................................... 113 Mediating Role of Teamwork, Engagement, and Maintenance ...................... 115 Implications for Team Effectiveness and Social Capital Theories .................. 116 Practical Implications ............................................................................................ 118 Limitations ............................................................................................................. 119 Future Research ................................................................................................... 120 Conclusion ............................................................................................................ 124 LIST OF REFERENCES ............................................................................................. 126 BIOGRAPHICAL SKETCH .......................................................................................... 139
7 LIST OF TABLES Table page 5 1 Descriptive statistics and correlations for networks .......................................... 105 5 2 Network measurement model ........................................................................... 105 5 3 Mediator measurement model .......................................................................... 106 5 4 Study items and factor loadings ........................................................................ 107 5 5 Descriptive statistics and correlations for study variables ................................. 109 5 6 Structural models .............................................................................................. 110 5 7 Tests of indirect effects ..................................................................................... 110
8 LIST OF FIGURES Figure pag e 3 1 Model of hypothesized relationships ................................................................... 77 5 1 Structural model relationships .......................................................................... 111
9 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TEAM NETWORK MULTIPLEXITY, SYNERGY, AND PERFORMANCE By Eean R. Crawford May 2011 Chair: Jeffery A. LePine Major: Business Administration In this dissertation I examine the implications of team social structure for work team functioning and performance. I apply a social network perspective to examine the balance of team structural coh esion and structural repulsion, where structural cohesion is indicated by the density of overlapping instrumental and positive expressive ties, and structural repulsion is indicated by the density of overlapping instrumental and negative expressive ties. I then examine the implications of these social structures for team performance through their associations with synergistic team process gains indicated by teamwork team motivation indicated by team engagement and team process losses indicated by team rel ationship maintenance. I test the predictions of this model with a sample of 109 student teams operating in the Master of Business Administration (MBA) programs of a large southeastern university. Overall, findings suggest that the structure of social relationships in the team matters for team effectiveness. Structural cohesion was beneficial for team performance because of its association with increased teamwork. It was also associated with increased team engagement and reduced team maintenance, bit it did not exhibit indirect relationships with team performance through these mechanisms. On the other hand, structural repulsion was detrimental for team
10 performance because of its association with reduced teamwork. It was also associated with decreased team engagement and increased team maintenance, but it did not exhibit indirect relationships with team performance through these mechanisms.
11 CHAPTER 1 INTRODUCTION Team based structures have become an increasingly important unit of work organization (Devine, Clayton, Philips, Dunford, & Melner, 1999). Recent estimates suggest that 82 percent of companies with 100 employees or more use some form of work teams, and of those companies using work teams, over half of employees within the company are members of teams (Gordon, 1992). Scholars argue that groups and teams are the central building blocks for getting work done in organizations (Goodman, Ravlin, & Schminke, 1987) As such scholars have focused a great deal of attention to understanding the factors contributing to the effective structure, function, and performance of teams (Mathieu, Maynard, Rapp, & Gilson, 2008). Researchers have long been interested in the implic ations of social interaction processes for team effectiveness (Bales, 1953; Bion, 1961; Shaw, 1964; Guzzo & Shea, 1992). While scholars are converging on the perspective that teams are complex, dynamic systems (Ilgen et al., 2005; McGrath, 1997; McGrath, A rrow, & Berdahl, 2000), and are increasingly recognizing that team member interaction occurs in discontinuous, complex, and nonlinear patterns (Kozlowski, Gully, Nason, & Smith, 1999; Kozlowski & Klein, 2000; Mathieu et al., 2008; Stewart, Fulmer, & Barri ck, 2005) current conceptions of team social interaction have yet to account for this complexity. Researchers have long recognized that team social interaction may be categorized as either instrumental concerning the exchange of work and task related information, or expressive concerning the exchange of affect, support, or friendship (Bales, 1953; Ibarra & Andrews, 1993; Lincoln & Miller, 1979; Stewart et al., 2005), but scholars disagree on the relative importance assigned to instrumental and expressive
12 member interactions as predictors of team effectiveness (Guzzo & Shea, 1992). Some schools of thought heavily emphasize the importance of expressive interactions in models of effectiveness (e.g., Bion 1961; Schutz, 1967), while others give more weight to instrumental interactions (e.g., Hackman, 1987), while yet others, such as sociotechnical systems theorists, emphasize both (e.g., Kolodny & Kiggindu, 1980; Rousseau, 1977; Trist, 1981). Though differ ent schools of thought have emphasized the relative importance of either instrumental, expressive, or both types of interaction, each of these perspectives has failed to consider the simultaneous overlap of these types of interaction among members in the t eam. In addition, when considering expressive interactions, researchers tend to focus solely on positive manifestations of expressive interaction (e.g., friendship, liking, support), while negative manifestations of expressive social interaction (e.g., adv ersarial, disliking, d istrusting) receive far less attention (Labianca & Brass, 2006). Thus, while researchers recognize that team interaction occurs in both instrumental and expressive forms, current conceptions of social interaction are underdeveloped b ecause they do not consider the implications of the overlap in these types of interaction, nor do they consider the importance of both positive and negative forms of expressive interaction. Thus, the purpose of my study is to develop a more replete concept ion of team social interaction, which I call team chemistry based on the intersection of instrumental and expressive team interactions in both their positive and negative forms. I use the concept of multiplexity in social network analysis to model the ove rlap of instrumental and positive expressive ties, and instrumental and negative expressive ties, respectively, in the team. I then propose that team chemistry consists of
13 team structural cohesion defined as the density of positive multiplex ties in the team, and team structural repulsion defined as the density of negative multiplex ties in the team. I propose to test the implications of this model of team chemistry for team effectiveness in a large sample of teams operating in the Master of Business Admini stration (MBA) programs of a large southeastern university.
14 CHAPTER 2 LITERATURE REVIEW Teams and Team Member Interaction A team, as defined by Hackman (1987) and Guzzo and Shea (1992), is a social system with the following properties: It is an enti ty with boundaries perceived by its members and nonmembers familiar with it; it has one or more tasks to perform resulting in discernable and potentially measurable group products; it has a differentiation of roles and some degree of interdependence; and i t operates within an organizational context. Hackman (1987) notes that this is actually a fairly inclusive statement, with the definition applying, for example, to a group of executives deciding where to locate a new plant, a team of manufacturing workers assembling a product, a group of students writing a case assigned by their instructor, a healthcare team tending to the needs of a set of patients, or a group of economists analyzing the budgetary implications of a proposed new public policy. Hackman does exclude from this definition other sets of people commonly referred to as groups, such as social groups (no task), reference groups (not an intact social system), groups of employees reporting to the same manager but who have their own, individual tasks to perform (no group task), and freestanding groups (no organizational context). The dominant paradigm for studying team effectiveness is an input process output (I P O) model, first put forth by McGrath (1964) and later expanded upon by Steiner (1972), McGrath (1984), and Hackman (1987). This model emphasizes that inputs consisting of individual (e.g., member characteristics and abilities), team (nature of the task and group leadership), and organizational (contextual and environmental influences) factor s combine to influence team processes, which refer to team member
15 interactions to transform inputs into outcomes (Marks, Mathieu, & Zaccaro, 2001). Outcomes are the results achieved or products created by the team and generally include the quality or quant ity of output as well as team member affective reactions such as team satisfaction and viability (Hackman, 1987; Mathieu et al., 2008). The major proposition underlying this model is that the key to understanding why some groups perform better than others lies in the varying level and quality of team member interactions, and that better team interactions are more likely to emerge from optimal arrangements of team inputs (Hackman, 1987). Researchers have since updated this I P O model to consider a broader array of mediators including cognitive, motivational, and affective states of teams, as well as to account for the episodic and cyclical, as opposed to linear, nature of team performance. Ilgen, Hollenbeck, Johnson, and Jundt (2005) proposed an input mediat or output input (IMOI) model to account for these concerns. They suggested that substituting M for P reflects the broader range of variables that are important meditational influences on team effectiveness. Adding the extra I at the end recognizes the notion of cyclical performance feedback whereby the outcomes at the end of one performance episode can become inputs for the next. Finally, they eliminate the hyphens to suggest that causal linkages may not be linear or additive, but rather nonlinear or conditional, recognizing that researchers have documented and should continue to investigate interaction effects between inputs and processes (I x P), between processes (P x P), and between processes and affective/motivational emergent states (P x ES). I n either the case of I P O models or IMOI models, the nature of the mediators between team inputs and outcomes is derived from the interactions between team
16 members. Katz and Kahn (1978, p. 187) argued that the substance of collective organization is the p attern of interaction among its members (Katz & Kahn, 1978: 187). Collective structures emerge and persist through the interactions of members of the collective. Absent collective interaction, a collective structure simply does not exist (Morgeson & Hoffma n, 1999). Thus, the basic unit of analysis of a collective phenomenon is not the individual, nor the team, nor the dyad, but the relationship or interaction between the individuals in the collective (Brass, 1995). Traditional Approaches to Team Member Inte raction Because collective structures originate in the interaction of individual level members, the nature of the emergence of team level constructs must be explicitly specified by theory. Kozlowski and Klein (2000) distinguished unit level constructs that emerge from the characteristics, behaviors, or cognitions of unit members as having either shared or configural properties. Shared unit constructs describe characteristics that are common to or shared by the members of the unit. Shared unit properties emerge as a consensual, collective aspect of the unit as a whole, and are based on composition models of emergence. Composition models assume isomorphism between manifestations of constructs at different levels, and rely on withinunit consensus (agreement) or consistency (reliability) to justify composition of the unit level construct. Configural unit constructs capture the array, pattern, or configuration of individual member characteristics or interactions. Unlike shared unit properties, configural unit pro perties are not assumed to converge or coalesce among members, and instead are based on compilation m odels of emergence. Compilation models do not assume isomorphism and convergence but assume instead discontinuity and complex nonlinear emergence of const ructs between levels. As a result, compilation models are not
17 concerned with agreement or consistency for aggregation, but rather rely on depicting patterns, distribution, and variability among member contributions to create unit level constructs. If the structure of a team construct is defined as the series or pattern of interactions between team members (Morgeson & Hoffman, 1999), then compilation models of emergence that provide the means for examining the configuration or pattern of interaction are a pr erequisite for depicting team structure. Interestingly, team constructs have most frequently been conceptualized and measured by adopting composition models of emergence. Chan (1998) proposed a typology of composition models consisting of additive, direct consensus, or referent shift models in which the team level variable is created by summing or averaging individual ratings (either by team members or subject matter experts) as to the extent to which a given team phenomenon occurs generally in the team (e. g., Mathieu et al., 2006; Marks, DeChurch, Mathieu, Panzer, & Alonso, 2005; Hirschfeld, Jordan, Feild, Giles, & Armenakis, 2006; Stewart & Barrick, 2000) The additive composition model creates a straightforward sum or average of individual team member sel f referent perceptions, such as the extent to which each member perceives they themselves perform a given process or fill a specific role in the team. The direct consensus composition model also averages member self ratings, but adds the requirement that t here be sufficient agreement or consistency within the individual ratings to justify aggregation to the team level. Statistics used to justify agreement and consistency include withingroup agreement ( rwg, James et al., 1984) and intraclass correlations (I CC and ICC, Bliese, 2000). Referent shift composition models also average individual member ratings and require sufficient withingroup agreement and consistency, but the referent
18 of individual ratings prior to aggregation is shifted from the members rating themselves to their rating the team or team members generally. The referent of the ratings is changed from the individual to the team creating a collective construct conceptually distinct from the original form. Whether composed by additive, direct consensus, or referent shift models, the measure of the collective construct results in a single index of the average extent to which a specific type of interaction is perceived to occur among team members. Despite the likelihood of team interactions occurring in nonuniform, complex patterns, team constructs rarely are described using compilation models of emergence1 (Kozlowski & Klein, 2000). Composition models sufficiently explain team interactions only if the assumption holds that all interactions, communications, and contributions occur uniformly among all members of a team. However, it is far more likely that members exert di sproportionate influences on team functioning by virtue of the positions that they occupy, their relative status, and so forth (Mathieu et al., 2008). For example, the construct of cohesion, originally defined by (Festinger, 1950: 274) as the resultant fo rces which are acting on the members to stay in a group consisting of interpersonal attraction, liking the group task, and group prestige or pride (Beal, Cohen, Burke, and McLendon, 2003; Mullen & Cooper, 1994), is configural in nature because it is based on interpersonal evaluations and relative differences in intergroup prestige. Yet cohesion is most often measured instead in terms of composition models of shared perceptions and average indices (e.g., Barrick, Stewart, Neubert, & Mount, 1998; Elfenbein & OReilly, 1 Recent studies adopting compilation models of team constructs include Stewart, Fulmer, & Barricks (2005) adoption of variance and skewness measures of team task and social roles; and LePines (2003) modeling of team role structure as recurring patterns of team messaging in an experimental task.
19 2007; OReilly, Caldwell, & Barnett, 1989; Stokes, 1983). For example, Barrick et al. (1998, p. 383) measured cohesion by indexing members shared agreement with statements such as Team members consistently help each other on the job and The members of this team got along well with each other. However, this approach is unable to capture whether all team members were uniformly liked, whether there were only one or two members who were especially liked (or disliked, for that matter), or whether there were subgroups that liked their own members, but not those of the other subgroups. Such is the case as well with helping members may have uniformly helped others, or a central few helped all others, or subgroups helped only their own members. Likewi se, the traditional approach is unable to capture the overlap in helping and liking (or disliking) at the relationship level among team members. In other words, certain individuals may consistently help and like each other but not share those types of over lapping relationships with other members of the team. Or it is even possible that, because of team task constraints, some members of the team must help another whom they simultaneously dislike a phenomenon that has important implications for the functioni ng of the team In each case, the traditional composition model approach with its emphasis on emergence through uniformity fails to provide information about these complex interactions, and thus traditional composition models of emergence are highly restri cted in their usefulness for conceptualizing and modeling nonlinear team interactions. Social Network Analysis As a Useful Alternative These limitations require a shift in thinking in order to advance our understanding of team interaction. Indeed, Ilgen and colleagues have noted that the empirical research on teams in organizational contexts is moving in the direction of increased
20 complexity, but this work still has a way to go to match developments in the conceptual domain (Ilgen et al., 2005, p. 519), and McGrath (1997, p. 16) has suggested that researchers need to borrow and invent new ways of thinking about groups and new tools for doing research on them conceptions and tools that allow us to seriously (not just rhetorically) conceptualize and study groups as complex, adaptive, dynamic systems. I propose the use of social network analysis as an ideal means to accomplish this purpose in the study of teams generally and for the study of team member interaction specifically. Because social network analysis is directly concerned with the pattern, content, and variation of the connections, or ties among actors, or nodes in a collective (Hanneman & Riddle, 2005), it represents a useful perspective on which to base compilation models of collective constr uct emergence that capture the more complex patterns of interactions in teams. As a simple example in the case of cohesion, a network approach to modeling cohesion, in contrast to the traditional approach (e.g., Barrick et al., 1998), would begin by asking members directly which of the other team members they liked (or which of the other members they helped). Increasing cohesion would be modeled by increasing density in the network of liking or helping. Further, overlap between individuals of multiple types of cohesion e.g., individuals who indicate that they both like and help each other is easily modeled through the network concept of multiplexity, which accounts for the simultaneous overlap of multiple types of relations between individuals. Increasing cohesion in this sense would be modeled by increasing density in the multiplex network of individuals that both like and help each other. This accounts for multiple types of cohesion at the relationship level, and team level
21 cohesion emerges as the configuration or pattern of these multiplex relationships in the team. Thus, a dopting a network perspective allows researchers to model the structural patterns of cohesion based on a compilation model of emergence rather than shared perceptions about cohesion based on traditional composition models of emergence. Viewing team interaction through a social network lens not only increases the clarity of our observation through improved methods and operationalizations, but more importantly it introduces altogether new concepts that are highly useful to increase our understanding and usage of compilation models of collective construct emergence. A network perspective enriches teams theory by allowing scholars to move beyond typical traditional composition models to examine the structural patterns of team interactions in a configural sense. This approach can also enrich network research by expanding scholars focus beyond the structural properties of networks to pay [more] attention to the kinds of relationships that bind a network together (Baker, Cross, & Wooten, 2003, p. 328329). Balkundi and Harrison (2006) showed through metaanalysis that social network structures have important implications for team performance and team viability. They assumed this was so because the pattern of ties facilitates or constrains the flow of interpersonal resources and information between members of the team, yet recognized that research is needed to pinpoint the actual information and resources that flow through network ties to influence these team outcomes. Below I provide a brief overview of social network analysis terminology, and in a subsequent section I review research on teams from a social network perspective.
22 Social Network Analysis2 Social network analysis seeks to model relati onships or interactions between sets of actors to depict the structure of a collective and how this structure impacts the functioning of the collective as well as influences actors within this collective (Wasserman & Faust, 1994). Because network structures are implicitly based on compilation models of emergence, they are well suited to describe configural concepts that depict the pattern or array of interactions or relationships within a team (Kozlowski & Klein, 2000). Thus, the purpose of social network analysis as applied to team interactions is to (a) observe team phenomena at the basic unit of analysis their interactions, (b) model the emergence of structure from the ground up as networks or patterns of interaction, and (c) examine how these team structures impact the functioning of the team as well as the individual members within it. I briefly introduce social network terminology and concepts below. Social Network Terminology The network perspective allows for new leverage in describing the configural approach to team phenomena by providing terminology and precise formal definitions for aspects of the structure of team interaction. The term social network refers to a set of actors and the ties or relationships among them. As I have described earlier, structure refers to patterns of relationships or interactions among units or individuals. Structural variables refer to quantities that measure aspects of structure. I first describe basic terminology adopted by social network analysts to describe the buil ding blocks of 2 The exposition of concepts in this section draw heavily from Wasserman & Faust (1994); Scott (2000); Hanneman & Riddle (2005); and Brass (1995). Other statements requiring specific citation have been referenced within the section accordingly.
23 social networks; and then, based on these definitions, I discuss the structural variables relevant to describing the team structures pertaining to this study. Actors (also called nodes or vertexes ) represent any discrete entity in the network, whether a person, an organization, a place, or a collective social unit. Applied to the present research, we view actors as individual members of a group or team. The social network perspective is also able to incorporate actor attributes in the traditional social science sense of actor characteristics and traits, but attributes of actors are not required for social network analysis since the focus is on the patterns of relationships among them. Ties represent the linkages between actors, which are formed by some type of connection, relationship, or interaction occurring between actors. Some examples (among a myriad) of ties include those created from one actors evaluation of another; by transfers of material r esources between actors; by actors joint associations or affiliations with some social event or other entity; by actors behavioral interaction and communication; by formally specified relationships between actors; or by biological, physical, or contractual relationships between actors. It is important to note that a distinct tie represents each different type of connection. Ties may be directed, in which one actor is the source (sender) and another actor is the destination (receiver) of the tie; or undirected in which the tie has neither a sender nor receiver but rather constitutes joint presence. Ties may be dichotomous the tie is either present or absent, or valueda range of numbers are used to give an indication of the strength, intensity, or frequency of the tie. A graph is a visual depiction of a network in which actors are represented with circles (usually), and ties between actors are represented with lines
24 co nnecting the circles. Actor attributes may be incorporated into the graph by modifying nodes with different colors or shapes (to represent categorical differences) or different sizes or shades (to represent continuous differences). Directed ties are repres ented by lines with arrowheads pointing to the actor receiving the tie. A directed tie that is reciprocatedboth sender and receiver nominate each other is usually depicted by a single line with doubleheaded arrows rather than by two separate lines with opposing arrows. A dyad consists of a pair of actors and the (possible) ties between them. A triad consists of a trio of actors and the (possible) ties among them. A subgroup of actors is any subset of actors and all ties among them. Wasserman and Faust (1994, p. 19) defined a group specifically as a finite set of actors who for conceptual, theoretical, or empirical reasons are treated as a finite set of individuals on which network measurements are made. A relation in graph theoretic terms, is the set o f all ties of one specific type among actors in a network. With these terms defined, I am able to adopt Wasserman and Fausts (1994, p. 20) more explicit definition of a social network as a finite set or sets of actors and the relation or relations defined on them. Types of Ties Network ties are viewed as pipes, or conduits, through which contents or resources flow (Podolny, 2001). As indicated earlier, while the number of different types of ties is limited only by the researchers imagination (Brass, 1995), network analysts have offered a general taxonomy of different types of ties based on different resources that are assumed to be exchanged through the tie (Tichy, Tushman, & Fombrun, 1979, p. 508; Brass, 1992; Brass, Galaskiewicz, Greve, Tsai, 2004; Pod olny & Baron, 1997): workflow ties involve the exchange of goods and services; communication ties involve
25 the exchange of information and ideas; friendship ties involve the exchange of affect; and authority ties involve the exchange of influence or power. However, network analysts have simplified this taxonomy still to recognize two broad categories of ties: instrumental ties that primarily involve the exchange of information as a resource, and expressive ties that primarily involve the exchange of affect as a resource (Brass, 1992; Ibarra, 1993; Balkundi & Harrison, 2006). Instrumental ties I nstrumental ties are those that arise in the course of work role performance and involve the exchange of jobrelated resources including information, expertise, profes sional advice, political access, and material goods or services (Ibarra, 1993). In this study of teams, instrumental ties are those involving the exchange of task related information, ideas, and work output. Instrumental ties are not necessarily prescribed merely by virtue of team membership. For example, self managing teams have discretion over how to structure their workflow and exchange of information and ideas among team members (Wageman, 2001), and researchers recognize that even in relatively similar environments, one should be able to observe quite different patterns of taskoriented behavior among groups at work (Shea & Guzzo, 1987). Expressive ties Expressive ties are informal social relations such as friendship and support that primarily involve the exchange of affect including liking, trust, and positive regard (Ibarra, 1993; Lincoln & Miller, 1979). Interestingly, network researchers focus on expressive ties has generally been on its positive aspects, while lesser attention has been focused on negative expressions of affect such as disliking, distrust, or adversarial intentions (Brass et al., 2004). Negative expressive relationships may be relatively rare
26 compared to positive ones (Baldwin, Bedell, & Johnson, 1997; Labianca, Brass, & Gray, 1998) but as Labianca and Brass (2006, p. 598) noted, their rarity belies their importance. They suggest that, owing to research on negative asymmetry that has shown that negative stimuli have been found to have greater explanatory power than positive or neutral stimuli in a diverse range of situations (Taylor, 1991), negative expressive relationships, despite their rarity, may have a greater effect on relevant outcomes than positive relationships. Thus, in this study of teams, I consider positive expressive ties as those based on exchanges of positive affect such as liking, trust, and friendship; and I also consider negative expressive ties based on exchanges of negative affect including disliking, distrust, and being adversaries. Multiplex ties Network researchers most often analyze networks of instrumental or expressive ties separately (e.g., Ibarra, 1993; Baldwin et al., 1997). However, relationships between actors are usually more complex, in that individual relationships are composed of simultaneously occ urring bundles of interaction (Burt & Schtt, 1985) including those that are both instrumental and expressive. This idea is captured in the network concept of multiplexity, which refers to connections between actors of multiple different types of ties (B rass, 1995; Scott, 2000). Max Gluckman (1967) originally made the distinction using the terminology of simplex (where persons are only linked in one context) and multiplex (where persons are linked in more than one context) relations and argued that t he central property of a multiplex relation is that it allows the resources of one relationship to be appropriated for use in others. Multiplexity is not a structural network variable itself, but it is used to depict structural variables in terms of only those actors with multiple types of ties existing between them
27 Studying configural patterns based on multiplex ties is important because multiplex ties capture networks in a level of detail and complexity that more closely approximates the reality of team interactions. Simplifying the natural complexity of relationships increases the likelihood of random errors in modeling team interaction, and accordingly, as in any data analysis, random errors in network data can be expected to attenuate standardized effects, amplify standard errors, and suppress evidence of true effects (Burt & Schtt, 1985). Research has shown that multiplex relationships are fundamentally different than those composed of simplex tie content in terms of their implications for work functioning (Methot, 2010). Because a multiplex tie can be viewed as a stronger form of relationship (Scott, 2000), multiplexity of ties has the potential to amplify the importance of configural patterns of team interaction. Researchers have acknowledged two compone nts to tie strength: communication frequency and emotional closeness (Granovetter, 1973; Marsden & Campbell, 1984; Burt, 1990). Communication frequency is how frequently individuals speak with each other, and emotional closeness is the level of affect associated with the interaction (Reagans & Zuckerman, 2001, p. 508). The concept of multiplexity makes it possible for researchers to combine both instrumental and expressive aspects into a single relational tie. In this study I will examine the intersection o f instrumental ties with both positive and negative expressive ties among members in the team. I define positive multiplexity as the overlap of instrumental and positive expressive ties between team members, and negative multiplexity as the overlap of inst rumental and negative expressive ties between team members. In sum, multiplexity is important because it will allow me to model multiple
28 types of interaction simultaneously in order to observe their joint influences on team functioning and effectiveness. Network Density I have thus far described basic network terminology and categories of ties, and now I describe how those building blocks compile to describe team network structure. The most common way to describe the pattern of relationships is with the co ncept of network density which refers to the connection, interrelatedness, and general level of linkage among the members of the network (Scott, 2000). Density is perhaps the most widely used structural variable (Scott, 2000, p. 69) and is important because it largely determines the level of information and resources that can move throughout the network (Balkundi & Harrison, 2006). Creating a density measure of team interaction first requires gathering information on the relationships or ties among team me mbers. This can be accomplished by providing individual team members with a list of all other members and instructing them to rate whether (for dichotomous ties) or the extent to which (for valued ties) they perceive to have a specific type of interaction with every other member of the team. Ratings are recorded in matrix form where the row labels indicate the team member giving the rating and column labels indicate the team member about whom the rating was given. Ratings are recorded in each cell of the ma trix corresponding to presence or absence (1s and 0s) of interaction between a particular pair of team members. Alternatively, ratings of the strength of interaction (range of values) can be recorded in the matrix. In the dichotomous case, ties are then mo deled between team member pairs with 1s in corresponding cells. In the valued case, the researcher may choose based on theory or conceptual reasons at which minimum level of strength to model ties. Density is then calculated as the number of
29 observed ties divided by the number of all possible ties (or the number of 1s present in the matrix divided by the total number of cells). Alternatively the researcher can employ a strengthbased measure of density that accounts for the value of each tie without dichotomizing (for an application see Reagans, Zuckerman, & McEvily, 2004, p. 114). In a network of N actors, the number of possible undirected ties is ( N2 N )/2; the number of possible directed ties is ( N2 N ). A complete network exists when every team member is connected to every other member, resulting in a density of 1.0, but such completeness is relatively rare, even in small networks (Scott, 2000). Density is an important structural pattern to model beca use the number of connections between team members is thought to indicate the teams capacity to coordinate its actions and thereby enhance its performance ( Reagans & Zuckerman, 2001). Teams with many members interacting with many other members are likely to have higher levels of information sharing and collaboration necessary for successful task completion, while teams with few members interacting are unlikely to be able or willing to share important knowledge and information to accomplish their tasks (Bal kundi & Harrison, 2006; Hansen, 1999). Further, teams whose members interact and communicate frequently are more likely to be able to build norms of support to smooth the functioning of the group ( Coleman, 1988; Reagans, Argote, & Brooks, 2005). Balkundi and Harrisons (2006) metaanalysis of 37 netwo rk studies showed that teams with higher densities of ties were associated with higher levels of team performance as well as team viability. Thus, the interconnectedness of the pattern of team member interactio n has important implications for the success of the team.
30 Density of Multiplexity While it is common to model network density based on a single type of relation (e.g., density of either instrumental or expressive ties), relationships among team members, as we indicated earlier, are usually more complex, in that they involve the simultaneous overlap of both instrumental and expressive ties (in both positive and negative forms). For this reason, and to more fully capture the level of complexity of team member interactions, it becomes important to model the structural properties of networks based on the overlap of instrumental and expressive ties or in other words, the structural properties of networks based on multiplex ties as a relation of their own. This ca n be accomplished by modeling the density of multiplex ties This type of density captures the connectedness of actors in the network through simultaneous instrumental and expressive ties. In this study, I examine the density of overlapping instrumental and positive expressive ties in the team and I label this density of positive multiplexity team structural cohesion. I also examine the density of overlapping instrumental and negative expressive ties in the team and I label this density of negative multiplexity team structural repulsion. My general proposition is the mix of both team structural cohesion and team structural repulsion, operating as simultaneous opposing forces, has important implications for team effectiveness beyond those accounted for by t he densities of instrumental, positive expressive, or negative expressive ties alone. Below I review the research literature applying network analysis in team contexts to illustrate that the networks of teams do have implications for team functioning and e ffectiveness. Teams As Social Networks The research applying a social network perspective to teams finds it s origins in laboratory experiments by conducted by social psychologists in the early 1950s seeking
31 to understand the impact of different designs of communication networks on team group efficiency in problem solving (see Shaw, 1964, for a review). For example, Leavitt (1951) conducted an experiment with a simple group problem solving task while constraining groups communication networks to conform to certain patterns. He found that groups whose communication networks were constrained to have high centralizationin this case a star pattern of communicationwere much more efficient at solving the problem, but he also found that members of these groups were much less satisfied with the group experience. On the other hand, Leavitt found that groups whose communication networks were constrained to have low centralizationin this case a circle pattern of communicationwere much less efficient at solving the problem, but the members of these groups were more satisfied with the group experience. After this early period, however, research on groups and teams in social psychology generally waned (McGrath, 1997; Levine & Moreland, 1990), and it wasnt until t he early 1990s, coinciding with the increasingly widespread adoption of teams in organizations (Gordon, 1992; Manz & Sims, 1993; Devine et al., 1999; Hoerr, 1989), that researchers gained renewed interest in applying a network perspective to groups and tea ms, albeit this time in fields other than social psychology, such as organizational psychology (Levine & Moreland, 1990). In the sections that follow I review research that has examined instrumental, expressive (positive and negative), and multiplex networ ks within and between teams, departments, or organizational subunits. Research on Instrumental Team Networks Research on instrumental team networks has focused mainly on work related communication and advice networks. In one of the earlier studies of instrumental networks within the team, Lucius and Kuhnert (1997) measured who are your top
32 choices to be associated with across several task scenarios among 129 cadets attending military college arranged into 29 squads (average team size was 4.4 members). They found marginal support indicating that the density of the squads instrumental choice networks was positively associated with military supervisor ratings of squad performance aggregated across 80 military field exercises, but surprisingly, density was not related to the squads self ratings of performance. However, given the small sample size (performance regressions were restricted to only 21 squads), this study may have lacked power to detect an association, if one existed. Additionally, the density of these instrumental choice networks was, surprisingly, not associated with a measure of team cohesion (using a traditional composition model measure), but it was positively associated with team member satisfaction. They acknowledged that they may not hav e observed a relationship between density and composition model cohesion because the measure of cohesion they used suffered from low reliability (alpha = .59). In a much larger sample, Reagans and Zuckerman (2001) measured the frequency of communication among work contacts in a sample of 224 corporate research and development teams and found that the density of communication frequency within the team was positively related to team productivity. They additionally found that the heterogeneous range of communication contacts to other teams was also positively related to focal team productivity. They argued that high network density improves performance through improved team coordination, and that high network heterogeneity enhance s performance through improved learning capacities and access to a broader range of external information, though they did not examine these mediators directly. In a similar study Reagans, Zuckerman, and McEvily (2004) replicated these
33 findings showing in a sample of 1,518 corporate project teams that both the density of internal team communication frequency and the heterogeneous range of external communication frequency were associated with faster project completion times. They additionally showed that diversity in functional background and tenure were negatively related to internal team density, but positively related to external team range. Cummings and Cross (2003) also examined communication frequency ties among 182 project teams from a worldwide corporation and found that the densi ty of communication ties, employing a strengthbased measure of density, was positively related to both manager and member rated team performance. Additionally they found that, controlling for the density of communication, hierarchical communication in the team, in which communication is ordered or tiered in a chain of command, was negatively associated with team performance. They also found that teams with coreperiphery structures of communication, in which communication revolved around a central core of members who communicated with a periphery of team members unconnected to each other, also exhibited decreased collective performance. Finally, teams with group leaders who maintained structural holes of communication between team members, rather than acti vely seeking to connect the team members around them, also exhibited diminished collective performance. These findings are interesting in that, while network features such as structural holes and hierarchy are said to yield benefits for individual accompli shment (Burt, 1992), these same structures may be detrimental for collective performance. Along those lines, Sparrowe, Liden, Wayne, and Kraimer (2001) conducted a field study involving 190 employees in 38 work groups from five different organizations and
34 found that at the individual level, an employees centrality in the employee advice network (indegree centrality, measured as the number of other employees who named the focal individual as someone they go to for help or advice on work related matters) was positively related to manager ratings of that employees individual performance. However, at the group level, the density of the work group advice network was not significantly related to ratings of work group performance. They did find marginal support indicating that the centralization of the advice network at the work group level, in which advice seekers tended to seek out a single individual in the work group for advice, was negatively related to group performance. They noted the paradox illustrated by their findings that at the individual level, centrality in the advice network was positively associated with individual performance, while at the group level, centralization of the advice network was negatively associated with group performance, suggesting that, in a manner similar to Cummings and Cross (2003), the determinants of group performance are different than those for individual performance. A more recent study suggests, however, that individual centrality in the withinteam advice network c an be beneficial for group outcomes, given that the central individual is also the teams formal leader. Balkundi, Barsness, and Michael (2009) conducted a study of 19 manufacturing teams in a wood products company and found that having a formal team leader with high in degree centrality in the teams advice network was positively related to team viability and negatively associated with team conflict. However, if this same leader broker ed many structural holes between subordinates in the withinteam advice network (subordinates sought advice from the leader but not each other measured by normalized betweenness centrality), the team
35 exhibited higher conflict and lower viability. Further, they found that t eam conflict mediated the relationships between leader centrality measures and team viability. These results parallel those of Cummings and Cross (2003) and Sparrowe et al. (2001) in that centralization of team advice networks, which was observed in Balkundi et al. (2009) as teams with leaders high in both indegree centrality and betweenness centrality, appear to create problems for team effectiveness. Hansen (1999) examined the benefits and drawbacks of strong and weak communication ties in terms of a different team outcomeinformation search and information transfer in a study of 120 new development projects undertak en by 41 divisions in a large electronics company. He found that weak communication ties between units facilitated the search for useful information in different divisions, but that those same weak ties inhibited the transfer of complex, noncodified information. On the other hand, strong communication ties between divisions restricted the search for information, but the transfer of complex information was facilitated by the strength of the commun ication ties. They suggested that having weak inter unit ties, because they are less costly to maintain in terms of time and energy, speeds up project completion when knowledge is not complex, but slows them down when the knowledge transferred is highly co mplex. Hansen also found that the density of withindivision communication ties was also positively associated with project completion. Finally, in one of the only studies to examine the development of instrumental team network structure longitudinally, Zaheer and Soda (2009) tracked the comembership networks (members of a current production team that had worked together as a part of a prior production team) of 501 production teams across a 12year span in
36 the Italian TV production industry. They found t hat the extent to which the current team spanned structural holes between teams (i.e., current team members bridge a hole between teams of a prior period because the current team members were selected from different prior teams, or current team members bri dge a hole between other current teams because they work in both teams) was largely a function of the extent to which the current team members had bridged structural holes in the past, but also was a function of the prior status and centrality of the teams that members were part of in the past. Conversely, current spanning of structural holes was negatively related to past team cohesion, indicating that teams whose members had worked together frequently in the past were less likely to bridge holes between c urrent teams. They found that current structural holes were positively related to team performance in terms of size of the TV production viewership audience, but that, ironically, it was the homogeneity, rather than the heterogeneity, of the teams prior production content that influenced performance across those structural holes. Research on Positive Expressive Team Networks Nearly all research focusing on positive expressive networks has examined friendship ties. Two early experiments during this time period examined the impact of friendship networks on team performance. First, Jehn and Shah (1997) examined the impact of friendship ties on task performance in an experimental setting with 53 threeperson teams of business students. Prior to the experimental task, the researchers asked each participant to list up to ten other students they considered friends. They then assigned two types of teams to complete both a cognitive task and a motor skills task. Friendship teams were those in which all three members assigned had mutually named each other as friends. Acquaintance teams were those in which each member
37 had named one other member as a friend without the other reciprocating the friendship. Though Jehn and Shah did not explicitly recognize it, the friendship densities they had created in these two types of teams were 1.0 for friendship teams (six of six possible friendship ties indicated) and .50 for acquaintance teams (three of six possible friendship ties indicated). No teams were assembled with mutual strangers, which would have had a density of 0 (zero of six possible friendship ties indicated). Findings from the experiment revealed that friendship groups, those with higher friendship density, performed better on both the cognitive and motor tasks. Additionally, this study was one of very few that provides additional data as to why this might have been the case. Jehn and Shah found that friendship groups exhibited higher group commitment, in that they appeared to be more committed to their task and t o each other and were generally more proud of their group in comparison to the acquaintance groups. In addition, the friendship groups exhibited higher group cooperation, in that each member appeared to put greater effort toward completing the task and tow ard helping each other. However, t hey found no difference between the groups in terms of their efforts to plan, make critical evaluations of their work, or their monitoring of each other. A second experiment highlights the implications for the structure of friendship ties between groups, as well as within them. Krackhardt and Stern (1998) conducted an organizational simulation with business students to observe whether the nature of their friendship ties within and between their assigned teams would influenc e the teams abilities to respond to an induced organizational crisis. Krackhardt and Stern used the friendship network within each of six student cohorts to create two types of simulated organizations. The first type of organization, called a natural or ganization, was created
38 to maximize the number of friendships occurring within the subunit teams. The second type, called an optimal organization, was created to maximize the number of friendships occurring between the subunit teams. At a certain point in the simulation, the researchers induce different crises requiring the organization subunits to respond immediately to unfamiliar and unlearned situations. Across the six trials, they found that the optimal organizations, those with greater friendshi ps between subunits, were able to respond more effectively. Krackhardt and Stern presumed the performance differences were due to variation in cooperation and conflict between the organizations, and provided anecdotal data taken from diary entries made during the trials that the subdivisions in the optimal organizations were seen to cooperate with each other in the face of the dilemmas that they encountered, while the subdivisions in the natural organizations frequently directed their members attention to protecting or enhancing the resources of their own division rather than those of the entire group. Three recent field studies have also examined the association of team friendship ties and performance. First, Mehra, Dixon, Robertson, and Brass (2006) exa mined friendship ties between the leaders of 88 sales groups of a financial services firm, and also the friendship densities within a subset of 28 of those sales groups ranging in size from 8 to 22 group members. They found that group leader centrality in the friendship network of sales group leaders was positively related to objective measures of performance including sales revenue and customer loyalty. They also found that friendship density among the subset of 28 groups was positively but not significant ly related to sales revenue, and significantly positively related to customer loyalty. Their reasoning for the former was that leader ties to leaders of other sales groups provided
39 faster and fuller access to novel information located outside the group, and their reasoning for the latter was that increased interconnectivity, or structural closure (Coleman, 1988) e nsure d that knowledge and support wa s easily shared among all group members. However, in neither case were these mechanisms examined directly. They suggested that the nonsignificance of the relationship between friendship density and sales performance was in part due to the lower power inherent in the small sample size of groups. Second, Balkundi, Kilduff, Barsness, and Michael (2007), in an ear lier study of the same 19 wood products manufacturing teams involved in Balkundi et al. (2009), examined both antecedents and outcomes of structural holes in the withinteam friendship networks. First, they found that neither ethnic nor gender diversity pr edicted structural holes in teams. They did find that age diversity was associated with fewer structural holes in teams, likely because the older members of teams formed mentoring relationships with younger members, as opposed to isolating them in rigid ag e clusters. In regards to performance, they found an invertedU shaped relationship between structural holes and team supervisor ratings of performance. They suggested that t eams with few structural holes were likely to be unable to be innovative, while teams with too many structural holes likely have difficulty coordinating yet they did not measure or test these proposed mediators. Finally, Oh, Chung, and Labianca (2004) found similar invertedU shaped associations between informal socializing ties (i.e., T o what extent did you go out with this person for social activities outside work such as going out to informal lunch, dinner, or drinks? ) and CEO ratings of work group performance in a study of 60 work groups
40 drawn from 11 different organizations in Korea. They suggested, but did not directly show, that work groups draw benefits of dense informal social ties through increased group solidarity, stronger norms of reciprocity, greater trust, and greater sanctions against self serving or opportunistic behav ior (e.g, free riding). However, teams only derive these benefits to a point, as they argued that excessive group density negatively impacts performance because strong ingroup biases can develop that can restrict access to outside information, creating ho mogenous and redundant information in the team Research on Negative Expressive Team Networks Only three studies have measured some aspect of negative expressive networks in a team setting. First, in addition to measuring advice networks, Sparrowe et al. ( 2001) also measured what they labeled hindrance networks (i.e., relationships with work group members who make it difficult for you to carry out your job responsibilities) among the members within the 38 work groups studied. They found that indegree centrality in the hindrance network being the one who made it difficult for others to do their jobs was negatively related to supervisor ratings of inrole and extrarole performance. They also found that density of the work group hindrance network was negatively related to group performance. They reasoned that performance suffers in groups where many individuals thwart the job performance of others by withholding information, opportunities, and resources. Second, in addition to measuring friendship ties, L abianca, Brass, and Gray (1998), also measured prefer to avoid relationships among 83 employees spread across 11 work groups of a university hospital center. They found that betweenunit prefer to avoid relationships were positively associated with per ceptions of between-
41 unit conflict, while betweenunit friendship were unrelated to betweenunit conflict perceptions. Additionally, in somewhat of a spillover effect, they found that the density of within unit prefer to avoid relationships was also positively related to betweenunit conflict perceptions, such that if members perceived they had negative relationships within their work groups, that appeared to influence their perceptions about the relations occurring between units as well. Interestingly, they did not examine whether prefer to avoid work group density was related to withinunit conflict perceptions. An important finding from this research is that the positive expressive tiefriendship was not related to conflict perceptions, while the negative expressive tieprefer to avoid was related. This is consistent with research on negative asymmetry (Taylor, 1991) that suggests that negative stimuli have greater influence than do positive stimuli across a range of perceptual outcomes. Last, Baldwin, Bedell, and Johnson (1997) examined adversarial networks (i.e., with whom do you have an adversarial or difficult relationship?) among 250 full time, residential MBA students divided into 62 teams within four cohorts, and found, controlling for cognitive ability indicated by Graduate Management Admissions Test (GMAT ) scores, that individual centrality in the adversarial network was negatively associated with an individuals satisfaction with team based learning as well as with general program satisfaction, but it was not associated with individual grades. At the team level, in group preference of adversarial ties (related to density, it is a measure of the extent to which team members named those within, as opposed to outside, their assigned teams as adversaries) was negatively related to team satisfaction, team workload sharing and perceptions of the effectiveness of team interactions, but was
42 surprisingly positively associated with team grades. A second counterintuitive finding was that team workload sharing was actually negatively related to team grades. They found these results disappointi ng but not inexplicable. In the first case, they speculated that adversarial relationships may have been instrumentally beneficial for task performance because it may have allowed team members to challenge each others assumptions and enhance decision qual ity. In the second case, they speculated that certain teams may have benefitted from relying on a highly intelligent star to carry the workload alone, lending credence to MBA instructors fears that teams allow one bright member to complete team projects to the detriment of the learning of the other team members. Research including Instrumental and Expressive Team Networks Thus far I have reviewed studies as having generally focused on either instrumental or expressive networks (whether positive or negati ve). However, a handful of studies have included both types of networks in the same study, though they have been analyzed separately within the study. For example, Baldwin et al. (1997) also examined friendship and communication networks, in addition to ad versarial networks. They found that withingroup communication was positively related to workload sharing and perceptions of effectiveness of team interaction but was not significant ly related to the team grades. They also found that withingroup friendshi p was positively related to perceived interaction effectiveness, which in turn was positively related to team grades. It was not, however, related to general satisfaction with the program. Interestingly, they found that the expansiveness of friendship (the degree to which the teams members chose members of other teams as friends) was negatively associated with team grades, suggesting that the teams that had the most socially extraverted members were not
43 necessarily the highest performing teams. Labianca et al. (1998) also measured communication and friendship networks in addition to the prefer to avoid network in their study of inter unit conflict. As mentioned earlier, friendships were not related to perceptions of inter unit conflict. The frequency of b etweenunit communication was also unrelated to perceptions of inter unit conflict. In an earlier study, Rosenthal (1997) examined individual friendship, advice, and discussion relationships among all the contacts both within and outside the team of the m embers of 15 cross functional teams in a robotics manufacturing company. She was interested in how the constraint created by the pattern of the team members ties would affect the focal cross functional teams performance. Constraint is an individual network measure indicating essentially the extent to which all of a persons contacts are connected to each other, thus constraining opportunities to broker between disconnected contacts (Burt, 1992). It is unclear from the study how the instrumental (e.g., adv ice, discussion) and expressive (e.g., friendship) networks were used to create the measure of individual constraint. However, after calculating a measure of constraint for each team member and then averaging those measures within the team, zeroorder correlational results indicated that the average level of team member constraint was negatively associated with supervisor rated team performance. Reagans et al. (2004) measured emotional closeness (respondents rating of how close [they] are with each [cont act] on a scale of especially close, close, not so close, and distant) in addition to communication frequency, but because they found the two measures were highly correlated, they presented their analysis based on communication frequency, noting t he results did not change if they conducted the
44 analysis based on the emotional closeness measure, or on the average of the two. Cummings and Cross (2003) also collected data on emotional closeness (labeled psychological closeness in their study) network s, and although it was not a focal aspect of their analysis, their results indicated that the density of the emotional closeness ties, in addition to the density of communication frequency ties, was positively associated with both manager and member rated team performance. Finally, in an attempt to summarize the team network literature, Balkundi and Harrison (2006) showed in a metaanalysis of 37 team network studies that the density of instrumental ties and the density of expressive ties (analyzed separ ately) are each positively related to both team task performance and team viability. They also examined whether team network structure was more likely a cause or a consequence of team performance and found strong inferential evidence that the former was th e case. Thus, network structures are more strongly positioned in time as antecedents to team performance, rather than as by products of it. They argued their findings provide compelling support for the view that social networks have important implications for team performance and viability, and urged researchers to increasingly incorporate social network concepts into theories of team effectiveness. Research on Multiplex Team Networks As described above, researchers have incorporated communication, friends hip, and adversarial networks into the same study (Bedell et al., 1997; Labianca et al., 1998), but to date no research has directly examined the overlap in any of these types of networks. The limited work that has been done has been conducted at the indiv idual level of analysis (Ibarra, 1995; Casciaro & Lobo, 2008; Methot, 2010). As mentioned earlier, Rosenthal (1997) collected data on individual friendship, advice, and discussion
45 relationships, but it was not apparent in her study how these networks were combined in her analysis. Only one study I am aware of has examined the overlap in multiple network relations in a team setting. In an experimental team simulation with first year undergraduate business students, Parise and Rollag ( 2010) examined the overl ap of prior instrumental ties (fellow students that had collaborated on prior school projects) with existing instrumental ties (students within the 10member team that collaborated a great deal during the twohour task simulation). The researchers collected data on the students prior existing friendship ties as well. They found that the density of overlap in current and prior work ties (multiplexity that the authors referred to as the reenacted work network) was positively associated with performance, b eyond the positive association of prior existing work or friendship ties with performance. Thus, the authors suggested that emergent work relationships based on preexisting work relationships are likely to be more productive (at least initially) than coll aboration between strangers. They suspected from their observations that this overlap may have impacted the degree of enthusiasm and competitive spirit seen while completing tasks but did not gather data concerning these mechanisms. Instead they called for research to evaluate the degree to which coordination, trust, and commitment are a factor in driving group performance based on team network structure. Limitations of Existing Team and Network Literatures The broad implication of this literature review is that the social network structure of team interaction has important implications for team functioning and effectiveness. However, this review also highlights several limitations in the existing team and network literatures. First, the general proposition underlying the traditional approach to modeling team phenomena in general is that team level constructs emerge from the interactions
46 of individuals in isomorphic, shared, and uniform compositions, rather than in discontinuous, complex, and nonlinear confi gurations, despite researchers acknowledgements that real team interaction likely emerges according to the latter rather than the former (Kozlowski & Klein, 2000; Mathieu et al., 2008; McGrath et al., 2000). The dominant use of traditional composition mod els to conceptualize and model team constructs that are configural in nature creates mismatches between the levels of team researchers theoretical and rhetorical complexity and the levels of researchers empirical and observational reality. Second, while some researchers have begun to apply a social network perspective as a useful alternative in order to capture more of the complexities of team interactions, the research has tended to focus on only one type of network relation, such as communication frequency (e.g., Reagans & Zuckerman, 2001), advice (e.g., Balkundi et al., 2009), or friendship (e.g., Mehra et al., 2006). Only limited research has examined more than one type of team network in the same study, and only three studies have looked at negative relations (e.g., Baldwin et al., 1997; Labianca, et al., 1998; Sparrowe et al., 2001). Thus, limited team research is beginning to examine instrumental and positive expressive networks, but more work is needed to examine these networks in addition to incor porating negative expressive networks that have been relatively under examined (Labianca & Brass, 2006). Related to this issue is that very little research provides any evidence, one way or the other, as to whether multiple measured networks are in fact di stinct, or whether they are measuring the same underlying relations. Thus, research is needed that validates the distinctiveness of instrumental, positive expressive, and negative expressive networks.
47 A more important corollary of the preceding limitation is that although some research does include multiple networks, to date no research has examined the multiplexity or simultaneous overlap, of those networks. There is no research linking multiplexity of instrumental and expressive ties positive or negativeto team outcomes of any sort. This is unfortunate because, as mentioned earlier, a multiplex tie is a unique relation of its own (Methot, 2010). Multiplexity captures the network in a level of detail that more closely approximates the reality of team interactions, and simplifying the natural complexity of team member relations may introduce random errors in network data that attenuate standardized effects, increase standard errors, and suppress evidence of true effects (Burt & Schtt, 1985). Thus, research is clearly needed which not only examines instrumental, positive expressive, and negative expressive networks, but also research that examines the multiplexity, or overlap, in these networks. A final limitation is that very little research has directly examined intervening processes that would explain why network structure is associated with team performance. Research has linked network features to either team performance (e.g., Reagans & Zuckerman, 2001; Reagans et al., 2004) or to some type of team proc ess or emergent state (e.g., Lucius & Kuhnert, 1997), but limited research has examined these associations jointly in the context of a broader structural model in order to examine whether the effects of one are mediated by the other, as implied by most models of team effectiveness. Jehn and Shah (1997) found that group commitment and cooperation mediated the relationship of friendship density with task performance as tested in a Baron and Kenny (1986) causal steps fashion, but this experimental finding
48 has not yet been replicated in a natural group setting with teams of a longer duration, and the Baron and Kenny approach has been shown to have low power for testing mediation effects (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002). Both Krackhardt and St erns (1998) and Parise and Rollags ( 2010) evidence on the meditational influence of cooperation and coordination was merely anecdotal. Only Balkundi and colleagues (2009) study found that conflict mediated the relationship between the team leaders cent rality in the advice network and team viability. Thus, on the whole, although researchers propose that the effects of network structures on team performance occur because of benefits associated with improved coordination, cooperation, solidarity, support a nd effort, these associations remain largely the function of argument and assumption rather than empirical observation and testing. No studies to date have comprehensively modeled the indirect links from network structure to performance through various pro posed mechanisms.
49 CHAPTER 3 THEORY AND HYPOTHESES The preceding review makes clear that a social network perspective is a useful approach for the study of team phenomena and that social network structures have important implications for team effectiveness. However, multiplex network structures have not be en examined, nor have their linking mechanisms to performance. Accordingly this dissertation seeks to make the following contributions to the team and network literatures. First, I develop a more complete conceptualization of social interaction in teams i ncorporating instrumental, positive expressive, and negative expressive relationships. This formulation is composed of both team structural cohesion, defined as the density of overlapping instrumental and positive expressive relationships in the team; and team structural repulsion, defined as the density of overlapping instrumental and negative expressive relationships in the team. These two components act as countervailing social forces that pull team members closer together or push them further apart. I c ontribute to network research by accounting for instrumental and both positive and negative ties, whereas previous conceptions often take for granted that expressive ties are positive (Brass et al., 2004; Labianca & Brass, 2006) and prior research fails to account for their overlap. I contribute to team research by illustrating how social network concepts allow compilation models of emergence to be operationalized. This approach directly addresses recent calls for researchers to incorporate to a greater deg ree social network concepts and variables that allow social relations in groups and teams to be captured and understood with greater complexity (Kozlowski & Klein, 2000; Klein & Harrison, 2007; Mathieu et al., 2008; Balkundi & Harrison, 2006).
50 Second, I i ncorporate theory from both the team effectiveness (e.g., Hackman, 1987) and social capital (e.g., Coleman, 1988) literatures to examine a systematic set of variables that explain how and why these components of team network structures influence team performance. In essence, I argue that structural cohesion and repulsion affect team performance through their influence on group synergy, consisting of maximizing team process gains, minimizing team process losses, and increasing team motivation (Hackman, 1987) I examine team process gains in terms of team members use of effective teamwork processes (Marks, Mathieu, & Zaccaro, 2001), while I examine team process losses in terms of team relationship maintenance ( Wright, 1984). I draw from motivational research on individual employee engagement (Kahn, 1990; Rich, LePine, & Crawford, 2010) to conceptualize team engagement as a representation of team level motivation. I make a contribution to the networks literature by systemati cally examining mechanisms that have previously been assumed or only anecdotally observed as explanations as to why team structure is associated with team performance. I make a contribution to team effectiveness literature by exploring how team network str uctures function as antecedents to these intervening mechanisms. In sum, I propose that social network analysis applied to team interaction will be useful to model both the contents and structure of team chemistry among team members, and that those struct ures will have important implications for team performance. Further, I include a set of intervening mechanisms to explain why these network structures influence team performance. I then empirically test these predictions in a structural model systematicall y including the indirect links from team network
51 structure to team performance through these proposed mediators. The model in Figure 3 1 summarizes the combination of these predictions. Team Synergy and Performance Because one of the main components of the definition of a team is that it has tasks to perform that result in discernable products, the key aspect of team effectiveness I focus on in this study is team task performance, defined by Hackman (1987) as the productive output of the group as evaluated against the standards of the people who receive and/or review the output. The criteria is appropriate because, as Hackman suggested, if a groups output is not acceptable to either its clients or to supervisors charged with evaluating the groups performa nce, then it cannot be considered effective. Further, for teams in organizational settings, it is rare for valid and reliable objective criteria to be available (for a notable exception see Mehra et al., 2006). In addition, Hackman argued that the implicat ions of performance for a group often depend far more on others assessments of the groups output than on any objective index, even if those assessments are made on whatever objective measures may be available. Hackmans (1987) model of effectiveness emphasizes three classes of variables proposed to impact team task performance, including the design of the group (e.g., the structure of the teams task; the size of the group), the organizational context (e.g., the reward and information systems influencing the group; the material resources available to the group), and group synergy (e.g., process gains and losses; motivation) that results from team members social interactions. In this current study, because of the nature of the sample, the first two classes of variables are relatively fixed in that the teams all complete similar tasks and are of similar sizes, and they have the same
52 reward systems and material resources available to them. Because of the lack of variance in these first two categories of variables, they are of less relevance to explain differences in team performance in this context. On the other hand, because the nature of team interaction in this context is not specified, structured, or mandated, I expect substantial differences to emerge in the group synergy available to the teams as a result of their varying patterns of social interactions. Shea and Guzzo (1987) acknowledged that even under the constraints of similar group designs and operating environments, teams will differ substantially in their patterns of working together, and this should be especially the case where decisions on how to structure team member interaction are left to the members discretion. In relation to group synergy, Steiner (1972) originally modeled team productivity as the difference between a teams maximum possible productivity minus its process losses. Actual productivity fails to match maximum possible productivity because of losses stemming from less than optimal ways of combining members efforts into a group pr oduct. Hackman (1987) added that actual productivity is also a function of process gains in addition to process losses, and that productivity is affected by a groups spirit or its members willingness to work hard to make the team its best. Hackman holds that the balance of synergy, whether positive as process gains and motivation exceed process losses, or negative as process losses overwhelm process gains and motivation, will serve to explain why some groups are able to excel given certain performance co nditions, and why others are not. In the sections that follow I describe theory and hypotheses for how elements of synergy affect team performance,
53 and then how the network structures of team chemistry function as antecedents to that synergy and performanc e in turn. Team Process Gains Team process gains involve getting more from a team than you would expect from the capabilities of individual members. Process gains result in new resources and capabilities that did not exist before the team created them (Hackman, 1987). One manner by which teams achieve these resource gains is through their teamwork processes. Teamwork processes are members interdependent acts that convert inputs into outcomes through cognitive, verbal, and behavioral activities directed toward organizing task work to achieve collective goals (Marks et al., 2001). Though researchers have disagreed on the meaning of team processes (Kozlowski & Bell, 2003), Marks and colleagues identified the content domain of team process interaction with ten narrow processes that map onto three higher order categories of transition action, and interpersonal processes. These higher order dimensions are embedded in a time based framework that stipulates that teams perform different types of processes at differ ent times corresponding to performance episodes distinguishable periods of time over which task work is performed and feedback becomes available. Transition processes describe team member activities that occur between performance episodes. During these ph ases, member interaction and communication focuses primarily on evaluation and planning activities to guide team accomplishment of objectives. These narrow processes consist of (a) mission analysis, (b) goal specification, and (c) strategy formulation and planning. Action processes describe team member activities that occur during performance episodes. During these phases, member interaction and communication focuses on activities that relate directly toward
54 the accomplishment of objectives. These narrow pr ocesses consist of (d) monitoring progress toward goals, (e) resource and systems monitoring, (f) team monitoring and backup, and (g) coordination of action sequence and timing. Finally, interpersonal processes describe team member activities that are ongoing and salient at all times and is argued to lay the foundation for the effectiveness of the other processes. Occurring throughout transition and action phases, this member interaction and communication is focused on managing interpersonal relationships. These narrow processes consist of (h) conflict management, (i) motivating and confidence building, and (j) affect management or regulating member emotion. Teamwork processes are integral to effective team performance because they allow team members to coordinate and combine their separate ideas, inputs, and expertise into collective group output. Transition processes allow team members to identify and prioritize team goals and develop plans as to how to achieve them. Teams that fail to identify goals or pl an how to achieve them risk misallocating team member efforts until it is too late to recover (Gersick, 1988). A team that disregards transition processes likely ends up with no shared understanding of the teams purpose (Marks et al., 2001). On the other hand, teams that agree on specific, difficult goals and develop strategies for how to achieve them should exhibit higher collective performance (Locke & Latham, 1990; OLeary Kelly, Martocchio, & Frink, 1994). Action processes allow team members to coordi nate their activities, track their progress, and provide feedback to members so that they are aware of their likelihood of achieving success within a given period of time. Teams that are unaware of their progress cannot provide appropriate feedback necessary to make adjustments in order
55 to achieve goals (DeShon, Kozlowski, Schmidt, Milner, & Wiechmann, 2004). Alternatively, teams that more extensively coordinate their activities and report regularly on progress are more likely to understand when they need t o speed up, slow down, or help each other out to get back on track to successfully complete their work (Marks et al., 2001; Marks, DeChurch, Mathieu, Panzer, & Alonso, 2005). Interpersonal processes allow teams to develop conditions for managing interpers onal relationships in a way that preempts or amicably resolves conflicts, encourages team members to maintain high levels of performance, and regulates team member emotions for productive team uses (Marks et al., 2001). If team members know they will be working together for any significant duration, they are less able to simply ignore unpleasant interpersonal relations until the team disbands, and their ability to smoothly manage interpersonal relations will significantly contribute to the effectiveness of the teams goal accomplishment (Bradley, White, & Mennecke, 2003; Druskat & Kayes 2000; Mathieu & Schulze, 2006). Taken together, the effective use of each type of teamwork processes facilitates team task accomplishment that should relate positively to team performance (Mathieu, Gilson, & Ruddy, 2005). Recent metaanalytic evidence has shown that teamwork processes are indeed significantly predictive of team performance, and this relationship is even stronger as team tasks become more interdependent (LePin e, Piccolo, Jackson, Mathieu, & Saul, 2008). Thus, I hypothesis the following: Hypothesis 1 : Teamwork processes are positively associated with team performance. Team Process Losses Team process losses refer to directing team members time, energy, and ta lent away from task accomplishment in order to manage interdependencies of collective
56 interaction (Hackman, 1987). Some losses are due to coordination loss, viewed as overhead costs inherent in all group task settings where team members must direct time away from direct task accomplishment to coordinate members activities (Steiner, 1972; Hackman, 1987). However, other losses are due to motivation loss (Steiner, 1972), often conceived in terms of social loafing, where team members contribute less effort to the group task than they would were they to perform the task individually (Latan, Williams, & Harkins, 1979). However, most social loafing research occurs in experimental studies where individual participants complete relatively simple tasks where they are only led to believe they are working in a group context (Karau & Williams, 1993). Drawing from the literature on social relationships and distractions (e.g., Wright, 1984; Winstead, Derlega, Montgomery, & Pilkington, 1995; Jett & George, 2003), anoth er form of process loss can be captured by the concept of team relationship maintenance This literature acknowledges that while social relationships such as good friendship are typically beneficial, a potential by product of increasing social closeness is the difficulty of maintaining these strong forms of relationships Relationship maintenance refers to the time, effort, and difficulty of coping with misunderstandings, disagreements, and incompatibility of goals in social relationships (Winstead et al., 1995). Adapted to the t eam level, I define team maintenance as the time, effort, energy and difficulty that members of a team experience in building and managing their social relationships. Team maintenance reduces team performance because time directed t oward building and maintaining social relationships reduces the overall level of energy and attention that would otherwise be applied to team task performance (Cohen, 1980).
57 More specifically, in a team context, the additional requirements of having to expend effort to resolve disputes, smooth disagreements, and provide assurances that team members are on positive standing with each other may deplete members energy to the point where they becom e less capable of dealing with problems, are less committed to working on tasks, and have less desire to see the team succeed and achieve its goals (Maslach & Leiter, 1997). These efforts may come to be viewed as distractions whereby individuals feel time and energy is being wasted on activities that are not instrumental in completing the tasks they are performing (Jett & George, 2003). When team members become exhausted and frustrated under the influence of these additional taxes on their capabilities and resources they will not be able to perform well because their energetic resources are diminished (Bakker, Demerouti, & Verbekke, 2004). Issues that distract team members from dedicating all of their energy and attention to task focused activity can reduce their effectiveness at performing that activity by limiting the amount of time available for and increasing the amount of interruptions in task completion. Recent evidence has shown that at the individual level, difficulty in maintaining interpersonal re lationships is negatively associated with job performance ( Methot, 2010). However, to date, no research has examined the relationship between team maintenance and team performance. Therefore, I hypothesize: Hypothesis 2 : Team maintenance is negatively ass ociated with team performance. Team Motivation Guzzo and Dickson (1996, p. 313) note that motivation in teams has received more theoretical than empirical attention. The purpose of this has been to understand motivation at a collective level rather than st rictly confining the construct to an individual analysis. At the individual level, motivation refers to the allocation of personal resources
58 toward role performance, and also how intensely and persistently those resources are applied (Kanfer, 1990). Two concepts attempting to represent motivation at the team level are group potency (Guzzo, Yost, Campbell, & Shea, 1993) and collective efficacy (Bandura, 1982). Both refer to shared team member beliefs that the team can be effective, but the distinction among them generally made that potency refers to general efficacy beliefs across situations while collective efficacy refers to task or situationspecific efficacy beliefs (Stajkovic, Lee, & Nyberg, 2009). However, these two concepts only speak in terms of team members beliefs in its capabilities to be effective rather than in terms of the direct motivational implications of how the team actually employs those capabilities in order to accomplish its tasks. A related concept is team commitment (Jehn, Northcraft, & Neale, 1999; Bishop & Scott, 2000), which is a team adapted version of Mowday, Porter, and Steerss (1982) Organizational Commitment Questionnaire (OCQ) that emphasizes an individuals identification with, and involvement in, a particular organization ( or team). However, this concept focuses on team members affective appraisals of their team experiences, rather than their actual investment of effort and energy on the teams behalf. Another concept that more directly and more completely captures the inv estment of team members personal resources toward role performance implied by team motivation is team engagement. Kahn (1990, p. 700) originally defined engagement at the individual level as the simultaneous employment and expression of a person s preferred self in task behaviors that promote connections to work and to others, personal presence (physical, cognitive, and emotional), and active, full performances. Engaged individuals harness their full selves in active, complete work role performances
59 by driving personal energy into physical, cognitive, and emotional labors. Kahn (1992) described engaged individuals as being psychologically present, fully there, attentive, feeling, connected, integrated, and focused in their role performances. They are op en to themselves and others, connected to work and others, and bring increasing depths of their personal resources to perform. Engagement is observed through the behavioral investment of personal physical, cognitive, and emotional energy into work roles (K ahn, 1992). People exhibit engagement when they become physically involved in tasks, whether alone or with others; cognitively vigilant, focused, and attentive; and emotionally connected to their work and others in the service of their work (Kahn, 1990). K ahns engagement concept is clearly motivational because it directly refers to the investment of personal resources toward role performance. And while it was originally conceptualized at the individual level, it is possible that it exhibits emergent group properties as team members each exhibit characteristic levels of personal engagement that come to define a shared level of team engagement. Researchers have argued that individuals spread their level of engagement to others in their work groups (Bakker, 2008), and empirical research has demonstrated that engagement has collective group properties, independent of the engagement of individual members (Bakker, van Emmerik, & Euwema, 2006; Salanova, Llorens, Cifre, Martnez, & Schaufeli, 2003; Salanova, Agut, & Peir, 2005). Thus, I define team engagement as team members shared investment of their collective physical, cognitive, and emotional energies into team role performance. While Kahn (1990) did not explicitly outline a relationship between engagement and performance, recent research by Rich, LePine, and Crawford (2010) provides
60 theoretical and empirical evidence at the individual level as to why this would be the case. Engagement should be beneficial for work role performance because individuals who are highly engaged not only focus their physical effort in the pursuit of role related goals, they are also cognitively vigilant and emotionally connected to the endeavor (Kahn, 1990; Ashforth & Humphrey, 1995). On the other hand, individuals who are highly dis engaged withhold their physical, cognitive, and emotional energies, and this is reflected in team task activity that is, at best, robotic, passive, and detached (Goffman, 1961; Hochschild, 1983; Kahn, 1990). Investment of physical energy contributes to per formance because it facilitates the accomplishment of tasks and goals through increased levels of effort over sustained periods of time (Kahn, 1992; Brown & Leigh, 1996). Investment of cognitive energy contributes to performance because behavior that is mo re vigilant, attentive, and focused reduces errors resulting from failures to see, take note of, or be attentive to ones work role (Weick & Roberts, 1993). Investment of emotional energy contributes to performance because individuals become more authentic and complete during role performance in a way that enhances connection among coworkers in the pursuit of task accomplishment (Ashforth & Humphrey, 1995). In regards to how engagement translates to performance at the team level, as a team on the balance h as more individuals that are personally engaged in their team work roles, and as this engagement spreads among team members to reinforce a characteristic level of team engagement, a highly engaged team should exhibit superior performance. These types of te ams are those that are willing to invest the time and energy to do what is necessary to produce highquality work (Maslach & Leiter, 1997) because the members value their membership in the group and find it rewarding to work
61 collaboratively with their team mates, and as a result they are willing to work especially hard on group tasks (Hackman, 1987). Thus, I hypothesize the following: Hypothesis 3 : Team engagement is positively associated with team performance. The Role of Team Chemistry in Relation to Synergy and Performance The notion of team chemistry is often invoked as an after the fact explanation in popular press articles as to why sports teams succeed or fail irrespective of the individual talents of the members composing the team (e.g., McGregor, 2006; Gauthier Villars, 2010). However, occurrence of the concept in the management and groups or teams literatures is surprisingly sparse. In fact, a search of the terms team chemistry or group chemistry in the Psy cINFO, Academic Source Premier, and Business Source Premier databases yielded only two articles. The first was a survey of United States Olympic team coaches as to what factors they attributed their teams successes and among them they listed strong team chemistry but without defining what team chemistry was ( Gould, Greenleaf, Guinan, & Chung, 2002). The second was a study examining pharmaceutical companies decisions to select and sponsor clinical research organizations to conduct clinical trials. One of the deciding factors listed was good team chemistry, described as a positive relationship between the sponsor and the clinical research organization, which are required to work closely together (Glass & Beaudrey, 2008). Much to my surprise, I found no articles or chapters in the groups and t eams in organizations literature even discussing the concept. An expanded search of the sports psychology literature yielded some studies that discussed team chemistry loosely without actually defining the construct. For example, in an article describing applied team building interventions with sports teams, Yukelson (1997) stated that the objective of a successful team building program is to enhance
62 team chemistry while getting everyone to work together toward common goals. However, Yukelson never actuall y defines what is meant by team chemistry, though he indirectly alludes to what it is by quoting the coach of a university hockey team who believed his team had good chemistry (p. 85): This is one of the best teams we have ever had in terms of camaraderie and team chemistry As the season progressed, the chemistry of the team became real tight, we became more and more of a family, almost like a brotherhood. The team did everything together, they would eat together, socialize together, pick each other up in practice, genuinely ask how each other is doing, and, in general, interact like a close knit family. Yukelson suggested that teams that sports teams that develop good chemistry become very much like families; that no matter what, good or bad, everyone sticks together for the betterment of the team. Teammates may not like each other, they may not agree with each other, they may have very little in common with one another, yet they know they belong to each other and are bound together by these family ties (p. 85). Only Bond (2007) has even discussed chemistry in a work setting, but this was in the context of a case study of one organization trying to achieve greater workplace diversity. However, Bond (2007, p. 1) did describe good chemistry as a work situation where things are going well because of naturally occurring forces of nature. In this situation, individuals mix and complement one another well, and as a result they may also sy nergistically bring out something new and positive in one another. Bad chemistry implies an interpersonal mess, even an explosion (p. 1), as personal elements collide and refuse to transform peacefully on contact. The players in this second scenario may simply not mix, like oil and water, or worse, may bring out something quite negative in one another. In personal life, people can choose friends and associates based on natural chemistry and decide to foster those relationships
63 that offer a relatively spontaneous connection. In the workplace, however, people often need to work closely with people in situations where there is little natural connection. I have referred to team chemistry as the mix of structural team cohesion, or the density of overlapping instrumental and positive expressive ties within the team, with structural team repulsion, or the density of overlapping instrumental and negative expressive ties within the team. Structural cohesion results when members of the team both work closely together and like, trust, and are friends with each other. Structural repulsion results when members of the team work closely together but dislike, distrust, and are adversaries with each other. Below I discuss several components of this definition that are im portant to recognize. The first component is that, for team chemistry to exist, team members must develop instrumental ties or in other words depend on each other to some degree to accomplish their work (Guzzo & Shea, 1992; Thompson, 1967; Van de Ven, D elbecq, & Koenig, 1976). Team members who work in complete isolation cannot develop team chemistry because, absent team member interaction, a collective structure cannot exist (Morgeson & Hoffman, 1999). In research practice, it is usually taken for grante d that members of the team depend on each other equally (whatever the level) and is often modeled simply by using composition model indices of respondents shared perceptions of interdependence ( e.g., Campion, Medsker, & Higgs, 1993; Campion, Papper, & Med sker, 1996; Langfred, 2007). However, interdependence is configural in nature in that it describes relational dependence requiring coordination among team members a system described by Kozlowski and Klein (2000, p. 17) as requiring networked contributions. Tesluk, Mathieu, Zaccaro, and Marks (1997, p. 201) even used net work
64 diagrams to depict their adaptation of Thompsons (1967) patterns of task interdependence or workflow. Because members of a self managing team have flexibility in the degree to which they choose to interact and depend on each other to accomplish work (Wageman, 2001), the network approach to team chemistry requires modeling the structure of task interdependence directly by depicting as instrumental ties who depends on whom to accomplish their work, team member by team member, rather than only indicating whether members agree that they generally depend on one another to accomplish their work. Thus, one requisite component of team chemistry is that team members work more rather than less closely together. The second component required for team chemistry i s that team members must have some expressive relation between themselves. While it is often the case that the nature of expressive ties is assumed to be positive (e.g., Ibarra, 1993), it is more likely that among the various relations within a single team some expressive ties are positive while others can be negative. Further, the formation of negative relationships is not the mere opposite of positive relationship formation (Labianca & Brass, 2006). Research has shown that friendships take longer to devel op and are included in more finegrained ranking distinctions, while negative relationships develop much faster and are included in more coarsegrained categories such as rival or enemy (Wiseman and Duck, 1995). Thus, a fuller conception of team chemistry involves both positive and negative expressive ties. A tie between two people is positive if it conveys positive affect, such as liking or love, socioemotional support, trust or friendship (Baker, Cross, & Wooten, 2003) Positive ties vary in type, as wel l as strength or quality. Strength, for example, can range from mild liking to what Dutton and Heaphy (2003) call High Quality
65 Connections. A tie between two people is negative if it conveys negative affect, such as disliking, distrust, animosity, or hos tility. It represents a recurring set of negative judgments, feelings, and behavioral intentions toward another persona negative person schema (Labianca & Brass, 2006). Labianca and Brass note that negative ties are rare, with recent empirical studies sug gesting that they make up only 1 to 8 percent of the total number of relationships in an organization (e.g., Baldwin et al., 1997; Gersick, Bartunek, Dutton, 2000; Kane & Labianca, 2005; Labianca, Brass, & Gray, 1998); but despite their rarity, they sugges t that, because of the principle of negative asymmetry (Taylor, 1991), negative ties may have even larger implications for task performance and socioemotional outcomes than positive ties. The third required component of team chemistry is tie overlap. The overlap of instrumental and positive expressive ties creates structural cohesion, a force that pulls members closer together and reinforces their bonds. The overlap of instrumental and negative expressive ties creates structural repulsion, a force that pus hes members further apart to relieve their tension. A simple analogy relating structural cohesion and repulsion to bar magnets may serve to illustrate this point. The force generated between magnets is a joint function of two factors: magnet position and m agnet polarity3. If magnets are positioned far apart, the orientation of their polarities matters little because the distance precludes them from exerting any force on each other. On the other hand, if the two magnets are positioned close together, then the orientation of their polarities matters a great deal. Closely positioned magnets with polarities that match generate a strong force that quickly pulls the magnets together (in the case of magnets, a match 3 The force is also a function of the relative strengths of each magnet, but for the sake of simplicity in the analogy, I assume magnets of equal strength.
66 is positively charged pole oriented toward a negatively charged pole). Alternatively, closely positioned magnets with polarities that mismatch generate a strong force that repels them apart (a mismatch is either two positively charged poles or two negatively charged poles oriented toward each other). In a team, the instrumental ties represent team member positions, while the expressive ties represent polarities. If team members are far apart, or in other words they lack or have very weak instrumental ties, then the orientation of their express ive ties, whether positive or negative, activates very little force that either attracts or repels. Because the individuals dont work together, their affective feelings towards each other have few implications for team effectiveness. On the other hand, if team members work closely together, or have strong instrumental ties, then the orientation of their expressive ties matters a great deal more as exchanges of positive affect reinforce and strengthen the bonds of team members working together while exchang es of negative affect destabilizes them and breaks them apart. In sum, the chemistry of the team is based on strong ties. Multiplex ties are strong forms of ties (Scott, 2000) that are capable of generating social forces that pull and push team members beyond those of simplex ties. The density of overlapping instrumental and positive expressive ties creates structural cohesion, or positive team chemistry resulting not just from working closely with other team members, but from liking, trusting, and being f riends with those one must work closely with. Likewise, the density of overlapping instrumental and negative expressive ties creates structural repulsion, or negative team chemistry that does not result from merely disliking other team members, but from di sliking, distrusting, and being adversaries with those that one must work closely with.
67 Because researchers have suggested and shown that instrumental ties have a tendency to overlap with positive expressive ties (Borgatti & Foster, 2003; Balkundi & Harris on, 2006; Reagans et al., 2004; Ibarra & Andrews, 1993; Ibarra, 1995; Methot, 2010), an interesting argument is whether highly interdependent team members are even able to develop negative ties, such that team structural repulsion never comes into existenc e. In other words, working closely constrains the development of negative affect between team members. There is no empirical evidence to test this argument one way or the other, but Labianca and Brass (2006) argued that while highdensity and hightaskint erdependence networks should serve to reduce the likelihood of negative relationship formation (because of the high monitoring ability and incentives to minimize disruptions), highdensity instrumental networks will serve to amplify the impact of negative relationships that do occur on task performance and socioemotional outcomes. In these descriptions we see elements of team chemistry as I have proposed it, consisting of team structural cohesion and team structural repulsion. The notion of teams creating s trong bonds (e.g., family ties, Yukelson, 1997) is captured through the network concept of multiplexity as stronger than normal ties. The concept of structural cohesion implies the willingness of teams to stick together because of strong bonds forged t hrough close working relationships with those who trust and like each other. The concept of structural repulsion recognizes that negative relationships with team members act as a counterveiling force that would try to break the team apart because team memb ers are unwilling or unable to continue working closely with those whom they dislike and distrust. Bond (2007) recognized these opposing forces in discussing both good chemistry and bad chemistry. Greve, Baum, Mitsuhashi, and Rowley
68 (2010) have also recognized these dual forces of cohesion and repulsion (their label was friction) in studies of interorganizational alliance networks. Positive Multiplex Density (Structural Cohesion) The metaargument for why network structure influences team performanc e is because of its influence on team social capital (Coleman, 1988; Portes, 1998), defined as the aggregate of the actual or potential resources which are linked to possession of a durable network of relationships in a social system (Bourdieu, 1985). More specifically, a dense, cohesive network structure is a form of social capital that allows a team to make more effective use of its individual and collective resources, which is assumed to contribute to the effectiveness of team production. Increasing clos ure (Coleman, 1998), or density, of positive multiplex ties ensures the existence of sufficient strong bonds between team members that coordination is facilitated, information spreads more quickly, and norms against deviant or opportunistic behavior are more easily enforced (Portes, 1998). Gittells (2003) theory of relational coordination argues that highquality connections, such as the positive multiplex ties of structural cohesion, promote more effective coordination of interdependent work because they result in more frequent, timely, and accurate communications dedicated to team problem solving. Increased proportions of positive multiplex ties in the team foster shared understanding of team goals, shared understanding of team members knowledge, and inc reased levels of mutual respect. Other research has shown that strong positive relationships facilitate knowledge transfer that improves group or organizational performance (Hansen, 1999; Tsai, 2001). In essence, increasing structural cohesion promotes greater and more effective use of teamwork processes. It fosters greater attention to transition phase
69 planning and goal setting; action phase monitoring, feedback, and coordination; and interpersonal conflict and confidence management throughout. Thus, I hypothesize: Hypothesis 4 : Structural cohesion is positively associated with teamwork processes. Social capital created from structural cohesion enhances teamwork resources, but it also promotes team energetic resources in the form of increased team engagem ent. Kahn (1990) described the psychological conditions that precede individual decisions to engage more completely in work role performances as experienced meaningfulness in which individuals sense a return on investments of the self through feeling wort hwhile, useful, and valuable; psychological safety in which individuals feel able to invest oneself without fear of negative consequences because situations are trustworthy, secure, predictable, and clear in terms of behavioral consequences; and personal availability in which individuals feel prepared to drive their physical, cognitive, and emotional energies into role performances. Kahn recognized that one of the most important factors driving the development of these conditions for individuals was the supportiveness of the social relationships with those whom they interact with the most. Along these lines, Dutton and Heaphy (2003) argued that positive, highquality relationships indeed afford the opportunity and safety for individuals to craft identities that fit who they are or who they wish to be, which makes them feel worthwhile. These valued iden ti ties help them derive positive meaning about the work they are doing. Baker et al. (2003) supplemented this line of argument even more directly with qualit ative interviews asking members of work groups to report the physical and psychological consequences of positive relationships with those they worked with (labeled as energizing relationships in their work). They found that nearly all of the 66 informant s they interviewed described some physical consequence of the energizing
70 relationships such as feeling more there, more up, more aroused, more intense, and more attuned to what is going on, and that this was reflected in voice inflection, body language, how they felt, and was even contagious to others. They also described cognitive consequences such as being more attentive to and processing information more rapidly and more thoroughly, often reflected in being able to retrieve more ideas from memory or be ing able to make connections to other ideas more quickly in a way that generated new insights. All of the informants described emotional consequences as a result of the energizing relationships such as feeling more enthused and drawn into the issues of a c onversation or problem solving session, and this enthusiasm carried over into commitment to interactions and a willingness to devote discretionary time to issues after interactions. In sum, the highquality and energizing connections involved in a team wit h high structural cohesion facilitates greater creation and investment of team members physical, cognitive, and emotional energy that should be reflected as greater engagement into team role performance. Thus, I hypothesize the following: Hypothesis 5 : S tructural cohesion is positively related to team engagement. While it seems natural to assume that a high degree of positive multiplexity in a team is universally beneficial, excessive structural cohesion in a team may have drawbacks as well, a fact recognized by Ingram and Zou (2008) who argued that members to positive multiplex relationships experience both benefits and tensions from those relationships. Bond (2007) argued that positive relationships among coworkers are not naturally occurring and require effort to be created and maintained. Along these lines, Katz and Kahn (1978) also argued that social structures are essentially contrived, not held together by any set of biological givens, and require considerable energy to be maintained. Indeed, as Wrig ht (1984) argued, growth in friendship is reflected in
71 increasingly broad and detailed acquaintance, and as such, the more one learns about a friend, the greater potential one has to become aware of aspects of that friends behavior that one may not like. This can result in incompatibilities in motives, interests, or momentary goals. Thus, creating and maintaining positive multiplex ties among team members requires additional attention, effort, and energy beyond that which is already being dedicated directl y toward task performance. In this sense, maintaining structural cohesion in the team can be viewed as a team challenge stressor (Lazarus & Folkman, 1984; LePine Podsakoff, & LePine, 2005; Crawford, LePine, & Rich, 2010), a demand which if met results in considerable benefits to the team, but the extra effort required to meet such a demand can result in depleted team energy Granovetter (1973) noted that, in general, strong ties are more difficult to maintain. In addition, Portes (1998) argued that high s tructural cohesion, despite its benefits, may also restrict individual members freedom to act by create excessive claims and expectations on team members, and may even be the source of downward leveling norms. Research has shown that the density of team s ocial relationships was positively associated with performance, but only up to a point (Oh et al., 2004), suggesting that there are costs associated with maintaining high density within the team, and other research has shown that positive multiplex relationships are indeed more difficult to maintain (Methot, 2010). This maintenance can be seen as a tax on team members resources, the management of which requires effort, and in turn creates additional draw s on team members energy. Thus, I hypothesize the following: Hypothesis 6 : Structural cohesion is positively related to team maintenance. I have to this point argued that structural cohesion acts as a force that pulls members together to make greater use of teamwork processes and increases team
72 engagement, but that energy and attention must also be directed away from task accomplishment in order to maintain these relationships. I had previously argued that this set of mechanisms contributes to team effectiveness in turn through their balance of team process gains, losses, and team motivation. What I have implicitly described is a model in which team structural cohesion exhibits indirect effects on team performance through its association with these intervening mechanisms. I expect structural cohesion to exhi bit positive indirect effects with team performance through its positive relationships with teamwork processes and team engagement, while I expect it to exhibit a negative indirect effect through its positive relationship with team maintenance. Hypothesis 7 : Structural cohesion has positive indirect relationships with team performance, as mediated by (a) its positive relationship with teamwork and ( b ) its positive relationship with team engagement, and (c) a negative indirect relationship with team perform ance, as mediated by its positive relationship with team maintenance. Negative Multiplex Density (Structural Repulsion) Just as positive multiplex density impacts performance because it builds a teams socia l capital, negative multiplex density reduces per formance because it creates team social liabilities (Labianca and Brass, 2006) described as decrements to the availability and use of the potential resources available to a team resulting from structures of increasingly strong, direct negative relationships between team members. The presence of negative relationships between team members that must work together impedes their exchange of performanceenhancing information and may also result in covert and overt behaviors, such as attempts to harm each other (Pondy, 1967; Pruitt & Rubin, 1986), that are disruptive to accomplishing work and achieving goals (Labianca & Brass, 2006).
73 In addition, Labianca and Brass (2006) do not equate negative multiplex relationships with conflict episodes but argue instead th at they may lead to conflict and other behaviors such as avoidance efforts and workflow redesigns that will have negative repercussions for the team. In these situations team members are more likely to find ways to operate as individuals attempting to avoi d getting in one anothers way rather than to attempt to enhance their performance as a group (Maslach & Leiter, 1997). Lang f red (2007) provided empirical evidence supporting this line of reasoning, showing that self managing team members unintentionally r estructured their workflow processes inefficiently as a response to conflict. Because of negative multiplex relationships, team members may ignore or retreat from setting goals and making plans to achieve them during transition phases; they may give less attention during action phases to tracking and communicating information related to performance progress, failing to give needed assistance to each other as a result; and they may have greater difficulties managing the interpersonal tensions and confidence deflations resulting from negative interactions. In sum, structural repulsion pushes team members apart such that it reduces their desire to coordinate their efforts as an integrated work group. Thus, I hypothesize the following: Hypothesis 8 : Structural repulsion is negatively related to teamwork processes. Team negative multiplex density will also likely be detrimental to team motivation in terms of its members willingness to invest their energies into team role performances. Kahn (1990) noted that among separate samples of camp counselors and architects, negative interactions among coworkers that communicated a lack of care, appreciation, and respect diminished the parties sense of meaningfulness and feelings of being valuable and worthwhile. One camp counselor noted that negative
74 interactions diminished his willingness to invest energy in the work as a result, commenting that, you put the energy where it will be appreciated (p. 708). In such situations, team members will prefer to be psychologically absent rather than investing energy on the teams behalf (Kahn, 1990). Negative multiplex density also erodes the psychological safety in the team as it fosters a threatening and stifling climate where members fear the dangers of taking risks because mistakes will be held against them (Kahn, 1990; Edmondson, 1999). In order to cope with the perceived dangers, team members may adopt a less authentic and more detached approach so as to not run the risk of offending others to the point of reinforcing already negative relationships (Goffman, 1961; Hochschild, 1983). One architect described such an approach with a coworker as knowing to stay away from him. But the problem is, I have to deal with him So we interact, but I keep it short I get monotonic, almost moronic with him (Kahn, 1990, p. 709). Negative multiplex density can also reduce members perceived availability for investing them selves into role performances as negative relationships can deflate members confidence and security that they even have the resources and energy necessary to produce for the team. Strong negative relations between teamm ates can undermine their self confidence and their willingness to express themselves because they become increasingly self conscious and focused on how others perceive them, rather than on how they might contribute to the team (Kahn, 1990). Negative comments about the teams competence can reduce members confidence levels (Marks et al., 2001), leading them in a downward spiral that drags down both team confidence and performance over time (Lindsley, Brass, & Thomas, 1995). In sum, structural repulsion
75 crea tes a psychologically disadvantageous team climate of reduced meaning, safety, and availability that undercuts members willingness to invest their energies into active and full team role performances. Thus, I hypothesize the following: Hypothesis 9 : Structural repulsion is negatively related to team engagement. Negative multiplex density can also be expected to contribute to team maintenance to an even greater degree than positive multiplex density and for different reasons. T he demands of social relationships and the patterns of difficult interpersonal communications within them were the original impetus for studying the causes of strain (Maslach, 1993). The social environment for most employees, and for members of teams specifically, is comprised of the day to day interactions within their immediate work groups (Maslach & Leiter, 1997). Negative expressive ties among those who must work closely together severely undermine the sense of solidarity and mutual support that are used as resources to accompl ish team goals. They are an emotional drain on the energy those team members need for creative and productive task work as the team in conflict works against itself. As these resource bases are fractured, the team members can be expected to more quickly ac crue the anxieties and frustrations associated with perceiving that they will be unable to meet the team task demands they must confront. In this sense, high structural repulsion in the team can be viewed as a team hindrance stressor (Lazarus & Folkman, 1984; LePine et al., 2005), a socially taxing demand that is appraised as having the potential to harm team productivity, growth, and gain, and as a result triggers negative emotions (e.g., fear, anxiety, anger) collectively in the team. These emotions are unpleasant and team members will be required to consume energy coping with them that would otherwise be available for task productive
76 uses. Research has shown that negative social interactions require greater cognitive processing than positive ones (Taylor 1991; Porath & Erez, 2007) resulting in fewer cognitive resources available for task performance activity. This additional draw on team resources creates a smaller reserve of team energy dedicated to team task performance. In essence, structural repulsion causes team members to waste time and energy dealing with unpleasant interactions rather than accomplishing tasks and goals (Maslach & Leiter, 1997). Thus, a social liability of increasing structural repulsion is that it results in additional team mainte nance. Hypothesis 10: Structural repulsion is positively related to team maintenance Taking these arguments together, structural repulsion creates social liabilities for teams in the form of reduced teamwork, reduced team engagement, and increased maintenance Given the previous arguments that these elements of team process gains, losses, and motivation influence team performance, I expect that team structural repulsion will exhibit indirect effects on team performance through these intervening mechanisms More specifically, team structural repulsion will have negative indirect relationships with team performance through its negative associations with teamwork processes and team engagement, and through its positive association with team maintenance Hypot hesis 11: Structural repulsion has negative indirect relationships with team performance, as mediated by (a) its negative relationship with teamwork, (b) its negative relationship with team engagement, and (c) its positive relationship with team maintenanc e
77 Figure 31. Mode l of hypothesized relationships Positive Expressive Ties Instrumental Ties Positive Multiplexity Negative Expressive Ties Negative Multiplexity Structural Cohesion Structural Repulsion Team Chemistry Team Process Gain (Teamwork) Team Motivation (Engagement) Team Process Loss (Maintenance) Team Performance + + + + + Network density of team: +
78 CHAPTER 4 METHODS Sample and Procedure Students enrolled or beginning enrollment in the Fall 2010 semester in the Master of Business Administration (MBA) programs of a large southeastern university p articipated in this study The MBA a dmissions office request ed that the 823 students arranged in 175 teams across 1 6 MBA programs (including traditio nal, executive, professional, and distanceI nternet programs) participate in the study. Because t he research methodology required that as many members as possible from each team complete the survey s, participation was strongly encouraged but remained voluntary and confidential. Students were asked to complete two surveys. The first was administered during the initial half of the semester and consisted of questions to assess team members instrumental, positive expressive, and negative expressive ties on a team member by team member basis. This survey also assessed team level control variables including team member tenure, as well as traditional team level referent shift (Chan, 1998) measures of task interdependence and cohesion. The second survey was admini stered during the latter half of the semester before any final grades were assigned and assessed referent shift perceptions regarding teamwork processes, team engagement, and team maintenance. During participant recruiting, participants were informed and gave their consent that, by participating in this study, they would be granting the researchers access to their team project grades as well as their background information that was collected by the MBA office as a part of the admissions process. Participan ts were assured that only
79 the principal investigator of this study would have access to this information and it would be used exclusively for research purposes. Furthermore, participants were assured that all personally identifying information would be des troyed, such that no participant would be able to be identified based on their responses. For each of the two rounds of surveys, I sent a personalized email with a unique link and password to each student to access his or her survey. I sent up to four follow up reminder emails over a period of two weeks following the initial distribution of the survey links to encourage nonrespondents to participate. Surveys were completed online (via www.surveygizmo.com) and transmitted to a secure server. A total of 629 of the 823 students completed the first survey for a response rate of 76%. A total of 574 students completed the second survey for a response rate of 70%. To be included in the analysis, teams could be missing no more than one members response to the first survey involving the network ties, and no more than two members responses to the second survey involving the team mediators. These inclusion criteria resulted in 109 usable teams composed of 511 individuals for a team response rate of 62% Average team size was 4.69 members. Respondents were 73 % male and 27 % female; 65% were Caucasian, 10 % were Hispanic, 8 % were Asian, 3 % were African American, 5% indicated Other, and the remaining 9 % did not specify their ethnicity. Average age was 29. 4 8 years (SD = 5 17 ). Following the conclusion of the semester, I contacted each of the 38 professors instructing in the 50 classes for which the participating MBA student s were registered to ask for available team project grades to be used as the studys measure of student team performance. I sent up to three reminder emails to professors who did not respond
80 to my initial inquiry. A total of 34 of the 38 professors (representing 43 of the 50 classes) responded for professor response rate of 89% (represent ing 86% of classes). Of those who responded, 25 professors (representing 31 classes) indicated that they assign ed team projects, while 9 professors (representing 12 classes) indicated that they did not. I followed up with the 25 professors who assigned tea m projects with documentation of the students signed consents to have their team project grades released to me. Of these 25 professors 20 (representing 25 classes) provided the student team project grades, while 5 professors (representing 6 classes) were unwilling or unable to provide the team project grades. Thus, team project scores were obtained for 81 % of the 31 classes in which team projects were assigned. All 16 cohorts of MBA students had at least one professor provide team project grades. Across all cohorts, i n only courses in which team projects were assigned, team based projects constituted 34% of the students total work. Across all courses (including those in which no team projects were assigned), team based projects constituted 23% of the students total work. In accordance with students consent, the MBA admissions office provided student background demographic information collected during the admissions process, including student age, gender, ethnicity, and Graduate Management Admissions Test ( GMAT ) composite score. After matching this information with the students survey responses as well as the provided team project scores, all personally identifying information was destroyed such that students and teams were only identified by anonymous identification codes.
81 Measures Network Relations I use d a roster method to collect complete withinteam relations among team members whereby each team member was presented a list of his or her teammates and was asked to assess the extent to which they hav e instrumental, positive expressive, and negative expressive ties with each other team member (1 = To a very small extent 2 = To a small extent, 3 = To some extent, 4 = To a regular extent 5 = To a good extent, 6 = To a great extent, 7 = To a very great extent NA = 0 = Does not apply ). The raw data collected from the respondents was used to create three independent team networks corresponding to instrumental ties, positive expressive ties, and negative expressive ties. Square matrices were created for each network as an array of cells with the random identification numbers of team members listed as labels for both the rows and columns (Scott, 2000). By convention, the value in each cell corresponds to the response of the person listed in the cell row about the pe rson listed in the cell column, or in other words, t he row person is rating his or her relationship with the column person. Instrumental networks To assess team instrumental networks, I asked each team member to indicate for every other team member how much do you interact with this teammate to get your group work done (e.g., communicating either in person, on the phone, or via email, webinar, or video conferencing)? This measure of work related communication frequency is similar to that used by Reagans and Zuckerman (2001), Reagans et al. (2004), and Cummings and Cross (2003).
82 Positive expressive networks The most common positive expressive network is that of friendship. To measure friendship relations, I asked team members to indicate for every other team member t o what extent would you say you are friends with this teammate (for example seeing them socially outside of class, sharing extracurricular hobbies, or confiding in them about personal matters ) ? The additional stipulation clarifying what is meant by friendship is included to ensure greater consistency in respondents understanding of the definition of friendship (Fischer, 1982). In addition, because Baker et al. (2003) suggested that measures of positive expressive networks include ot her types of positive ties, I also measured liking relations ( How much do you like this teammate [ by liking, I mean thinking about this person helps you feel positive, upbeat, and energized] ? ) and trust relations ( How m uch do you trust this teammate [ that is, you would be willing to depend on this person without having a way to check up on them] ? ). In the same way that confirmatory factor analysis (CFA) can be conducted on individual items of scales to confirm the extent to which they measure a common underlying construct, a CFA can be conducted on individual network relations to gauge the extent to which they measure the same underlying connection between actors. Thus, given appropriate confirmatory evidence, the three positive relations will be averaged to create a single network of positive expressive ties. Negative expressive networks Negative expressive networks are less commonly studied, but these relations can include adversarial networks (e.g., Baldwin et al., 1997), as well as more direct measures of negative affective exchanges such as dislike and distrust (Labianca & Brass, 2006). To measure adversarial relations, I asked respondents to indicate in
83 regards to every other team member t o what extent would you say you have a difficult r elationship with this teammate ( for example your personalities clash, you feel tension between you, or you tend to disagree or argue) ? To measure disliking relations I asked H ow much would you say you dislike this teammate ( by disliking, I mean thinking about this person really drains, annoys, or deenergizes you) ? To measure distrust relations I asked H ow much do you distrust this teammate ( that is you have to monitor them, or youd feel safer having things in your own hands rather than under this person s control ) ? ) Given appropriate confirmatory evidence from a network CFA, the three negative relations will be averaged to create a single network of negative expressive ties. Positive and negative multiplex networks After creating the instrumental, posi tive expressive, and negative expressive networks, I calculated positive and negative multiplex networks. Calculating multiplexity can be done in two ways. The first, used by Methot (2010) involves dichotomizing the valued data in the instrumental, positi ve expressive, and negative expressive networks at a given value such that responses greater than or equal to that value indicate the presence of a tie and responses less than that value indicate the absence of the tie. The positive multiplex network would then be created by combining the instrumental and positive expressive matrices such that in the resulting multiplex matrix, a zero (0) indicates there is neither an instrumental nor a positive expressive tie; a one (1) indicates there is an instrumental but not a positive expressive tie; a two (2) indicates there is a positive expressive tie but not an instrumental tie; and a three (3) indicates there is both an instrumental and a positive expressive tie a positive multiplex tie. The same procedure would be used to combine the instrumental and negative expressive
84 networks to create the negative multiplex network. Combining the multiple relations in this way yields a qualitative typology of the kinds of network relations that are present among the actors or nodes (Hanneman & Riddle, 2005). Methot notes, however, that the values in the combined matrices do not represent the strength of a tie (such that a zero would be the weakest form and a three would be the strongest). Rather, they are nominal values that si mply distinguish between the qualitative types of ties that exist between a pair of actors. The disadvantage of this approach is that it sacrifices variance by having to dichotomize the network relation data. An alternative approach, used by Parise and Rollag ( 2010) allows the researcher to preserve the tie strength values (as opposed to dichotomizing for presence versus absence) by combining the instrumental and expressive matrices using cell wise matrix multiplication. In this approach, the value of the cell corresponding to two actors in the instrumental matrix is multiplied by the value of the cell corresponding to the two actors in the ex pressive matrix, and the resulting value gives an indication of the strength of their multiplex interaction. For positive multiplexity, the strength values would be indicated as follows: 7 X 7 = 49: means that the person had very strong instrumental ties with someone whom they also had very strong positive expressive ties. 7 X 0 = 0: means that the person had very strong instrumental ties with someone whom they had no positive expressive ties. 0 X 7 = 0: means that the person had no instrumental ties with someone whom they had very strong positive expressive ties. 0 X 0 = 0: means that the person had neither instrumental nor positive expressive ties with the other. 4 X 4 = 16: means that the person had regular instrumental ties with someone whom they also had regular positive expressive ties.
85 For negative multiplexity, the strength values would be indicated as follows4: 7 X 7 = 49: means that the person had very strong instrumental ties with someone whom they also had very strong negative expressive ti es. 7 X 0 = 0: means that the person had very strong instrumental ties with someone whom they had no negative expressive ties. 0 X 7 = 0: means that the person had no instrumental ties with someone whom they had very strong negative expressive ties. 0 X 0 = 0: means that the person had neither instrumental nor negative expressive ties with the other. 4 X 4 = 16: means that the person had regular instrumental ties with someone whom they also had regular negative expressive ties. Structural cohesion and repulsion I then calculated s tructural cohesion as the valued density of the withinteam positive multiplex network for each team, respectively. I calculated structural repulsion as the valued density of the within team negative multiplex network for each team likewise. The valued density measure accounts for both the number and strength of each multiplex tie as a proportion of all possible multiplex ties in the network Because the network data was directed, the maximum number of possible ties for each team network was ( N2 N ) as opposed to ( N2 N )/2 for undirected (or symmetrized) data. Thus, in a team of five members, there were 20 possible ties of each type of multiplexity, both positive and negative. The multiplex density calculation sums the values of the team members multiplex ties and divides it by their maximum number such that t he valued positive multiplex density ranges from 0, no positive multiplex relations between team members ; to 49, maximum strength positive multiplex relations 4 For didactical clarity, I present these values as negative. However, because of network software requirement s, these values were treated as positive numbers in the analysis, such that a higher positive number indicated a stronger negative multiplex relationship.
86 between all team members. Likewise the valued negative multiplex density ranges from 0, no negative multiplex relations between team members ; to 49, maximum strength negative multiplex relations between all team members5. Team Perceptions Team perceptions were m odeled using a referent shift consensus composition model of emergence (Chan, 1998) in which team members rated their perceptions of the team or its members generally in regards to the following variables. Teamwork processes Teamwork processes were assesse d with 27 items from Mathieu and Marks (2006) measure of the Marks et al. (2001) teamwork processes. Example transition process items include asking members to rate to what extent their team actively worked to Identify our main tasks, Set goals for the team, Ensure that everyone on our team clearly understands our goals, and Develop an overall strategy to guide our team activities. Example action process items include to what extent the team actively worked to regularly monitor how well we are mee ting our team goals, Monitor and manage our resources, Develop standards for acceptable team member performance, Assist each other when help is needed Balance the workload among our team members, Smoothly integrate our work efforts, and Coordi nate our activities with one another. Example interpersonal process items include to what extent the team actively worked to Deal with personal conflicts in fair and equitable ways, Maintain group harmony, Develop confidence in our teams ability to perform well, 5 Once again, for didactical clarity, I present these values as negative. However, as before, these values were treated as positive numbers in the analysis, such that a higher positive number indicated greater negative multiplex density.
87 Encourage each other to perform our very best, and Keep a good emotional balance in the team. All items were assessed on a sevenpoint scale (1 = To a very small extent, 2 = To a small extent, 3 = To some extent, 4 = To a regular extent, 5 = To a good extent, 6 = To a great extent, 7 = To a very great extent). Cronbach alpha reliability for this scale at the individual team member level was .97. A high level of within group agreement (median rwg = .8 1 ) and significant intraclass correlat ion coefficients (ICC = .20; ICC = .50) justify aggregating the individual referent shift ratings to the team level. Team engagement Team engagement was measured with the Rich et al. (2010) job engagement scale with 20 items adapted for use at the t eam level. Team physical engagement is indicated by six items asking respondents to rate their extent of agreement as to, when the team members do their group project work, whether We w ork with high intensity, We e xert our full effort, We devote a lot of our energy, We t ry our hardest to perform well, We s trive as hard as we can to complete our project work, and We e xert a lot of energy. Team cognitive engagement is indicated by six items including We are mentally focused, We give our full at tention to our assignments We think hard about our projects We concentrate completely, We are focused on the group project work we do, and We pay a lot of attention to our assignments Team emotional engagement is indicated by eight items including We are emotionally connected, We put our emotions into what we do, We are really into our projects, We really care about what we do, We put our feelings into our projects We bring a lot of passion, We work with a sense of excitement and We are enthusiastic. All items were assessed on a sevenpoint scale (1 = Strongly disagree, 2 = Moderately disagree, 3 =
88 Slightly disagree, 4 = Neutral, 5 = Slightly agree, 6 = Moderately agree, 7 = Strongly agree). Cronbach alpha reliability for this scale at the individual team member level was .97. A high level of withingroup agreement (median rwg = .81) and significant intraclass correlation coefficients (ICC = .13; ICC = .36) justify aggregating the individual referent shift ratings to the team level. Team maintena n ce Team maintenance was measured with five items created specifically for this study including Building relationships with my team members takes a lot of time Maintaining relationships among team members uses up a lot of our energy It is difficult to main tain friendships with teammates, Interpersonal interactions in our team can distract us from accomplishing our tasks, and Maintaining our team's relationships takes time away from getting our work done. All items w ere assessed on a sevenpoint scale (1 = Strongly disagree, 2 = Moderately disagree, 3 = Slightly disagree, 4 = Neutral, 5 = Slightly agree, 6 = Moderately agree, 7 = Strongly agree). Cronbach alpha reliability for this scale at the individual team member level was .85. This scale also exhibited a high level of withingroup agreement (median rwg = 76), but the intra class correlation coefficients (ICC = .06; ICC = .20) were lower than those of other reported variables, indicating that although teams tended to agree on their characteristic levels of team maintenance, team maintenance exhibited less betweenteam variability than did other variables. However, Bliese (2000) noted that ICC(1) values are commonly observed in the range of .05 to .20, and that ICC(2) values typically range from .00 to .50. Taken together then, this evidence suggests it is also appropriate to aggregate these individual referent shift ratings to the team level.
89 Team Performance Team performance was measured as the average of student team project scores across classes. As described earlier, team project grades were provided for 81% of the classes in which team projects were assigned. Within each of the 16 cohorts, students functio ned as consistent team members in each of their classes, ensuring that team membership was constant across all assigned projects. Projects were generally graded on a 100point scale, and in the limited cases where a different scale was used (e.g., graded o n a 10point or 50point scale) I arithmetically converted the scores to a 100point scale. Control Variables I controlled for several team characteristics that might provide alternative explanations of team performance. First, I considered team tenure, s ince the amount of time members have worked together may have impacted the formation and cultivation of their interpersonal relationships, as well as their performance on team projects. I asked each team member to indicate, in months, how long they had bee n teammates with every other member of the team (acknowledging that some team members may have switched prior to the semester of study). I then averaged these personby person tenures to create an index of team tenure. To control for any demographic differ ences in performance, I considered team composition characteristics including the percentage of the team that was male as well as the percentage of the team that was Caucasian. Also, in order to control for the simple explanation that smarter teams are hig her performing teams, I controlled for team cognitive ability by including the team average GMAT score. Next, because the task structure of the assigned team projects might influence the extent to which team members interact to accomplish their work, I
90 con trolled for team members perceptions of task interdependence using a threeitem scale from Campion et al. (1993) Items include, I cannot accomplish my tasks without information or material s from other members of my team, Other members of my team depend on me for information or material s needed to perform their tasks, and Within my team, tasks performed by team mem bers are related to one another. Items were assessed on a sevenpoint scale (1 = Strongly disagree, 2 = Moderately disagree, 3 = Slightly disagree, 4 = Neutral, 5 = Slightly agree, 6 = Moderately agree, 7 = Strongly agree). Cronbach alpha reliability for this scale at the individual team member level was .66. A high level of withingroup agreement (median rwg = 75 ) and significant intracla ss correlation coefficients (ICC = .11 ; ICC = .3 4 ) justify aggregating the individual ratings to the team level. Finally, in order to examine the non redundancy of the configural network approach to team interactions, I also controlled for a traditional composition model of shared perceptions of self reported team cohesion an eight item scale from Dobbins and Zaccaro (1986) Items were measured on the same sevenpoint scale as task interdependence, and included, If given the chance, I would choose to leave my team to join another (reverse scored), The members of my team get along well together The members of my team will readily defend each other from criticism by outsiders I feel that I am really a part of my team I look forward to being w ith the members of my team each day I find that I generally do not get along with the other members of my team (reverse scored), I enjoy belonging to this team because I am friends with many team members and The team which I belong to is a close one Cronbach alpha reliability for this scale at the individual team member level was .90. A high level of withingroup agreement (median rwg = .83) and significant intraclass
91 correlation coefficients (ICC = .27; ICC = .61) justify aggregating the individual ratings to the team level.
92 CHAPTER 5 RESULTS Network Characteristics Before examining the social structure of each team in terms of positive and negative multiplex density and its influence on team functioning and team performance, I first examined the descriptive statistics and correlations among the networks themselves across all teams. These results, presented in Table 51, indicate that instrumental ties (Interact, mean = 4.82, s.d. = 1.44) and positive expressive ties ( Like, mean = 5.29, s.d. = 1.47; Trust, mean = 5.50, s.d. = 1.54; Friends, mean = 3.89, 1.98) occurred with greater strength and frequency than did negative expressive ties (Dislike, mean = 0.73, s.d. = 1.14; Distrust, mean = 0.96, s.d. = 1. 41 ; Difficult mean = 0 72 s.d. = 1.06). Of note, although the average value of the negative expressive ties is low, the standard deviations of these ties are similar to those of the instrumental and positive expressive ties Also, the minimum and maximum values (not reported in the tabl e) for all of the seven networks were 0 and 7, respectively, indicating that the full range of each rating scale was used in regard to each network including the negative expressive ties Taken together, these descriptive suggest that when individuals had negative expressive relationships with other team members, they were willing to indic ate it at least to some extent. The correlations between the networks indicate that positive expressive ties were positively related to instrumental ties (average correlat ion = .49) while negative expressive ties were negatively related to instrumental ties (average correlation = .28) The average correlation among the positive expressive ties was .56, while the average corre lation among the negative expressive ties was 59. The average correlation
93 between the positive and negative expressive ties was .37. This pattern of correlations indicates that positive expressive ties and negative expressive ties tended to be more strongly related within the positive and negative categories than between them. This pattern also indicates that, while positive and negative expressive networks are negatively related, the magnitude of this relation is small enough to suggest that a substantial portion of these relationships may cooccur. In this regard, the correlations between mirroredopposite networks are also revealing. Compared to the range of correlations where 0.00 indicates independence and 1.00 indicates a perfect negative relationship, the magnitudes of these correlations would seem to indicate that of the three pairs, trust versus distrust comes closest to measuring strict opposites (correlation = .62) while friend versus difficult appear to be measuring more independent networks (correlation = .19). The correlation between like and dislike networks was .49. However, even the large negative relationship between trust and distrust networks is only 62% of the way toward perfect negative collinearity, indicating that there is still substantial room for these relations to occur independently. While many of the correlations in Table 51 appear to be large in magnitude, the question remains as to whether these correlations could simply be observed by random chance. To answer this question, I needed to compare my observed correlati on to the standard error of a sampling distribution of correlations that one would expect from re peated sampling. However, in this case classical formulas underestimate the standard error because my network data violates assumption of independence. Underes timates of the standard error increase the likelihood of making a Type I error, or concluding that two network relations are significantly associated when the association could be driven
94 by chance. The appropriate alternative to test for network associations and correctly estimate standard errors is called Quadratic Assignment Procedure or QAP correlation (Hanneman & Riddle, 2005). To conduct this procedure, I used a nonparametric bootstrapping method to generate 20,000 random sub samp les of my network data, and the correlations from each of the 20,000 sub samp les were used to estimate a sampling distribution of correlations for each network. To test the hypothesis that there is an association between networks, I compared the observed correlation to the s ampling distribution of correlations generated from the random subsampling of networks. The proportion of correlations as large as (or as small as in the case of negative correlations) indicates the probability of obtaining my observed correlation by chance. Results of these QAP correlation tests indicate that, of the correlations presented in Table 52, the probability of observing the correlation by chance was less than .05 only for liking and trust network ties. The probability of observing the correlat ion by chance was less than .10 for interact and liking ties, for interact and trust ties, for liking and friend ties, for liking and dislike ties, for trust and distrust ties, for dislike and distrust ties, and for dislike and difficult ties. Taken together, this pattern indicates that although the magnitude of the correlations between the networks on the whole appear s to be large, a significant proportion of these correlations have a decent probability of merely occurring because of random chance. Thus, t his evidence would suggest that each of these types of networks captures unique information about the social relationships that team members have amongst themselves. Network Measurement Model As I indicated earlier, I conducted a CFA to confirm the extent to which liking, trust, and friendship networks reflected a common underlying relation of positive expressive
95 ties, and the extent to which disliking, distrust, and difficult networks reflected a common underlying relation of negative expressive ties. To conduct this analysis, I specified four models to test the factor structure of my measured networks using MPlus 4.0 (Muthn & Muthn, 2006) In Model 1, I specified three factors: Instrumental, Positive Expressive, and Negative Expressive, which were allowed to correlate I then loaded each network onto its corresponding higher order factor. The instrumental factor was indicated by the interact ion network only (by fixing the residual variance of the interact network to 0); the positive expressive factor was indicated by liking, trust, and friendship networks; the negative expressive factor was indicated by disliking, distrust, and difficult networks. In Model 2, I estimated a twofactor model by combining the instrumental and positive expressive factors. In Model 3, estimated a twofactor model by combining the positive expressive and negative expressive factors. In Model 4, estimated a single factor model refle cted by all seven networks. As can be seen by the fit statistics presented in Table 5 2 Model 1 consisting of three factors produced the best fit in an absolute and relative sense ( 2  = 148.12; CFI = .91; SRMR = .063; RMSEA = .149) Factor loadings o f the like ( .86) trust (.81) and friend (.59) networks on the positive expressive factor were strong and significant at p < .01. Factor loadings of the dislike (.86), distrust (.73), and difficult (.72) networks on the negative expressive factor were als o strong and significant at p < .01. Specifying either twofactor model (M2 to M1 2 [ 1 ] = 3 22 p = .0 7; M3 to M1 2 [ 2 ] = 223 23, p < .0 1) or a onefactor model ( 2 [ 2 ] = 223 23 p < .01 ) resulted in significant decrements in fit of the estimated model with the underlying data. Based on this confirmatory evidence, I averaged the like, trust, and friendship networks to
96 create the positive expressive network, and I averaged the dislike, distrust, and difficult networks to create the negative expressi ve network. Mediator Measurement Model Consistent with convention, I also assessed the fit of the data involving the team mediators to a measurement model prior to assessing their substantive relationships with the network variables and team performance. In Model 1, I loaded each individual item onto its corresponding higher order factor (teamwork, team engagement, and team maintenance), which, by default, I allowed to correlate. As the fit statistics in Table 53 indicate, Model 1 fit the data fairly well ( 2 [1,271] = 3262.51; CFI = .72; SRMR = .082 ; RMSEA = 120 ) Additionally, all of the loadings of the items onto their respective latent variables, shown in Table 54 were statistically significant and strong (average estimate/standard error = 9.04; ave rage loading = .80 ). I also compared this model to four other more parsimonious models that combined latent constructs to assess discriminant validity of the latent variables. In the alternative models, I combined team engagement and teamwork (Model 2), team engagement and team maintenance (Model 3), teamwork and team maintenance (Model 4), and all three mediators into a single factor model (Model 5). The statistics and fit indices in Table 53 show that none of the alternative models fit the data as well as the hypothesized model. Chi square difference tests indicate statistically significant differences in model fit favoring Model 1. Taken together, this analysis supports the adequacy of my measures to test my hypothesized structural relationships. Descri ptive Statistics and Correlations Table 54 reports the means, standard deviations, and zeroorder correlations among all study variables. As shown in the table, the study variables (when applicable)
97 each possess an acceptable degree of internal consistency reliability. The exception is the task interdependence control variable that is slightly below the conventional cutoff of .70 at .66. Among the control variables, traditional self report cohesion is positively related to team tenure, the percentage of the team that is male, the percentage of the team that is Caucasian, and team task interdependence. Between the control variables and the study variables of interest, task interdependence and self repor t cohesion are both positively related to team structural cohesion (positive multiplex density), and self report cohesion is negatively related to team structural repulsion (negative multiplex density). Task interdependence is also positive ly related to te amwork and team engagement, but it exhibits no significant relationship with team maintenance. Self report cohesion is also positively related to teamwork and team engagement, and it is negatively related to team maintenance. Of the control variables, only team tenure is significantly associated with team performance. It exhibits a counterintuitive negative relationship, however, such that teams that have been together longer on average tend to be poorer performing teams. I n examining preliminary support f or study hypotheses, a s expected structural cohesion (positive multiplex density) is positively related to teamwork and team engagement. Contrary to expectations, structural cohesion is negatively related to team maintenance. A s expected, structural repul sion (negative multiplex density) is n egatively related to teamwork and positively related to team maintenance; in addition while it was negatively related to team engagement, this zeroorder relationship was not significant. Finally, teamwork, team engag ement, and team maintenance are positively related to
98 team performance, but only the relationship of teamwork with team performance approaches significance ( p < .10). Analysis of Independence Because the sample of student teams was drawn from 16 different MBA cohorts, I assessed the possibility of a cohort level effect distorting variance in the team performance data (such that student team grades could likely be higher or lower by virtue of the lenient, strict, or range restricted grading by the professor s associated with the teams specific MBA coho rt class) by conducting a oneway ANOVA to produce estimates of variance in team performance at the team level within cohorts and between cohorts. Results revealed that MBA cohort membership did in fact explain significant variation in the team performance dependent variable ( F [ 15, 93] = 3 .8 9 p < 001; ICC[ 1] = .30). Thus, because a significant portion of the variance in team performance scores is found between the cohorts relative to within them, the observations of team performance are not independent of cohort membership, and the necessary assumption of i ndependence for ordinary least squares regression is violated. As a result, in testing my hypotheses, I needed to estimate my models using a method that accounts for the clustering of the team performance data by cohort. Structural Models I used M p lus 4.0 (Muthn & Muthn, 2006) to estimate structural model s to test my hypotheses Like other structural equation modeling programs, M p lus allows for the simultaneous examination of a model of theoretical relationships while correcting for the biasing effects o f measurement error. Unlike other programs, M p lus also has the cap ability to provide maximum likelihood parameter estimates with standard errors and chi square test statistics that are robust to the nonindependence of nested outcome
99 data. G iven non indepe ndence was the case in my data, I estimate d structural model s with the Mplus multilevel analysis ( type = TWOLEVEL; estimation method = MLR) using a categorical identification number to specify each teams cohort membership. Consistent with the theory outl ined above, I specified a partially latent structural model at the team level of analysis where team structural cohesion and team structural repulsion simultaneously predicted teamwork processes, team engagement, and team maintenance, which in turn predict ed team performance. I also included the control variables team tenure, percent male, percent Caucasian, team average GMAT, task interdependence, and self report cohesion as independent predictors of team performance. To correct for measurement unreliabili ty in teamwork processes, team engagement, team maintenance task interdependence, and self report cohesion, I set the residual variances of these factors equal one minus Chronbachs alpha reliability for the corresponding scale reported in Table 55. I fi rst specified a model including both direct and indirect paths from the network variables to team performance (Model 1) and a second, more parsimonious model eliminating the direct paths from the network variables to team performance (Model 2). This latter model corresponds exactly to the hypothesized model. Comparing Model 2 to Model 1 tests whether significant direct effects remain between the network variables and team performance after including the mediators. A statistically non significant decrement in fit would provide evidence for full mediation of the network variable effects, while a statistically significant decrement in fit would imply partial mediation, provided that the indirect effects through the mediators are statistically significant. Results reported in Table 56 indicate that the fit of Model 1 ( 2 [ 21] = 75. 26; CFI = .7 9 ; SRMR = 086 ; RMSEA = 154) is very similar to that of
100 Model 2 ( 2 [ 23 ] = 79 81 ; CFI = .7 8 ; SRMR = 087; RMSEA = 151) Becaus e the MLR estimation methods chi square test statistics are only asymptotically (rather than exactly) distributed according to a regular chi square distribution, the chi square statistics for the comparison and nested models must be corrected by their res pective scaling factor s when testing for significant decrements in fit. For this reason the chi square test statistic is slightly adjusted from the simple difference between the chi squares reported in Table 56. The scaled c hi square difference test ( 2 [ 2 ] = 3 37 p = .19) indicates that there was no significant decrement in model fit of the hypothesized Model 2 with the direct paths from the network variables to team performance removed relative to the fit of Model 1 with the direct paths included. As a result, all hypotheses were tested on the more parsimonious Model 2 where the direct paths from the network variables to team performance were removed. Tests of Hypotheses for Team Synergy and Performance The standardized path estimates from Model 1 are depicted in Figure 5 1. These estimates indicate support for Hypothes i s 1 in that the path from teamwork processes to team performance ( = .40, p < .05) was positive and statistically significant. Hypotheses 2 and 3 were not supported, however. Contr ary to expectations, team maintenance was positively related to team performance ( = .17 p < .05 ), while team engagement was negatively related to team performance ( = 18 p < .05 ). While these estimates are statistically significant the hypotheses are not supported because the estimates are in the opposite direction of what I hypothesized. Thus, my results supported my expectation that teams reporting more extensive use of teamwork would have higher levels of performance. On the other hand, my result s did not support my expectations that teams exhibiting higher levels of team motivation, and that were able
101 to focus less time and energy on relationship maintenance would have higher levels of performance. Instead, the opposite was the case. Teams reporting higher motivation in the form of greater team engagement exhibited lower performance, while teams reporting increased levels of time and attention toward maintaining team relationships exhibited higher performance. However, before concluding that team motivation in the form of team engagement is negatively associated with team performance, it is important to note that the estimated relationship may actually be an artifact of multicollinearity between team engagement and teamwork processes as the zeroorder correlation between these two study variables reported in Table 55 is quite high ( r = .72, p < .01). Accordingly, the estimated negative relationship is more likely a statistical artifact rather than a substantive relationship. Tests of Hypotheses for Structural Cohesion to Team Synergy Standardized path estimates in Figure 51 provided support for Hypothesis 4 in that the path from structural cohesion to teamwork processes ( = .53, p < .05) was positive and statistically significant. This provides support for my expectation that increasing structural cohesion facilitates coordination, communication, and information sharing reflected in the effective use of teamwork processes. Hypothesis 5 was also supported in that the path from structur al cohesion to team engagement ( = .44, p < .05) was positive and statistically significant. This supported my expectation that increasing density of highquality positive multiplex connections in the team would be associated with team members collective willingness to invest greater amounts of physical, cognitive, and emotional energy into team role performance. Contrary to my expectations, Hypothesis 6 was not supported. Although the path from structural cohesion to team maintenance was significant ( = .29, p < .05), it was in the opposite
102 direction of my expectation. While I expected that structural cohesion would have drawbacks in the form of extra time and energy required to maintain those social connections (a positive hypothesized relationship), t he relationship was actually negative in that teams that reported stronger positive multiplex ties apparently viewed maintaining those ties as less burdensome. Tests of Hypotheses for Structural Repulsion to Team Synergy Standardized path estimates in Fi gure 51 indicated support for Hypothesis 8 in that the path from structural repulsion to teamwork processes ( = .26, p < .05) was statistically significant and negative. This provided support for my expectation that teams with increasing prevalence of s trong negative relationships between team members would impede their exchange of information and work related communication important for coordinating teamwork activity. Hypothesis 9 was supported in that the path from structural repulsion to team engageme nt ( = .15, p < .05) was negative and statistically significant This result was consistent with my expectation that negative multiplexity would be detrimental to team members willingness to invest their energies into active, complete team role performances. Hypothesis 10 was also supported in that the path from structural repulsion to team maintenance ( = .29, p < .05) was positive and statistically significant. This indicates that, consistent with my expectation, negative multiplexity is also detrimental in that it additionally taxes team members energy and resources that must be devoted to coping with unpleasant team interactions rather than to activities focused on team task accomplishment. Tests of Indirect Effects Estimates of indirect effects of structural cohesion and structural repulsion on team performance through the three proposed mediators are presented in Table 57.
103 Hypothesis 7a was supported in that the indirect effect through teamwork processes ( = 21, p < .05) was positive and significant. Hypothesis 7b was not supported in that the indirect effect of structural cohesion on team performance through team engagement was actually negative ( = .0 8 p < .05) when I expected a positive relationship. This is due to t he unexpected negative relationship between team engagement and team performance. Next, although Hypothesis 7c technically was supported in that I found an offsetting negative indirect effect of structural cohesion on team performance through team maintena nce ( = .05, p < .05) the observed negative indirect effect was created from the unexpected negative association of structural cohesion with team maintenance, and the unexpected positive association of team maintenance with team performance. Taken toget her, the product of these two direct effects still resulted in the expected negative indirect effect but for precisely the opposite reasons. I expected teams with increasing structural cohesion to need to devote increased energy and attention to maintainin g those relationships and the maintenance of those relationships to be detrimental for team performance. Instead, structural cohesion was associated with less team maintenance, and the resulting decrease in team maintenance was associated with decreased team performance. Thus, support for Hypothesis 7c was qualified in that although it was statistically supported, it was not logically supported. Taken together, results for structural cohesion in this study indicate that it was only positively associated wit h team performance through its relationship with teamwork processes. Turning to structural repulsion, Hypothesis 11a was supported as evidenced by the indirect effect of structural repulsion on team performance through teamwork
104 processes ( = 11 p < .05) that was negative and significant. On the contrary, Hypothesis 11b was not supported as indicated by the nonsignificant indirect effect through team engagement ( = .03, p > .05). Hypothesis 11c was also not supported as indicated by the nonsignificant indirect effect through team maintenance ( = 05, p > .05) Thus, it appears that structural repulsion is negatively associated with team performance, but the indirect effect is manifest in this study only through its negative association with teamwork processes. T he non significant chi square difference test in favor of the more parsimonious structural Model 2 over structural Model 1 indicated that there were no significant decrements in fit after removing the direct paths from structural cohesion an d structural repulsion to team performance. Thus, the effects of structural cohesion and structural repulsion on team performance were fully mediated, and it appears that this mediation occurs primarily through the teamwork processes mechanism Finally, a s shown in Figure 51, the study variables including the controls jointly explained 31% of the variance in team performance.
105 Table 51. Descriptive statistics and correlations for networks Mean s.d. 1 2 3 4 5 6 7 1. Interact 4.82 1.44 -2. Like 5.29 1.47 .54 -3. Trust 5.50 1.54 .53 .68* -4. Friends 3.89 1.98 .40 .56 .44 -5. Dislike .73 1.14 .29 .49 .47 .24 -6. Distrust .96 1.41 .38 .44 .62 .25 .61 -7. Difficult .72 1.06 .18 .34 .31 .19 .65 .51 -n = 511 nodes reporting between 1,691 and 1,705 ties for each network. Correlation significance values obtained from quadratic assignment procedure (QAP) correlation test based on 20,000 permutations. p < .05 one tailed; p < .10, one tailed. Table 52. Network measurement model Structure 2 df CFI SRMR RMSEA 2 ( df ) Model 1: 3 Factors 148.12 12 91 063 149 Model 2: 2 Factors, I & PE combined 151.34 13 91 064 144 3.22 (1) Model 3: 2 Factors, PE & NE combined 371.35 14 77 100 223 223.23 (2) Model 4: 1 Factor 371.35 14 77 100 223 223.23 (2) n = 511. I, instrumental network; PE, positive expressive network; NE, negative expressive network. CFI, comparative fit index; SRMR, standardized root mean square residual; RMSEA, root mean 2 tests relative to Model 1. p < .0 5; p = 07
106 Table 53. Mediator measurement model Structure 2 df CFI SRMR RMSEA 2 ( df ) Model 1: 3 Factors 3,262.51 1,271 72 082 120 Model 2: 2 Factors, TE & TW combined 4,24 0 02 1,273 59 110 146 977.51 (2) Model 3: 2 Factors, TE & TM combined 3,492.29 1,273 69 089 126 229.78 (2) Model 4: 2 Factors, TW & TM combined 3,469.67 1,273 69 084 126 207.16 (2) Model 5: 1 Factor 4,45 0 48 1,274 56 111 151 1187.97 (3) n = 109. TE, team engagement; TW, teamwork; TM, team maintenance. CFI, comparative fit index; SRMR, standardized root mean square residual; RMSEA, root mean 2 tests relative to Model 1. p < .0 5
107 Table 54. Study items and factor loadings Team engagement Factor loading We work with high intensity. 84 We exert our full effort. 89 We devote a lot of our energy. 86 We try our hardest to perform well. 84 We strive as hard as we can to complete our project work. 82 We exert a lot of energy. 85 We are mentally focused. 85 We give our full attention to our assignments. 87 We think hard about our projects. 84 We concentrate completely. 84 We are focused on the group project work we do. 92 We pay a lot of attention to our assignments. 89 We are emotionally connected. 72 We put our emotions into what we do. 68 We are really into our projects. 85 We really care about what we do. 79 We put our feelings into our projects. 67 We bring a lot of passion. 85 We work with a sense of excitement. 80 We are enthusiastic. 78 Teamwork To what extent does your team actively work to Identify our main tasks? 67 Identify the key challenges that we expect to face? 68 Determine the resources that we need to be successful? 79 Set goals for the team? 76 Ensure that everyone on our team clearly understands our goals? 80 S et specific timelines for each of our goals? 72
108 Table 54. Continued. Develop an overall strategy to guide our team activities? 80 Prepare contingency ("if then") plans to deal with uncertain situations? 69 Know when to stick with a given working plan, and when to adopt a different one? 79 Regularly monitor how well we are meeting our team goals? 79 Use clearly defined metrics to assess our progress? 72 Know whether we are on pace for meeting our goals? 80 Develop standards for acceptable team member performance? 71 Balance the workload among our team members? 71 Assist each other when help is needed? 82 Communicate well with each other? 90 Smoothly integrate our work efforts? 88 Coordinate our activities with one another? 91 Deal with personal conflicts in fair and equitable ways? 78 Show respect for one another? 81 Maintain group harmony? 88 Take pride in our accomplishments? 79 Develop confidence in our team's ability to perform well? 88 Encourage each other to perform our very best? 84 Share a sense of togetherness and cohesion? 91 Manage stress? 86 Keep a good emotional balance in the team? 86 Team maintenance Building relationships with my team members takes a lot of time. 87 Maintaining relationships among team members uses up a lot of our energy. 95 It is difficult to maintain friendships with teammates. 62 Interpersonal interactions in our team can distract us from accomplishing our tasks. 50 Maintaining our team's relationships takes time away from getting our work done. 66 n = 109. All factor loadings are significant at p < .01.
109 Table 55. Descriptive statistics and correlations for study variables Mean s.d. 1 2 3 4 5 6 7 8 1. Team tenure in months 9.55 6.02 -2. Percent male 0.73 0.18 .01 -3. Percent Caucasian 0.65 0.26 .10 .13 -4. Team average GMAT 600.48 42.41 .12 .13 .14 -5. Task interdependence 4.73 0.72 .14 .04 .05 .05 (.66) 6. Self report cohesion 5.89 0.67 .25** .24* .20* .05 .27** (.90) 7. Structural cohesion 22.45 6.77 .03 .14 .03 .00 .37** .61** -8. Structural repulsion 3.09 2.01 .09 .10 .06 .05 .07 .33** .02 -9. Team engagement 5.47 0.64 .18 .08 .06 .01 .37** .49** .42** .14 10. Teamwork 5.51 0.67 .01 .04 .05 .02 .35** .57** .51** .25* 11. Team maintenance 2.75 0.66 .05 .12 .15 .03 .00 .39** .23* .23* 12. Team performance 91.67 4.04 .28** .14 .17 .16 .04 .05 .08 .07 n = 109. Coefficient alpha reliabilities on diagonal. p < .10, p < .05, ** p < .01.
110 Table 55. Continued 9 10 11 12 1. Team tenure 2. Percent male 3. Percent caucasian 4. Team GMAT average 5. Task interdependence 6. Self report cohesion 7. Structural cohesion 8. Structural repulsion 9. Team engagement (.97) 10. Teamwork .72** (.97) 11. Team maintenance .28** .36** (.85) 12. Team performance .07 .16 .07 -n = 109. Coefficient alpha reliabilities on diagonal. p < .10, p < .05, ** p < .01. Table 56. Structural models Structure 2 df CFI SRMR RMSEA 2 ( df ) Model 1: Model with all direct and indirect paths 75.26 21 .79 .086 .154 Model 2: Hypothesized model 79.81 23 .78 087 .151 3.37 (2) n = 109. CFI, comparative fit index; SRMR, standardized root mean square residual; RMSEA, root mean square error of 2 tests relative to Model 1. p < .0 5 Table 57. Tests of indirect effects Indirect effect through Relationship Team e ngagement Teamwork Team maintenance Structural cohesion to team performance .0 8 21 05* Structural repulsion to team performance .03 11 05 n = 109. Estimates from Structural Model 2 All estimates are standardized. p < .05 one tailed
111 Figure 51. Structural model relationships. All paths standardized. Dashed line indicates path not hypothesized in Figure 3 1. Control variables estimates are from direct relationships with team performance. p < .05, onetailed
112 CHAPTER 6 DISCUSSION Summary of Findings Scholars have long been interested in the implications of social interaction for the functioning and effectiveness of work teams and are increasingly recognizing that teams social interaction occurs in complex, discontinuous, and nonlinear patterns. However, current conceptions of team social interaction have yet to account for this complexity. Social network analysis provides an ideal avenue for examining the complexity of social relationships in teams, but network research to date has generally examined either instrumental (work or task relate d) or positive expressive (social friendships) relationships in isolation. I examined a more replete conception of team social interaction based on the intersection of instrumental and expressive relationships in both their positive and negative forms. I used the concept of positive multiplexity to capture the overlap in instrumental ties with positive expressive ties such as liking, trust, and friendship. I used the concept of negative multiplexity to capture the overlap in instrumental ties with negative expressive ties such as disliking, distrust, and difficult or adversarial relationships. I then drew on social capital theory and team design theory to examine the implications of teams densities of positive multiplex relationships, termed structural cohesion, and negative multiplex relationships, termed structural repulsion, for team functioning and effectiveness. Altogether, there is evidence that the structure of social relationships in the team matters for team effectiveness. Specifically, I found that structural cohesion had positive indirect effects on team performance through its positive association with teamwork processes. Although I expected structural cohesion to have drawbacks for team
113 performance in terms of increased time and energy required to maintain those relationships, this turned out not to be the case, as structural cohesions negative indirect effect on team performance was actually created by a decrease in team relationship maintenance, which was in turn unexpectedly positively related to team performance. In addition, although I expected additional benefits of structural cohesion to team performance to occur through increased team motivation, this also turned out not to be the case as team engagement was unexpectedly negatively related to team performance. However, this latter relationship may be more likely due to a statistical artifact caused by the multicollinearity of team engagement with teamwork rather than being a substantive relationship of its own. Thus, in this sample, struct ural cohesion was beneficial for team performance primarily through more extensive teamwork. As for structural repulsion, I found that it had a negative indirect effect on team performance through its negative association with teamwork processes. Although I expected additional negative indirect effect s through decreased team engagement and increased team maintenance, neither was the case as these indirect effects were not significant. Thus, in this sample, structural repulsion was detrimental for team perfo rmance but mainly through its association with less effective teamwork. Notably, e ach of these effects hold after simultaneously controlling for other factors that might influence team performance, including team tenure, team demographic makeup, team intelligence, team task interdependence, and self reported team cohesion. Theoretical Contributions Positive and Negative Multiplex Relationships One of the most important contributions of this research is that it examined multiple forms of both positive and negative expressive ties in conjunction with
114 simultaneously occurring instrumental ties. To my knowledge, this is the first study to provide evidence on the relationships between multiple forms of positive and negative expressive ties. Although it would seem reasonable to assume that positive and negative relationships would be manifest as opposite ends of a single continuum, my observed correlations are much smaller in magnitude than they would need to be if this were actually the case. In addition, evidence from a confirmatory factor analysis suggests that modeling positive expressive and negative expressive ties as separate factors is a much better fit to the underlying data than is modeling them as a single combined factor. Thus, an important cont ribution of my research is evidence suggesting the distinctiveness of positive and negative expressive ties within teams, and even within individual relationships, providing additional support to the notion that individual social relationships are richer and more complex than might be initially assumed. A second major contribution is that I examined the multiplexity, or the simultaneous overlap, of instrumental ties with both positive and negative expressive ties. This multiplexity is useful to capture more of the richness and complexity involved in the bundles of multiple simultaneous interactions of interpersonal relationships. In addition, t he densities of positive and negative multiplexity were correlated only at .02, providing further evidence for the distinctiveness of positive and negative expressive relationships. Further each type of density exhibited unique structural effects with team functioning and effectiveness variables, indicating that both provide incremental contribution in predicting impor tant team outcomes. Taken together, these results support examining the joint occurrence of instrumental with positive and negative expressive ties as a more replete conception of the social structure of teams.
115 Mediating Role of Teamwork, Engagement, and Maintenance A third important contribution is that I specified and tested the extent to which mechanisms suggested by research on team synergy serve as mediators of the relationships between team social network structures and team effectiveness. This is an important contribution because researchers frequently assume that network structures affect team performance through improved coordination, cooperation, and solidarity, but these associations have yet to be tested in a comprehensive meditational model. To date only Balkundi et al. (2009) have found that conflict mediated the relationship between team leader centrality in the advice network and team viability. I examine a broader range of mediating mechanisms, including team process gains in the form of tea mwork processes, team process losses in the form of team maintenance, and team motivation in the form of team engagement. I show that each of these mediators has important relationships with the social structure of the team In particular, increasing densi ty of positive multiplex relationships is associated with increased use of teamwork, increased team engagement, and decreased team relationship maintenance. On the other hand, increasing density of negative multiplex relationships is associated with decreased use of teamwork, decreased team engagement, and increased team relationship maintenance. However, only teamwork processes played a particularly important mediating role in relationships between team social structure and team performance. More specifically, teams with increasing density of positive multiplex relationships had increased team performance primarily because of their more effective use of teamwork processes such as planning and coordination rather than because of the level of team motivation or team relationship maintenance. Similarly, t eams with increasing density of negative multiplex relationships had decreased team performance
116 primarily because of less effective teamwork rather than the level of team motivation or team relationship maint enance. Thus, while each of these mediators is associated with team social structure, it appears that association of social relationships with teamwork processes is what really matters for team performance. These findings are a valuable contribution in that they expand our knowledge of the mechanisms that are responsible for the effects of team network structure on team effectiveness outcomes. Implications for Team Effectiveness and Social Capital Theories This study has important implications both for team effectiveness theory (Hackman, 1987) and group social capital theory (Oh et al., 2006). First, team effectiveness theory has traditionally focused on group design elements consisting of task structure group size and diversity, or group norms; and contextual design elements such as the group reward system, technical resources, and training opportunities. The assumption is that by focusing on optimal arrangements of these design features, team effectiveness improves through the natural outflow of increases in members levels of effort and application of knowledge and skill. My research shows that another key element to be incorporated into team design theory is the structure of social relationships among team members. A network of strong positive relations hips between team members is associated with increases in the levels of team member effort and use of teamwork processes, while networks of strong negative relationships are associated with decrements in teamwork and increases in process losses of managing difficult relationships instead of focusing on task accomplishment. For these reasons, the structure of social relationships in a team is an important input to be incorporated along with other features of group design in team effectiveness theory. Additio nally, I have also drawn from the literatures on personal engagement (Kahn,
117 1990) and personal relationship maintenance (Wright, 1984) to further develop the notion of group synergy within team effectiveness theory by articulating conceptualizations of team motivation in terms of team engagement, and team process loss in the form of team relationship maintenance. Although the relationships of these two concepts with team effectiveness were somewhat unexpected, after additional refinement t hese concepts may yet prove useful in allowing team researchers to further examine how the elements of group design influence team effectiveness through these mediating process criteria of effectiveness. My research also has important implications for social network theory and group social capital theory in particular. Specifically, my work reinforces the importance of examining both the potential benefits and the potential liabilities of social relations hips in a more complete social ledger (Labianca & Brass, 2006) by includ ing positive and negative expressive ties as separate relationships rather than opposites of a single expressive continuum. This will allow network theory to develop a more complete understanding of the different ways in which positive and negative relationships contribute to and detract from individual and group effectiveness. A second implication is that network theory consider not only both positive and negative expressive ties, but also their multiplex overlap with instrumental ties, and how these stronger, richer descriptions of the complexity of social relationships have the potential to explain outcomes beyond the explanation of ties that were previously assumed to be solely instr umental or expressive in nature. Finally, while group social capital theory emphasizes resources that groups accumulate by virtue of their social structures in terms of access to information, political connections, trust, and emotional support, my
118 research further develops the nature of group social capital resources in terms of increased teamwork, team engagement, and reduced maintenance Thus, this research helps to elaborate group social capital theory by further articulating the nature of the resources that accumulate to groups with beneficial social structures. Practical Implications From a practical standpoint, my findings indicate that social relationships in teams matter for team effectiveness, and should be taken into account when considering how t o create teams with the best possible chance of succeeding. Team members who like, trust, and are friends with each other tend to coordinate better together, they work harder together, and they spend less time off task managing relationships, while team me mbers who dislike, distrust, and have difficult relationships with each other tend to coordinate less, invest less energy, and are distracted from task work in order to deal with the difficulties of being together. Thus, when designing work teams, it is im portant to consider the potential social interactions among team members In particular, if team members are having a hard time getting along, their teamwork and team performance will suffer. Instead of forcing them to work it out, one simple implication i s that an easier fix might be to break the team apart and rearrange it with new members so that the new team has a better chance to get along with each other To the extent that it is possible to know among a set of potential team members who is friends wi th whom, one simple idea is to form teams based on existing friendships, or to allow teams to form on their own based on their existing friendships if friendships are unknown. If potential team members are all relatively new acquaintances, allowing some ti me for social relationships to form before teams are selected could facilitate this process. If there is simply no opportunity to form social relationships before forming teams, team members
119 should be encouraged to develop positive social relationships thr ough extracurricular socializing and other activities that build liking and trust among team members. Ultimately, to the extent that leaders and managers of teams can develop positive social relationships and limit or remove negative social relationships among team members, teamwork, team engagement, and team performance should improve. Limitations There are several limitations in my work to be considered before my results are used to guide managerial practice. F irst, the sample is comprised of student teams in an MBA program rather than work teams in actual organizations. Thus, the degree to which my findings generalize to other work teams of longer duration in differing places of employment needs to be assessed. Second, my findings should not be interpreted as indicating causal relationships between team social structure and team functioning and effectiveness. I did not manipulate (nor could I have) the level of team social relationships in order to examine their causal influence on teamwork, team engagem ent, and team relationship maintenance. However, this concern is limited somewhat by prior experimental research (Jehn & Shah, 1997; Krackhardt & Stern 1998) that manipulated the levels of team friendships to find causal effects on team performance outcom es. Third, although group social capital theory implies that group social structure precedes team performance, it is possible that the relationship occurs in reverse, such that high performing teams are more likely to evaluate their social relationships more positively than lower performing groups. While I made an effort to address this possibility by collecting team social network data earlier in the semester before final team performance grades were assigned, this concern is not totally mitigated because many teams in the sample had been in existence and had received
120 performance feedback prior to the semester I conducted the study. Because prior semester team performance data was not available for these teams, I could not control for the effect of prior te am performance on current social relationships nor on current semester performance, though I did include team tenure as a control to address this to the greatest extent possible. Future research should measure network relationships both before and after mu ltiple performance evaluation episodes to further investigate the underlying temporal nature of network relationships and additionally address causality implica tions of network relationships. Fourth, one of the limiting factors in my ability to explain variance in team performance was the relative range restriction in the team project scores provided by the MBA course professors. Although the scores were provided on a 100point scale, only a very small portion of that scale was typically in use. For example, in nine of the sixteen MBA cohorts, the standard deviation in team project grade scores was less than three points. In the other seven cohorts, the standard d eviations in team project grade scores ranges from 4.31 to 10.13 points. While multilevel analysis was able to control for the unequal variances in team performance distributions across the cohorts, the limited range of scores within many of the cohorts li kely attenuated the strength of my observed relationships. Future research might observe stronger effects of team network structure on team performance if the team performance outcome discriminates to a greater extent between highand low p erforming teams Future Research Although future research could address the limitations I mentioned above, this research illuminates several additional interesting questions that scholars might address to move our knowledge even further forward. For example, given that positive and
121 negative social relationships in teams are associated with team functioning and effectiveness, what predicts how these structural relationships form in the first place? The homophily principle suggests that relationships form between similar people at a higher rate than among dissimilar people ( McPherson, Smith Lovin, & Cook, 2001) Thus, the extent to which personal similarities and differences predict the formation of different types of social relationships in teams is worthy of continued inv estigation. Yet what characteristics are encapsulated by the term similar is often taken for granted without examining how individuals belong to multiple social groups simultaneously. As a result, individual identity characteristics on which to base similarity judgments can be defined in numerous ways, including in terms of gender, race/ethnicity, age, marital status, occupation, parental status, educational background, social class, place of residence, place of work, position relevant to others in a networ k (i.e. central, leader, or opinion leader), behavioral patterns (achievement versus delinquent), values, beliefs, attitudes, personality, etc. Indeed, self concept is a cognitive organization of various identities and attributes negotiated and developed f rom interactions in social roles and group memberships ( Gecas, 1982; Ibarra, 1999). Thus, in any context, the reasoning that similarity perceptions among actors increase the probability of tie formation between them (and thus the formation of a team networ k structure composed of individual dyadic linkages based on homophily effects) implicitly assumes that the researcher know s which aspects of identity have been considered as most important in the similarity comparison process of the individuals who are exploring potential tie formation. The likelihood of a network tie formation based on similarity of characteristics may depend in turn on how salient a given set of characteristics is to a particular
122 situation. In other words, in a knowledgeintensive work setting, similarity in intelligence may be a more relevant characteristic for tie formation than is race or gender. In a creative context, similarity in openness to experience may be more relevant than age or educational background. Thus, when examining homophilous network tie formation, researchers should be careful to investigate which types of characteristics are most salient to a given context and whether similarity of those characteristics versus less salient characteristics is a stronger driver of network tie formation. A second avenue of future research involves examining the implications of positive and negative multiplex relationships for other outcome criteria including team creativity, viability, member satisfaction, and customer satisfaction. It i s possible that the effects of structural cohesion and repulsion differ across the varying criteria. For example, it is possible that for team creative performance, the relationships are the opposite in that increasing structural cohesion actually stifles creativity through greater pressures toward conformity and consensus, which constrains innovation and novel thinking necessary to produce creative solutions. On the other hand, teams with increasing structural repulsion may exhibit greater willingness to c hallenge, argue, and fight with each other to the point where takenfor granted assumptions are reexamined and novel solutions emerge. Thus, whereas team task performance is benefitted from structural cohesion, the opposite may be true for team creative performance. However, while teams with members who have difficult and adversarial relationships may produce more innovative and creative solutions, these same teams may suffer from a lower degree of team member satisfaction and decreased viability because o f the unpleasantness of the experience of dealing with conflict and argument within the team. Thus, future research
123 would be useful that examines the implications of positive and negative relationships in teams beyond what I have considered here. A third area of future research involves investigating w hether social relationships matter more or less as teams become increasingly virtualized and geographically dispersed In this context it may be that friendship relationships become less relevant simply owing to the fact that fewer opportunities to socialize are available to a virtual or geographically disperse team. On the other hand, it may be that trust relationships become increasingly important because of the reduced capability of team members to monitor each other on a consistent basis. In regard to negative relationships, it may be that disliking relationships become less relevant since the virtual environment limits the frequency and duration with which team members must come in contact, and as a result team members are more easily able to tolerate limited instances of unpleasantness that would otherwise distract more heavily from task accomplishment. On the other hand, it may be possible that distrust relationships become even more dysf unctional for team effectiveness in virtual or geographically disperse team environments as members worry about the reliability of the work of other team members, and, because of the uncertainty, end up wasting time and effort duplicating efforts as a contingency. Finally, an area of future research more specific to the MBA student sample setting involves examining the extent to which social relationships formed during the MBA program translate into future collaboration and working relationships following graduation. Many s tudents enroll in an MBA program in part because of the education, but also in part because of the opportunity to expand their professional networks. MBA program admissions officers often tout the opportunity to network as one of the major
124 benefits of enro lling in a universitys program. For this reason, future research could collect detailed network data on the quality of the relationships MBA students form within their cohorts during the program of study, and follow up with students at regular intervals f ollowing graduation (e.g., two years, four years, etc.) with a list of the members from their original MBA cohort to ask with whom they currently collaborate, from whom have they obtained employment opportunities, and with whom have formed other professional relationships Observing the correlation between the quality of the relationships formed during the MBA program and the number and quality of future collaboration relationships would provide fascinating empirical evidence as to whether MBA programs are able to accomplish one of their purported goals of giving their students a new network that will become highly useful in the future. Conclusion This study is the first in the organizational literature to examine the relationships between multiplex instrume ntal ties with positive and negative expressive ties among team members and resulting team performance through a systematic set of synergistic team mechanisms I found that positive multiplex density termed structural cohesion, had a positive indirect rel ationship with team performance mainly through its association with increased teamwork S tructural cohesion was also associated with increased team engagement and reduced team maintenance, bit it did not exhibit indirect relationships with team performance through these mechanisms. On the other hand, negative multiplex density termed structural repulsion, had a negative indirect relationship with team performance mainly through its association with decreased teamwork Structural repulsion was also associat ed with decreased team engagement and increased team maintenance, but it did not exhibit indirect relationships with team performance through
125 these mechanisms. Taken together, these results show that the balance of positive and negative social structures a mong team members has important implications for team functioning and effectiveness.
126 LIST OF REFERENCES Ashforth, B. E., & Humphrey, R. H. (1995). Emotion in the workplace: A reappraisal. Human Relations, 48, 97 125. Aub, C., & Rousseau, V. (2005). Team goal commitment and team effectiveness: The role of task interdependence and supportive behaviors. Group Dynamics: Theory, Research, and Practice, 9, 189 204. Baker, W., Cross, R., & Wooten, M. (2003). Positive organizational network analysis and energizing relationships. In K. S. Cameron, J. E. Dutton & R. E. Quinn (Eds.), Positive organizational scholarship : Foundations of a new discipline (pp. 328342). San Francisco: Berrett Koehler Publishers, Inc. Bakker, A. B. (2009). Building engagement in t he workplace. In C. L. Cooper, & R. J. Burke (Eds.), The peak performing organization (pp. 50 72) Abingdon, UK : Routledge. Bakker, A. B., Demerouti, E., & Verbeke, W. (2004). Using the job demands resources model to predict burnout and performance. Human Resource Management, 43, 83104. Bakker, A. B., van Emmerik, H., & Euwema, M. C. (2006). Crossover of burnout and engagement in work teams. Work & Occupations, 33, 464 489. Baldwin, T. T., Bedell, M. D., & Johnson, J. L. (1997). The social fabric of a t eam based M.B.A. program: Network effects on student satisfaction and performance. Academy of Management Journal, 40 13691397. Bales, R. F. (1953). The equilibrium problem in small groups. In T. Parsons, R. F. Bales & E. A. Shils (Eds.), Working papers in the theory of action (pp. 111 161). New York: Free Press. Balkundi, P., Barsness, Z., & Michael, J. H. (2009). Unlocking the influence of leadership network structures on team conflict and viability. Small Group Research, 40, 301 322. Balkundi, P., & Harrison, D. A. (2006). Ties, leaders, and time in teams: Strong inference about network structure's effects on team viability and performance. Academy of Management Journal, 49 49 68. Balkundi, P., Kilduff, M., Barsness, Z. I., & Michael, J. H. (2007). Demographic antecedents and performance consequences of structural holes in work teams. Journal of Organizational Behavior, 28, 241 260. Bandura, A. (1982). Self efficacy mechanism in human agency. American Psychologist, 37, 122147.
127 Baron, R. M., & Kenny, D. A. (1986). The moderator mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 11731182. 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, 377 391. Beal, D. J., Cohen, R. R., Burke, M. J., & McLendon, C. L. (2003). Cohesion and performance in gr oups: A metaanalytic clarification of construct relations. Journal of Applied Psychology, 88, 989 1004. Bion, W. R. (1961). Experiences in groups New York: Basic Books. Bishop, J. W., & Scott, K. D. (2000). An examination of organizational and team com mitment in a self directed team environment. Journal of Applied Psychology, 85, 439450. Bliese, P. D. (2000). Withingroup agreement, nonindependence, and reliability: Implications for data aggregation and analysis. In K. J. Klein, & S. W. J. Kozlowski (Eds.), Multilevel theory, research, and methods in organizations: Foundations, extensions, and new directions. (pp. 349381). San Francisco: Jossey Bass. Bond, M. A. (2007). Workplace chemistry: Promoting diversity through organizational change. Lebanon, NH: University Press of New England. Borgatti, S. P., & Foster, P. C. (2003). The network paradigm in organizational research: A review and typology. Journal of Management, 29, 9911013 Bourdieu, P. (1985). The forms of capital. In J. G. Richardson (Ed.), Handbook of theory and research for the sociology of education (pp. 241258). New York: Greenwood. Bradley, J., White, B. J., & Mennecke, B. E. (2003). Teams and tasks. Small Group Research, 34, 353 387. Brass, D. J. (1992). Power in organizations: A social network perspective. In G. Moore, & J. A. Whitt (Eds.), Research in politics and society ( V ol. 4, pp. 295 323). Greenwich, CT: Jai Press. Brass, D. J., Galaskiewicz, J., Greve, H. R., & Tsai, W. (2004). Taking stock of networks and organizations: A multilevel perspective. The Academy of Management Journal, 47, 795 817. Brass, D. J. (1995). A social network perspective on human resources management. Research in Personnel and Human Resources Management, 13, 3979.
128 Brown, S. P., & Leigh, T. W. (1996). A new look at psychological climate and its relationship to job involvement, effort, and performance. Journal of Applied Psychology, 81 358368. Burt, R. S. (1990). Kinds of relations in american discussion networks. In C. Calhoun, M. W. Meyer & W. R. Scott (Eds.), Structures of power and constraint: Papers in honor of P eter M. B lau (pp. 411452). Cambridge, MA: Cambridge University Press. Burt, R. S. (1992). Structural holes: The social structure of competition. Cambridge, Mass.: Harvard University Press. Burt, R. S., & Schtt, T. (1985). Relation contents in multiple networks. Social Science Research, 14, 287 308. Campion, M. A., Medsker, G. J., & Higgs, A. C. (1993). Relations between work group characterist ics and effectiveness: Implications for designing effective work groups. Personnel Psychology, 46, 823 850. Campion, M. A., Papper, E. M., & Medsker, G. J. (1996). Relations between work team characteristics and effectiveness: A replication and extension. Personnel Psychology, 49 429452. Casciaro, T., & Lobo, M. S. (2008). When competence is irrelevant: The role of interpersonal affect in task related ties. Administrative Science Quarterly, 53, 655 684. Chan, D. (1998). Functional relations among const ructs in the same content domain at different levels of analysis: A typology of composition models. Journal of Applied Psychology, 83 234246. Cohen, S. (1980). Aftereffects of stress on human performance and social behavior: A review of research and the ory. Psychological Bulletin, 88, 82 108. Coleman, J. S. (1988). Social capital in the creation of human capital. The American Journal of Sociology, 94, S95 S120. Crawford, E. R., LePine, J. A., & Rich, B. L. ( 2010). Linking job demands and resources to employee engagement and burnout: A theoretical extension and metaanalytic test. Journal of Applied Psychology, 95, 834 848. Cummings, J. N., & Cross, R. (2003). Structural properties of work groups and their consequences for performance. Social Networks, 25, 197 210. Demerouti, E., Bakker, A. B., Nachreiner, F., & Schaufeli, W. B. (2001). The job demands resources model of burnout. Journal of Applied Psychology, 86, 499 512.
129 DeShon, R. P., Kozlowski, S. W. J., Schmidt, A. M., Milner, K. R., & Wiechmann, D. (2004). A multiplegoal, multilevel model of feedback effects on the regulation of individual and team performance. Journal of Applied Psychology, 89, 10351056. Devine, D. J., Clayton, L. D., Philips, J. L., Dunford, B. B., & Melner, S. B. (1999). Tea ms in organizations prevalence, characteristics, and effectiveness. Small Group Research, 30 678 711. Dobbins, G. H., & Zaccaro, S. J. (1986). The effects of group cohesion and leader behavior on subordinate satisfaction. Group & Organization Studies, 11, 203219. Druskat, V. U. & Kayes, D. C. (2000). Learning versus performance in short term project teams. Small Group Research, 31, 328 353. Dutton, J. E., & Heaphy, E. D. (2003). The power of highquality connections. In K. S. Cameron, J. E. Dutton & R. E. Quinn (Eds.), Positive organizational scholarship: Foundations of a new discipline (pp. 263 278). San Francisco: Berrett Koehler Publishers, Inc. Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44, 350 383. Elfenbein, H. A., & O'Reilly III, C. A. (2007). Fitting in: The effects of relational demography and personculture fit on group process and performance. Group & Organization Management, 32, 109142. Festinger, L. (1950). Informal social communication. Psychological Review, 57, 271 282. Fischer, C. S. (1982). What do we mean by friend? A n inductive study. Social Networks, 3 287 306. Gauthier Villars, D. (2010, June 22). World news: France's bad case of L es B leus. Wall Street J ournal, p. A.12. Gecas, V. (1982). The self concept. Annual Review of Sociology, 8, 1 33. Gersick, C. J. G. (1988). Time an d transition in work teams: Toward a new model of group development. Academy of Management Journal, 31, 9 41. Gersick, C. J. G., Du tton, J. E., & Bartunek, J. M. (2000). Learning from academia: The importance of relationships in professional life. Academy of Management Journal, 43, 10261044. Gittell, J. H. (2003). A theory of relational coordination. In K. S. Cameron, J. E. Dutton & R. E. Quinn (Eds.), Positive organizational scholarship: Foundations of a new discipline (pp. 279 295). San Francisco: Berrett Koehler Publishers, Inc.
130 Glass, H. E., & Beaudry, D. R. (2008). Key factors in CRO selection. Applied Clinical Trials, 17 52 6 0. Gluckman, M. (1967). The judicial process among the B arotse of n orthern R hodesia ( Z ambia) (2nd ed.). Manchester: Manchester University Press. Goffman, E. (1961). Asylums New York: Doubleday Anchor. Goodman, P. S., Ravlin, E., & Schminke, M. (1987). Understanding groups in organizations. In L. L. Cummings, & B. M. Staw (Eds.), Research in organizational behavior (Vol. 9 pp. 121 173). Greenwich, CT: JAI Press. Gordon, J. (1992). Work teams: How far have they come? Training, 29, 59 65. Gould, D., Greenleaf, C., Guinan, D., & Chung, Y. (2002). A survey of U.S. Olympic coaches: Variables perceived to have influenced athlete performances and coach effectiveness. The Sport Psychologist, 16, 229 250. Granovetter, M. S. (1973). The strength o f weak ties. American Journal of Sociology, 78, 13601380. Greve, H. R., Baum, J. A. C., Mitsuhashi, H., & Rowley, T. J. (2010). Built to last but falling apart: Cohesion, friction, and withdrawal from interfirm alliances. Academy of Management Journal, 53 302 322. Guzzo, R. A., & Shea, G. P. (1992). Group performance and intergroup relations in organizations. In M. D. Dunnette, & L. M. Hough (Eds.), Handbook of industrial and organizational psychology ( Vol. 3 pp. 269 313). Palo Alto, CA: Consulting Psy chologists Press. Guzzo, R. A., & Dickson, M. W. (1996). Teams in organizations: Recent research on performance and effectiveness. Annual Review of Psychology, 47, 307 338. Guzzo, R. A., Yost, P. R., Campbell, R. J., & Shea, G. P. (1993). Potency in grou ps: Articulating a construct. British Journal of Social Psychology, 32, 87 106. Hackman, J. R. (1987). The design of work teams. In J. W. Lorsch (Ed.), Handbook of organizational behavior (pp. 315342). Englewood Cliffs, NJ: PrenticeHall. Hanneman, R. A ., & Riddle, M. (2005). Introduction to social network methods Riverside, CA: University of California, Riverside (published in digital form at http://faculty.ucr.edu/~hanneman/nettext/ ) Hansen, M. T. (1999). The searchtransfer problem: The role of weak ties in sharing knowledge across organization subunits. Administrative Science Quarterly, 44, 82 111.
131 Harrison, D. A., & Klein, K. J. (2007). What's the difference? D iversity constructs as separati on, variety, or disparity in organizations. The Academy of Management Review, 32, 11991228. Hirschfeld, R. R., Jordan, M. H., Feild, H. S., Giles, W. F., & Armenakis, A. A. (2005). Teams' female representation and perceived potency as inputs to team outc omes in a predominantly male field setting. Personnel Psychology, 58, 893 924. Hobfoll, S. E., & Freedy, J. (1993). Conservation of resources: A general stress theory applied to burnout. In W. B. Schaufeli, C. Maslach & T. Marek (Eds.), Professional burno ut: Recent developments in theory and research (pp. 115 133). Philadelphia, PA: Taylor & Francis. Hochschild, A. R. (1983). The managed heart: Commercialization of human feeling. Berkeley: University of California Press. Hoerr, J. (1989, July 10). The payoff from teamwork. BusinessWeek, 5662. Ibarra, H. (1993). Personal networks of women and minorities in management: A conceptual framework. Academy of Management Review, 18, 56 87. Ibarra, H. (1995). Race, opportunity, and diversity of social circles in managerial networks. Academy of Management Journal, 38, 673703. Ibarra, H. ( 1999 ). Provisional selves: Experimenting with image and identity in professional adaptation. Administrative Science Quarterly, 44, 764 791. Ibarra, H., & Andrews, S. B. (1993). Power, social influence, and sense making: Effects of network centrality and proximity on employee perceptions. Administrative Science Quarterly, 38 277303. Ilgen, D. R., Hollenbeck, J. R., Johnson, M., & Jundt, D. (2005). Teams in organizations: From i nput process output models to IMOI models. Annual Review of Psychology, 56, 517543. Ingram, P., & Zou, X. (2008). Business friendships. Research in Organizational Behavior, 28 167 184. James, L. R., Demaree, R. G., & Wolf, G. (1984). Estimating within group interrater reliability with and without response bias. Journal of Applied Psychology, 69, 85 98. Jehn, K. A. (1995). A multimethod examination of the benefits and detriments of intragroup conflict. Administrative Science Quarterly, 40, 256282. Jehn, K. A., Northcraft, G. B., & Neale, M. A. (1999). Why differences make a difference: A field study of diversity, conflict, and performance in workgroups. Administrative Science Quarterly, 44, 741 763.
132 Jehn, K. A., & Shah, P. P. (1997). Interpersonal relationships and task performance: An examination of mediating processes in friendship and acquaintance groups. Journal of Personality & Social Psychology, 72, 775 790. Jett, Q. R., & George, J. M. (2003). Work interrupted: A closer look at the role of i nterruptions in organizational life. Academy of Management Review, 28, 494 507. Kahn, W. A. (1990). Psychological conditions of personal engagement and disengagement at work. Academy of Management Journal, 33, 692 724. Kahn, W. A. (1992). To be fully ther e: Psychological presence at work. Human Relations, 45, 321 349. Kanfer, R. (1990). Motivation theory and industrial and organizational psychology. In L. M. Hough (Ed.), Handbook of industrial and organizational psychology ( Vol. 1, 2nd ed., pp. 75170). P alo Alto, CA: Consulting Psychologists Press. Karau, S. J., & Williams, K. D. (1993). Social loafing: A metaanalytic review and theoretical integration. Journal of Personality & Social Psychology, 65, 681 706. Katz, D., & Kahn, R. L. (1978). The social psychology of organizations New York: John Wiley & Sons. Kolodny, H. F., & Kiggundu, M. N. (1980). Towards the development of a sociotechnical systems model in woodlands mechanical harvesting. Human Relations, 33, 623 645. Kozlowski, S. W. J., Gully, S., Nason, E., & Smith, E. (1999). Developing adaptive teams: A theory of compilation and performance across levels and time. In D. R. Ilgen, & E. D. Pulakos (Eds.), The changing nature of work performance: Implications for staffing, personnel actions, and development (pp. 240 292). San Francisco: Jossey Bass. Kozlowski, S. W. J., & Bell, B. S. (2003). Work groups and teams in organizations. In W. C. Borman, D. R. Ilgen & R. J. Klimoski (Eds.), Handbook of psychology: Industrial and organizational psychology ( Vol. 12, pp. 333375). Hoboken, NJ: John Wiley & Sons, Inc. Kozlowski, S. W. J., & Klein, K. J. (2000). A multilevel approach to theory and research in organizations: Contextual, temporal, and emergent processes. In K. J. Klein, & S. W. J. Kozlowski (Ed s.), Multilevel theory, research, and methods in organizations: Foundations, extensions, and new directions (pp. 390). San Francisco: Jossey Bass. Krackhardt, D., & Stern, R. N. (1988). Informal networks and organizational crises: An experimental simulat ion. Social Psychology Quarterly, 51, 123140.
133 Labianca, G., & Brass, D. J. (2006). Exploring the social ledger: Negative relationships and negative asymmetry in social networks in organizations. Academy of Management Review, 31, 596614. Labianca, G., B rass, D. J., & Gray, B. (1998). Social networks and perceptions of intergroup conflict: The role of negative relationships and third parties. Academy of Management Journal, 41, 55 67. Langfred, C. W. (2007). The downside of self management: A longitudinal study of the effects of conflict on trust, autonomy, and task interdependence in self managing teams. Academy of Management Journal, 50, 885900. Latan, B., Williams, K., & Harkins, S. (1979). Many hands make light the work: The causes and consequences of social loafing. Journal of Personality and Social Psychology, 37 822832. Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping New York: Springer Publishing Co. Leavitt, H. J. (1951). Some effects of certain communication patterns on g roup performance. The Journal of Abnormal and Social Psychology, 46 38 50. Lee, R. T., & Ashforth, B. E. (1996). A meta analytic examination of the correlates of the three dimensions of job burnout. Journal of Applied Psychology, 81, 123133. LeP ine, J. A., Podsakoff, N. P., & Le P ine, M. A. (2005). A metaanalytic test of the challenge stressor hindrance stressor framework: An explanation for inconsistent relationships among stressors and performance. Academy of Management Journal, 48, 764 775. Levine, J. M., & Moreland, R. L. (1990). Progress in small group research. Annual Review of Psychology, 41, 585 634. Lincoln, J. R., & Miller, J. (1979). Work and friendship ties in organizations: A comparative analysis of relational networks. Administrative Scie nce Quarterly, 24, 181199. Lindsley, D. H., Brass, D. J., & Thomas, J. B. (1995). Efficacy performing spirals: A multilevel perspective. Academy of Management Review, 20 645 678. Locke, E. A., & Latham, G. P. (1990). A theory of goal setting & task per formance. Englewood Cliffs, NJ : Prentice Hall, Inc. Lucius, R. H., & Kuhnert, K. W. (1997). Using sociometry to predict team performance in the work place. Journal of Psychology: Interdisciplinary and Applied, 131 21 32.
134 MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7, 83 104. Manz, C. C., & Sims, H. P. (1993). Business without bosses: How self managing teams are building highperforming companies New York: John Wiley & Sons Inc. Marks, M. A., DeChurch, L. A., Mathieu, J. E., Panzer, F. J., & Alonso, A. (2005). Teamwork in multiteam systems. Journal of Applied Psychology, 90, 964 971. Marks, M. A., Mathieu, J. E., & Zaccaro, S. J. (2001). A temporally based framework and taxonomy of team processes. Academy of Management Review, 26, 356 376. Marsden, P. V., & Campbell, K. E. (1984). Measuring tie strength. Social Forces, 63, 482501. Maslach, C. (1993). Burnout: A multidimensional perspective. In W. B. Schaufeli, C. Maslach & T. Marek (Eds.), Professional burnout: Recent developments in theory and research (pp. 1932). Philadelphia, PA: Taylor & Francis. Maslach, C., & Leiter, M P. (1997). The truth about burnout: How organizations cause personal stress and what to do about it San Francisco: Jossey Bass. Mathieu, J. E., & Marks, M. A. (2006). Team process items. Unpublished. Mathieu, J. E., Gilson, L. L., & Ruddy, T. M. (2006). Empowerment and team effectiveness: An empirical test of an integrated model. Journal of Applied Psychology, 91 97 108. Mathieu, J. E., & Schulze, W. (2006). The influence of team knowledge and formal plans on episodic team process performance relationships. Academy of Management Journal, 49, 605619. Mathieu, J., Maynard, M. T., Rapp, T., & Gilson, L. (2008). Team effectiveness 19972007: A review of recent advancements and a glimpse into the future. Journal of Management, 34 410 476. McGrat h, J. E. (1964). Social psychology: A brief introduction. New York: Holt, Rinehart and Winston. McGrath, J. E. (1997). Small group research, that once and future field: An interpretation of the past with an eye to the future. Group Dynamics: Theory, Research, and Practice, 1, 7 27. McGrath, J. E., Arrow, H., & Berdahl, J. L. (2000). The study of groups: Past, present, and future. Personality & Social Psychology Review, 4, 95 105.
135 McGrath, J. E. (1984). Groups: Interaction and performance. Englewood Cliff s, N.J.: PrenticeHall. McGregor, J. (2006, August 21). Game plan: First find the leaders. BusinessWeek Retrieved on June 17, 2010, from http://www.businessweek.com /magazine/content/06_34/b3998437.htm McPherson, M., SmithLovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27, 415 444. Mehra, A., Dixon, A. L., Brass, D. J., & Robertson, B. (2006). The social network ties of group leaders: Implications for group performance and leader reputation. Organization Science, 17, 64 79. Methot, J. R. (2010). The effects of instrumental, friendship, and multiplex network ties on job performance: A model of coworker relationships. Unpublished doctoral dissertation, University of Florida, Gainesville. Morgeson, F. P., & Hofmann, D. A. (1999). The structure and function of collective constructs: Implications for multilevel research and theory development. Academy of Management Review, 24, 249 265. Mowday, R. T., Steers, R. M., & Porter, L. W. (1979). The measurement of organizational commitment. Journal of Vocational Behavior, 14, 224247. Mullen, B., & Copper, C. (1994). The relation between group cohesiveness and perf ormance: An integration. Psychological Bulletin, 115, 210 227. Muthn, L. K., & Muthn, B. O. (19982007). Mplus user's guide. (Fifth ed.). Los Angeles, CA: Muthn & Muthn. Oh, H., Chung, M., & Labianca, G. (2004). Group social capital and group effecti veness: The role of informal socializing ties. Academy of Management Journal, 47 860 875. Oh, H., Labianca, G., & Chung, M. (2006). A multilevel model of group social capital. The Academy of Management Review, 31, 569 582. O'Leary Kelly, A. M., Martocch io, J. J., & Frink, D. D. (1994). A review of the influence of group goals on group performance. Academy of Management Journal, 37, 12851301. O'Reilly III, C. A., Caldwell, D. F., & Barnett, W. P. (1989). Work group demography, social integration, and turnover. Administrative Science Quarterly, 34, 21 37. Parise, S., & Rollag, K. ( 2010). Emergent network structure and initial group performance: The moderating role of preexisting relationships. Journal of Organizational Behavior 31 877 897
136 Parker, D. F., & DeCotiis, T. A. (1983). Organizational determinants of job stress. Organizational Behavior & Human Performance, 32, 160 177. Peters, L. H., O'Connor, E. J., & Rudolf, C. J. (1980). The behavioral and affective consequences of performancerelevant situational variables. Organizational Behavior & Human Performance, 25 79 96. Podolny, J. M. (2001). Networks as the pipes and prisms of the market. American Journal of Sociology, 107, 33 60. Podolny, J. M., & Baron, J. N. (1997). Resources and relationships: Social networks and mobility in the workplace. American Sociological Review, 62, 673 693. Pondy, L. R. (1967). Organizational conflict: Concepts and models Administrative Science Quarterl y, 12 296320 Porath, C. L., & Erez, A. (2009). Overlooked but not untouched: How rudeness reduces onlookers performance on routine and creative tasks. Organizational Behavior & Human Decision Processes, 109, 29 44. Portes, A. (1998). Social capital: Its origins and applications in modern sociology. A nnual Review of Sociology, 24, 1 24. Reagans, R., Argote, L., & Brooks, D. (2005). Individual experience and experience working together: Predicting learning rates from knowing who knows what and knowing how to work together. Management Science, 51 869 8 81. Reagans, R., & Zuckerman, E. W. (2001). Networks, diversity, and productivity: The social capital of corporate R&D teams. Organization Science, 12, 502 517. Reagans, R., Zuckerman, E., & McEvily, B. (2004). How to make the team: Social networks vs. d emography as criteria for designing effective teams. Administrative Science Quarterly, 49 101133. Rich, B. L., LePine, J. A., & Crawford, E. R. (2010). Job engagement: Antecedents and effects on job performance. Academy of Management Journal, 53, 617 635. Rosenthal, E. (1997). Social networks and team performance. Team Performance Management, 3, 288 294. Rousseau, D. M. (1977). Technological differences in job characteristics, employee satisfaction, and motivation: A synthesis of job design res earch and sociotechnical systems theory. Organizational Behavior & Human Performance, 19 18 42. Rubin, J. Z., Pruitt, D. G., & Kim, S. H. (1994). Social conflict: Escalation, stalemate, and settlement (2nd ed.) New York, NY England: Mc G raw Hill Book Com pany.
137 Salanova, M., Agut, S., & Peir, J. M. (2005). Linking organizational resources and work engagement to employee performance and customer loyalty: The mediation of service climate. Journal of Applied Psychology, 90, 12171227. Salanova, M., Llorens, S., Cifre, E., Martnez, I. M., & Schaufeli, W. B. (2003). Perceived collective efficacy, subjective well being, and task performance among electronic work groups: An experimental study. Small Group Research, 34, 43 73. Schutz, W. C. (1967). Joy: Expandi ng human awareness New York: Grove Press. Scott, J. (2000). Social network analysis: A handbook Thousands Oaks, CA: SAGE Publications. Shaw, M. E. (1964). Communication networks. In L. Nerkowitz (Ed.), Advances in experimental social psychology (Vol. 1 pp. 111 147). New York: Academic Press. Shea, G. P., & Guzzo, R. A. (1987). Group effectiveness: What really matters? Sloan Management Review, 28, 25 31. Sparrowe, R. T., Liden, R. C., Wayne, S. J., & Kraimer, M. L. (2001). Social networks and the perf ormance of individuals and groups. Academy of Management Journal, 44, 316325. Stajkovic, A. D., Lee, D., & Nyberg, A. J. (2009). Collective efficacy, group potency, and group performance: Metaanalyses of their relationships, and test of a mediation mode l. Journal of Applied Psychology, 94, 814 828. Steiner, I. D. (1972). Group process and productivity New York: 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, 43 135 148. Stewart, G. L., Fulmer, I. S., & Barrick, M. R. (2005). An exploration of member roles as a multilevel linking mechanism for individual traits and team outcomes. Personnel Psychology, 58, 343365. Stokes, J. P. (1983). Components of group cohesion: Intermember attraction, instrumental value, and risk taking. Small Group Research, 14, 163173. Taylor, S. E. (1991). Asymmetrical effects of positive and negative events: The mobil ization minimization hypothesis. Psychological Bulletin, 110 6785. Tesluk, P., Mathieu, J. E., Zaccaro, S. J., & Marks, M. (1997). Task and aggregation issues in the analysis and assessment of team performance. In M. T. Brannick, E. Salas & C. Prince (E ds.), Team performance assessment and measurement: Theory, methods, and applications (pp. 197 224). Mahwah, NJ: Erlbaum.
138 Thompson, J. D. (1967). Organizations in action. New York: McGraw Hill. Tichy, N. M., Tushman, M. L., & Fombrun, C. (1979). Social network analysis for organizations. Academy of Management Review, 4, 507 519. Trist, E. L. (1981). The evolution of sociotechnical systems: A conceptual framework and an action research program Toronto: Ontario Quality of Working Life Centre. Tsai, W. (2001). Knowledge transfer in intraorganizational networks: Effects of network position and absorptive capacity on business unit innovation and performance. Academy of Management Journal, 44 996 1004. Van de Ven, A. H., Delbecq, A. L., & Koenig Jr, R. (1976). Determinants of coordination modes within organizations. American Sociological Review, 41, 322338. Wageman, R. (2001). How leaders foster self managing team effectiveness: Design choices versus hands on coaching. Organization Science, 12 559 577 Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications New York: Cambridge University Press. Weick, K. E., & Roberts, K. H. (1993). Collective mind in organizations: Heedful interrelating on flight decks. Administrative S cience Quarterly, 38, 357 381. Winstead, B. A., Derlega, V. J., Montgomery, M. J., & Pilkington, C. (1995). The quality of friendships at work and job satisfaction. Journal of Social and Personal Relationships, 12, 199215. Wiseman, J. P., & Duck, S. (199 5). Having and managing enemies: A very challenging relationship. In S. Duck, & J. T. Wood (Eds.), Confronting relationship challenges (pp. 4372). Thousands Oaks, CA: Sage. Wright, P. H. (1984). Self referent motivation and the intrinsic quality of frien dship. Journal of Social and Personal Relationships, 1, 115130. Yukelson, D. (1997). Principles of effective team building interventions in sport: A direct services approach at P enn S tate U niversity. Journal of Applied Sport Psychology, 9 73 96. Zaheer, A., & Soda, G. (2009). Network evolution: The origins of structural holes. Administrative Science Quarterly, 54, 1 31.
139 BIOGRAPHICAL SKETCH Eean R. Crawford received his Ph.D. in organizational behavior/human resource m anagement from the Warrington College of Business Administration at the University of Florida He is currently an assistant professor in the Tippie College of Business at the University of Iowa. His research focuses on individuals willingness to personal ly invest themselves in their work roles, and how the complexity of managing interpersonal relationships, especially in team settings, influences that investment. He explores this interest through several topics including employee engagement, team effectiv eness, personality, and social networks. He has published his research in Academy of Management Journal Journal of Applied Psychology and Human Resource Management Review He serves as a reviewer for Journal of Management Applied Psychology: An International Review and Group & Organization Management He is a member of the Academy of Management and the Society for Industrial and Organizational Psychology.