J Bus Psychol DOI 10.1007/s10869-016-9481-6
ORIGINAL PAPER
Are Pre-Assembly Shared Work Experiences Useful for Temporary-Team Assembly Decisions? A Study of Olympic Ice Hockey Team Composition Dev K. Dalal 1 & Kevin P. Nolan 2 & Lauren E. Gannon 3
# Springer Science+Business Media New York 2016
Abstract The purpose of this study was to investigate if preassembly shared work experiences among temporary team members facilitate individual and team performance. Archival data from the 2014 Men’s Olympic Ice Hockey Tournament (12 teams, 25 players each) was used in the study. Measures of social network centrality were computed based on the pre-assembly shared work experiences among national team members derived from professional and amateur affiliations. These measures were used to predict objective individual and team performance criteria. Players’ closeness centrality scores, from pre-assembly shared work experience networks, positively predicted their goals, assists, and being involved in more positive than negative plays. Teams with less centralized pre-assembly shared work experience network structures tended to perform better than teams with more centralized pre-assembly shared work experience network structures. Temporary teams are commonly used by organizations to perform tasks that are specific, important, and of short duration. Because temporary teams have little time to develop the shared properties required for effective team functioning, * Dev K. Dalal
[email protected] Kevin P. Nolan
[email protected] Lauren E. Gannon
[email protected] 1
Department of Psychology, University at Albany, State University of New York, Albany, NY, USA
2
Department of Psychology, Hofstra University, Hempstead, NY, USA
3
Department of Psychological Sciences, University of Connecticut, Storrs, CT, USA
assembly decisions are of paramount importance. The results from this study suggest that centrality measures derived from pre-assembly shared work experiences are useful to consider when assembling temporary teams. Few studies have investigated the impact of pre-assembly shared work experiences on individual and team performance. Using objective data, this study provides evidence that pre-assembly shared work experiences relate to individual and team performance in temporary teams, supporting the need for expanded research in this area. Keywords Temporary teams . Team assembly . Team composition . Social network analysis . Sports data Driven by economic, strategic, and technological imperatives, the twenty-first century has witnessed a remarkable transformation from work organized around individual jobs to teambased work structures (Kozlowski and Bell 2013; Lawler et al. 1995). Whereas approximately 15 years ago, one half of all organizations utilized team-based work structures (Devine et al. 1999), today, an estimated 95 % of workers are members of one or more work teams (O’Leary et al. 2011). In short, team-based work has become the default method by which organizations conduct business (Hollenbeck and Jamieson 2015; Tannenbaum et al. 2012). It is important to note that not all team-based work is the same, nor are all teams similar in structure or function. Rather, various types of teams exist, differing in regard to fundamental features. A key distinguishing feature among teams is whether or not members remain together after tasks have been completed (Devine et al. 1999; Hollenbeck et al. 2012; Marks et al. 2001). Whereas some teams are formed to work together continuously on multiple projects with minimal turnover among members, other teams are formed to complete specific
J Bus Psychol
tasks in a finite amount of time with team members who may never reassemble in their entirety after disbanding (Tannenbaum et al. 2012). Temporary teams are becoming increasingly popular given that their dynamic nature allows for more purposeful combinations of members to accomplish specific tasks (Contractor 2013; Tannenbaum et al. 2012). As a result, it is common for individual workers to belong to multiple teams (O’Leary et al. 2011), each composed of members with varying amounts of experience working with one another (Contractor 2013; Tannenbaum et al. 2012). Like all teams, temporary teams must coordinate1 their actions effectively in order to be successful. The finite time during which temporary teams work together, however, significantly limits their ability to develop the kinds of processes required for effective coordination (Contractor 2013). Therefore, temporary teams must quickly learn how to work together in order to be successful (Tannenbaum et al. 2012). Further complicating coordination efforts, research suggests that simply assembling the highest performing individuals into a temporary team does not necessarily result in the highest performing team (Contractor 2013). Rather, features beyond team member knowledge, skills, and abilities must be considered during temporary team assembly (Guimera et al. 2005; Hollenbeck and Jamieson 2015; Uzzi et al. 2013). Because temporary teams must quickly learn to coordinate their efforts, we propose that one way to facilitate individual- and team-level performance is to leverage the work experiences shared among potential team members prior to team assembly when making composition decisions. Shared work experiences are collaborative performance episodes wherein individuals work together on tasks, therein becoming familiar with each other’s knowledge, skills, abilities, characteristics, and behavioral tendencies. These experiences differ from other forms of social interaction in that they develop specifically through work, and ultimately result in a greater understanding of one another’s work-related attributes (e.g., knowledge, skills, abilities, characteristics) and behavioral tendencies. As such, pre-assembly shared work experiences are distinct from, and arguably stronger than, workplace social or communication connections (i.e., not just communicating with or liking a coworker). Although these latter forms of connections may play a role in shared work experiences, just knowing, liking, or communicating with someone in the workplace are relatively weak forms of shared work experience. Shared work experiences vary in terms of quality, with 1 Throughout the manuscript, we use the term coordinate (and coordination) to encapsulate all cognitive, affective, and behavioral processes involved with team members successfully working as a unit to achieve their goals. Although this captures an array of processes, we use this single term as descriptive for ease of presentation. We recognize that various coordination processes will relate differentially to pre-assembly shared experiences—we highlight some of these differences in the Discussion section.
higher-quality experiences (i.e., collaborating closely with someone) resulting in greater familiarity than lower-quality experiences (i.e., observing others work). Two marketers who work together on the creation of an advertising campaign, for example, are likely to have a greater understanding of each other’s work-related knowledge, skills, abilities, and behavioral tendencies than two marketers who have only observed the campaigns the other has created because former pair’s collaboration has a greater effect on familiarity, making it a higher-quality shared work experience. If high-quality shared work experiences facilitate familiarity and familiarity bolsters team members’ ability to coordinate efforts, then considering the pre-assembly shared work experiences among potential members during temporary team assembly may enhance individual and team performance in temporary teams (Cummings and Kiesler 2008). Using a sample of men’s Olympic ice hockey teams, this study tests the proposition that the quality of pre-assembly shared work experiences among temporary team members influences individual- and team-level performance. The results of this study contribute to the theory of temporary teams by identifying an antecedent of performance that is believed to facilitate the development of coordinating processes that traditionally require time to develop (Ilgen et al. 2005). Likewise, the methodology of this study provides organizations with a technique that can be used to quantify the pre-assembly shared work experiences among potential temporary team members during assembly decisions.
Temporary Team Characteristics Temporary teams are assembled to accomplish a specific task, to be completed within a finite timeframe, and whose members are unlikely to reassemble in their entirety once the task has been completed (Altschuller and Benbunan-Fich 2010; Contractor 2013; Devine et al. 1999). As a way to tackle novel, innovative, and dynamic problems, temporary teams are replacing more traditional team structures—teams whose membership is constant and who work on multiple tasks together (Contractor 2013; Tannenbaum et al. 2012). Temporary teams come in a variety of forms (e.g., action teams, flash teams, ad hoc committees) and are assembled for a range of purposes. Medical trauma teams, for example, are composed of doctors, nurses, technicians, and aides who assemble to perform specific, urgent, and unpredictable procedures (Klein et al. 2006). Disaster response teams assemble police, fire, medical, and city/state personnel to address prevailing emergencies (Tannenbaum et al. 2012). Ad hoc committees are used to address a range of issues. The Administrative Offices of the United States Courts, for example, sponsors an ad hoc committee of geographically dispersed judges,
J Bus Psychol
lawyers, academics, and administrators to review the Criminal Justice Act Program (U.S. Courts 2015). This study examines temporary team dynamics using a sample of men’s Olympic ice hockey teams. As such, it is instructive to detail the key features shared among this type of temporary team and those commonly employed in organizations. First, Olympic ice hockey teams assemble for a specific, time-constrained purpose—a 2-week long tournament that is held every 4 years. Second, Olympic ice hockey teams spend little time together before they must begin working on their designated task. In 2014, for example, teams with members who played in the National Hockey League (NHL) were unable to practice together as complete units until just 2 days before the start of the tournament (Iverson 2014). Third, Olympic ice hockey teams are unlikely to reassemble in their entirety after their task (i.e., the Olympic tournament) has concluded. Although some players will appear in more than one Olympics, the 4-year period between Olympic games makes it highly unlikely that the same set of players will represent their country across multiple tournaments. Finally, the nature of the task in which Olympic ice hockey teams engage matches other temporary teams in terms of being dynamic, requiring intensive coordinated effort, and being of short duration. Consequently, Olympic ice hockey players must develop awareness of their teammates’ actions (Bourbousson et al. 2015), adapt to teammates’ playing styles (Barth et al. 2015), and establish leadership roles within the team (Fransen et al. 2015) in order to be successful. Based on this common set of features, Olympic ice hockey teams represent a viable sample for exploring the relationship between pre-assembly shared work experiences and temporary team performance, particularly given that this study is an initial investigation of a new effect (Zhu et al. 2015).
Facilitating Temporary Team Performance As the use of temporary teams has risen in organizations, team composition has become less stable and more varied (Tannenbaum et al. 2012). Members are commonly assembled and reassembled into teams as needed, and members may or may not have ever worked together in the past (Contractor 2013). Consequently, temporary teams are regularly composed of members with different specializations, personal characteristics, and work styles that must learn to coordinate their efforts effectively in a limited amount of time. As Tannenbaum et al. (2012) noted, BOrganizations need to find ways to accelerate team readiness and help orient and prepare new team members^ because Bteam are often launched quite quickly and new team members frequently join teams in progress.^ (p.9) Given their unique structure, the most useful way to accelerate temporary team readiness may be through composition decisions (Hollenbeck and Jamieson 2015).
Composition decisions play an integral role in providing teams with the resources (i.e., inputs) they need to be successful. According to the Inputs-Mediators-Outputs-Inputs (IMOI) model of team effectiveness (Ilgen et al. 2005), team member characteristics (inputs) influence team processes (mediators), which in turn influence performance (outputs). For teams working across multiple performance episodes, the shared perceptions, normative expectations, and compatible knowledge that result from initial episodes (outputs) become resources (inputs) that later serve to shape and refine the team processes employed in subsequent episodes (Kozlowski et al. 1999; Marks et al. 2001). In this way, effective team processes develop over time, with team performance influenced by how quickly the processes develop (Ilgen et al. 2005; Tannenbaum et al. 2012). Whereas traditionally stable teams have the ability to develop effective team processes over repeated performance episodes (Marks et al. 2001), temporary teams’ ability to develop in this way is meaningfully restricted by the duration of their life cycle. Consequently, compared to traditional teams, temporary teams must leverage their inputs more quickly and effectively in order to develop the processes required for successful performance (Tannenbaum et al. 2012). The tenets of the IMOI model suggest that the correct combination of team member inputs should facilitate quicker development of effective team processes. For this reason, assembly decisions are of paramount importance for temporary teams (Contractor 2013; Tannenbaum et al. 2012). Research on team assembly has historically focused on manifest or descriptive characteristics like size and demographic diversity. More recent research has examined team composition in terms of latent constructs like personality and cognitive ability (Kozlowski and Bell 2013). The attributes required for a team to function effectively, however, go beyond the core characteristics commonly evaluated in traditional selection contexts (see Hollenbeck and Jamieson 2015). Consideration of individual members, as separate Bindependent^ entities, provides only a limited understanding of the dependencies and systemic elements of team behavior (Carron et al. 2002; Gammage et al. 2001). To this end, it has been argued that team members’ social capital should be considered in conjunction with individual characteristics when making composition decisions (Hollenbeck and Jamieson 2015). A person’s social capital is the interpersonal relationships he/she has with others that enable successful functioning (Hollenbeck and Jamieson 2015; Kilduff and Tsai 2003). It is possible for social capital to develop among team members even before the team has formally assembled through the work individuals perform together in other capacities. In the context of team assembly, social capital information provides insight concerning interdependencies related to team coordination (Lusher et al. 2010). Being a relational measure, social
J Bus Psychol
capital is conceptually distinct from the individual-level constructs typically considered during team assembly (e.g., personality, cognitive ability), and it has the potential to enhance prediction of individual and team performance incrementally over those constructs (Contractor 2013; Hollenbeck and Jamieson 2015; Kilduff and Tsai 2003). Cummings and Kiesler (2008), for example, found that the performance of interdisciplinary research teams was better when team membered had experience working together in the past, with shared work experiences buffering the negative effects of distance and disciplinary differences among team members. Importantly, social capital information, such as shared work experiences, can be quantified and considered during team assembly using social network analysis (Hollenbeck and Jamieson 2015).
Social Networks Analysis Social network analysis (SNA) is a set of methodological tools that focus on the Brelationships among social entities, and on the patterns and implications of these relationships^ (p.3, Wasserman and Faust 1994). As such, SNA indexes why and how strongly elements of a network are related to each other (Kilduff and Tsai 2003; Wasserman and Faust 1994). Social networks information can be indexed for individual team members (the node level) as well as teams as a whole (the network level; Lusher et al. 2010). At the individual level, this study examines the extent to which centrality influences player performance. At the team level, the extent to which team performance is influenced by network density and centralization is examined.
Individual-Level Network Concepts Measures of centrality index the extent to which an individual is important to the network to which he/she belongs (Kenny et al. 2006). The measures are influenced by both the number and types of connections an individual has with others in the network, with larger values typically indicating a more central position within the network (Lusher et al. 2010). Multiple measures of centrality exist, each determining an individual’s network position using slightly different criteria. Differences among the centrality measures included in this study are discussed in detail next. Principally, however, these measures all index the extent to which temporary team members are centrally located in their respective networks using information concerning the quality of their pre-assembly shared work experiences. Centrality is generally expected to facilitate team members’ ability to coordinate their actions with others in their respective networks, thereby fostering better performance.
Betweenness Centrality Betweenness centrality refers to the frequency with which an individual serves as the most direct connection between two other individuals in a network. More formally, betweenness centrality is defined as the number of times an individual member lies on the geodesic connecting other pairs of network members (Kenny et al. 2006). Individuals who connect others in a network are important because they act as couriers, offering a connection between members who are themselves not directly connected (Bourbousson et al. 2015; Kilduff and Tsai 2003; Wei et al. 2011). Individuals high in betweenness centrality are typically able to coordinate with the members they connect, which subsequently has a positive influence on their performance (Abbasi et al. 2011; Kilduff and Tsai 2003; Mehra et al. 2006). In this study, compared to team members who are low in betweenness centrality, temporary team members high in betweenness centrality share more high-quality pre-assembly work experiences with more pairs of team members who themselves do not directly share pre-assembly work experiences. Sharing higher-quality work experiences with multiple pairs of members in the network should facilitate coordination between and among the members in ways that enhance individual-level performance. Hypothesis 1: Individuals’ betweenness centrality scores in pre-assembly shared work experience networks will positively predict individual-level performance after team assembly.
Closeness centrality Closeness centrality refers to the average distance between an individual and all other members in a network. Individuals having stronger relations with others in the network are higher in closeness centrality, whereas those with weaker relations are lower in closeness centrality (Freeman 1978; Kenny et al. 2006; Wei et al. 2011). In this study, temporary team members high in closeness centrality would share higher-quality pre-assembly work experiences with other members of their teams, whereas those who are low in closeness centrality would share lower-quality pre-assembly work experiences with other teammates. Sharing higher-quality pre-assembly work experiences with many others in one’s network should facilitate temporary team members’ ability to coordinate their actions throughout the team because of an understanding of other’s work tendencies thereby enhancing individual-level performance (e.g., Abbasi et al. 2011). Hypothesis 2: Individuals’ closeness centrality scores in pre-assembly shared work experience networks will positively predict individual-level performance after team assembly.
J Bus Psychol
Eigenvector Centrality Unlike other measures of centrality, eigenvector centrality does not assume that all connections within a network are equally important. Instead, eigenvector centrality indexes a member’s importance based on the extent to which that individual is connected to others in the network who are themselves well connected (Bonacich 2007; Kilduff and Tsai 2003). Individuals having stronger relations with others in the network who are well connected are higher in eigenvector centrality than individuals who have weaker relations with well-connected members (Ruhnau 2000). Through their connections with influential members of the network, individuals high in eigenvector centrality are able to coordinate their actions with important others in ways that enhance individual-level performance (Ferriani et al. 2009; Mehra et al. 2006). In this study, temporary team members high in eigenvector centrality would share higher-quality pre-assembly work experiences with members who are central to the team compared to low eigenvector centrality individuals who do not share high-quality pre-assembly work experiences with central teammates. Familiarity with the work patterns of team members who are themselves familiar with the work patterns of many other team members should facilitate coordination with these individuals in ways that enhance individual-level performance. Hypothesis 3: Individuals’ eigenvector centrality scores in pre-assembly shared work experience networks will positively predict individual-level performance after team assembly.
Degree Centrality Degree centrality indexes the number or strength of direct connections an individual has with other members who have direct connections back to the individual in a network. Individuals’ degree centrality increases as they share stronger connections directly with others and as these others share stronger connections with the individual (Kenny et al. 2006; Kilduff and Tsai 2003; Wasserman and Faust 1994). In this study, temporary team members who are high in degree centrality would directly share higher-quality work experiences with more team members, and these team members would reciprocate higher-quality pre-assembly work experiences with the individual compared to those who are low in degree centrality. Having greater familiarity with the work patterns of more members of the team should facilitate coordination and thereby enhance individual-level performance (e.g., Abbasi et al. 2011; Bourbousson et al. 2015). Hypothesis 4: Individuals’ degree centrality scores in pre-assembly shared work experience networks will
positively predict individual-level performance after team assembly.
Team-Level Network Concepts Connections among network members can also be indexed at the team-level providing information about the amount, quality, and/or distribution of social capital in a network (Hollenbeck and Jamieson 2015). The team-level measures included in this study address centralization and density. Centralization indices provide insight concerning the extent to which networks contain few influential members. Centralization tends to benefit team performance on simple tasks, whereas decentralization tends to benefit team performance on complex tasks (Barth et al. 2015). This is likely because highly centralized teams rely on a small subset of well-positioned members to coordinate actions and perform principle functions. When tasks are simple, the wellpositioned members are capable of performing much of the work on their own. However, when tasks are complex, greater coordinating efforts than these individuals are capable of providing are required (Bourbousson et al. 2015). Winning an Olympic Ice Hockey Tournament is a complex task that requires high levels of coordination among team members. Decentralized teams are, therefore, generally expected to perform better in the tournament than centralized teams. In this study, team-level centralization measures were calculated for each of the four individual-level centrality measures included in the study. In addition to measures of centralization, we also examine the relationship between network density and team performance. Whereas centralization measures index the extent to which a network contains few influential members, measures of network density index the extent to which all network members are connected to one another (Kenny et al. 2006). Temporary teams with high network density would have more pairs of members who are familiar with each other’s work patterns through high-quality pre-assembly shared work experiences; low network density teams would have fewer teammates who are familiar with each other’s work patterns because they share lower-quality pre-assembly work experiences. Because a denser team is expected to have more members who are familiar with each other’s work patterns through higher-quality pre-assembly shared work experiences, team coordination should be bolstered therein facilitating team performance (e.g. Barth et al. 2015; Mehra et al. 2006; Reagans et al. 2004). Hypothesis 5: The degree to which teams’ pre-assembly shared work experience networks are (a) decentralized (i.e., low centralization values across all four measures
J Bus Psychol
of centralization) and (b) dense positively relates to team performance.
Method Sample and Design This study utilized archival data from g = 12 men’s ice hockey teams competing in the XXII (i.e., 2014) Olympic Winter Games. The countries represented were as follows: Austria, Canada, Czech Republic, Finland, Latvia, Norway, Russia, Slovakia, Slovenia, Sweden, Switzerland, and The United States of America. With 25 players per team, the data set contained N = 300 individual-level entries. This N = 300 included goalies, and, as explained in greater detail below, although goalies were included in the computation of the social network variables, the nature of their performance data did not allow them to be included in the test of the study hypotheses. Data were acquired from official online records for the 2014 Men’s Olympic Ice Hockey Tournament as well as the professional (e.g., NHL, Kontinental Hockey League) and amateur (e.g., National Collegiate Athletic Association) ice hockey leagues in which the players competed prior to the start of the Olympic Games. Social Network Coding2 For each team, a single social network was estimated by quantifying the quality of shared work experiences among the players within each country’s team prior to the 2014 Olympics. A symmetric, square matrix of continuous link values was used as inputs for the SNA. To determine link values between pairs of teammates, a coding system was developed based on quality of shared playing experiences. Specifically, a link value for a pair of players could range from 0 (i.e., no shared experiences among the pair) to 63 (i.e., highquality shared experiences among the pair) and was derived based on five categories of shared work experiences: (1) do two players play in the same professional/amateur league? (2) Do two players play in the same professional/amateur division? (3) Do two players play on the same team? (4) Do two players play the same position? And (5) did two players play together in the 2010 Olympic games. To expand the coding to 63 possible values, the connections represented by categories 1, 2, and 3 were coded for the four seasons prior to the 2014 Games (i.e., the 4 years between the 2010 and 2014 Olympics), with chronologically more recent experiences being considered higher quality. That is, professional teammates 2 The full coding scheme used to determine link values is available from the corresponding author.
from 2011 were considered to share higher-quality pre-assembly work experiences than professional teammates from 2010. To provide an example of the coding system, the Swedish pair of Henrik Zetterberg and Daniel Alfredsson had the maximum link value of 63: Both players played for the NHL’s Detroit Red Wings in 2014; therefore, both play in the same league and division. In addition, both play the same position (i.e., forwards), and both played for the Swedish Olympic team in 2010. According to this coding scheme, these two players shared the highest-quality pre-assembly work experiences possible and should therefore be most familiar with each other’s work patterns. On the same Swedish team, however, Jimmie Ericsson shared no pre-assembly work experiences (i.e., link values of 0) with 11 of the other players—this was likely due to the fact that Ericsson was the only Swedish player to not play in the NHL at any time between 2010 and 2014.
Measures Individual-Level Social Network Measures Betweenness Centrality Betweenness centrality indexes the frequency with which an individual serves as the most direct connection between two other individuals in a network. High betweenness centrality score players more strongly connect two otherwise weakly connected or unconnected teammates.
Closeness Centrality Closeness centrality is computed based on the number and length of the shortest paths between a team member and other members in the matrix. An individual with a high closeness centrality score has stronger connections to other team members that are more directly connected to many individuals.
Eigenvector Centrality Eigenvector centrality indexes the extent to which a person is more strongly connected to influential team members by allowing connections to highimportance individuals to contribute more to the score. High scores on eigenvector centrality mean that the team member shares strong connections to team members who are wellconnected.
Degree Centrality Degree centrality is an index of the number of individual team members that are directly connected to the individual. Higher scores on degree centrality suggest that a person has stronger connections to and from other individuals in the network.
J Bus Psychol
Individual-Level Performance Measures Player Points Player points represent the aggregate number of goals and assists a player earned throughout the tournament. A player receives a point for either an assist or a goal; therefore, player points represent an aggregate scoring/offensive performance criterion. Player Plus/Minus Player plus/minus is a measure that represents a player’s overall impact on his team’s tournament performance. Specifically, a player receives a plus if he is on the ice when his team scores a goal and a minus if he is on the ice when the other team scores a goal. In this way, plus/minus quantifies the differential between players’ offensive and defensive performance during the tournament.
Degree Centralization Degree centralization is a measure of the extent to which ties among members is concentrated among a few members. High scores suggest that few individuals in the network have the most familiarity with the work patterns of other teammates, and these other teammates are familiar with the individuals’ work patterns; other team members are less informed of the work patterns of the network members. Density Density of the network is the ratio of the number of actual connections to possible connections. Again, density was indexed within each country’s Olympic team, and higher density values indicate that more of the possible connections are realized in the team. Team-Level Performance Measures
Player Penalty in Minutes Penalties in minutes index, across the tournament, how many minutes a player spent serving a penalty for an infringement of the rules. Players with higher penalty in minutes (PIM) have violated the rules more often, causing their team to play with fewer players for a length of time, and generally represent counterproductive performance. Team-Level Social Network Measures Measures of network-level centralization assess variability in the individual-level centrality measures and index the extent to which few individuals are influential/important to the team (Grund 2012; Wei et al. 2011). Centralization was measured within each country’s Olympic team. Betweenness Centralization At the network level, this indexes the extent to which a network is reliant on a small number of individuals to connect more pairs of network members, resulting in less familiarity with work patterns among the team as a whole. High scores indicate networks reliant on a few individuals to connect other team members. Closeness Centralization Closeness centralization is a measure of the extent to which only a few members in a network are closely linked to others. Higher scores for this index suggest that few groups of team members are familiar with each other’s work patterns while other team members are less familiar with the work patterns of their teammates. Eigenvector Centralization Eigenvector centralization indexes the extent to which connections to influential members are shared across the network. In other words, high values of eigenvector centralization indicate a network wherein only a few influential members are aware of each other’s work patterns, and a large portion of the network is not familiar with the work tendencies of influential individuals.
Team Goals Team goals are the aggregate number of goals scored by all team members. Team Assists Team assists are the aggregate number of assists by all the team members. Team Plus/Minus Team plus/minus is the direct aggregate of the individual player plus/minus ratings. Teams with higher plus/minus values scored more goals than they allowed. Team Wins Team wins is the number of games a team won. Tournament Points Tournament points are awarded to teams for wins in regulation, wins in overtime, and losses in overtime. Teams are awarded three points for a win in regulation time, two points for an overtime win, one point for an overtime lose, and no points for a regulation loss. This variable differs from wins in that it rewards teams for performances that lead to overtime even if they lose, but penalize teams for performances that lead to wins in overtime rather than in regulation. Team Face-Offs Won and Percentage Team face-offs won and percentage represent the number and percentage of faceoffs won by a team. All hockey games start and restart with a face-off in which the official drops the hockey puck between opposing players, and these players compete the gain control of the puck to start a possession. Face-offs are a team-level competition; however, as all five players must know their set positions, be aware of the direction the puck will be played, and the subsequent movement of other players to assure acquiring the puck. We consider face-offs as a team process variable as a team must first control the face-off before they can score.
J Bus Psychol
Team Final Rank This variable indexes the teams’ final rankings in the tournament based on tournament and round robin performance.
Simple Rating System Rating Simple rating system (SRS) is an advanced metric of team performance that takes into account a team’s strength of schedule and goal differential for a victory. At its core, SRS assesses a team’s performance based on the average margin of victory for a game which is then weighted by the opponent’s average goal differential such that playing weaker opponents contributes less to the SRS rating. Higher scores indicate that teams won by a larger margin against better opposition (STATS Insights 2015). Control Variable Regular Season Plus/Minus We controlled for the effects of players’ previous performance accomplishments in our main analyses using the plus/minus ratings they earned during the regular season before the start of the Olympic Tournament. Plus/minus ratings are aggregate measures of performance that include goals and assists and represent elements of offensive and defensive play. Controlling for season plus/minus allows us to control for individual player offensive and defensive skills which might influence tournament play independent of pre-assembly shared work experiences. Analyses Social Networks Analyses Social network variables were obtained using the ORA software developed by CASOS at Carnegie Mellon University. The symmetric, square matrix of link values coded by the research team was used as inputs to compute the social network metrics. Importantly, goalies were included in the computation of social network ties for all players and the teams (i.e., an N = 300), but goalies were Table 1
excluded from subsequent analyses as they do not share the same performance metrics that the non-goalie teammates share (i.e., an N = 262). This approach allows us to account for association with goalies in the emergence of team processes, even though they could not be included in the main analyses. Main Analyses As players are nested within teams, the need for multilevel models to account for the nesting within teams was assessed for each individual-level outcome variable. This was done by comparing the fit of a model excluding a random intercept to a model including a random intercept (Field et al. 2012). If the random intercept model fit significantly better, a random intercept for team was included in subsequent analyses. If the random intercept model did not fit better than the fixed intercept model, ordinary least squares regression was used. Given the limited power at level 2 (i.e., only 12 teams), the relations between network centralization, network density, and team-level outcomes were investigated using correlations to explore if an effect worthy of further exploration exists.
Results Tables 1 and 2 provide the means, standard deviations, and intercorrelations among individual-level and team-level variables, respectively. Although the alternative measures of centrality were correlated, they were not completely redundant indicating that the measures assess unique aspects of network centrality. Relations Between Individual Performance and Team Outcomes Recall that we focus on whether or not individual team member centrality measures predict individual team
Means, standard deviations, and intercorrelations among team-member variables
1) Betweenness centrality 2) Closeness centrality 3) Eigenvector centrality 4) Degree centrality 5) Pre-Olympic plus/minus 6) Player tournament points 7) Player tournament plus/minus 8) Player tournament PIM N = 262 *p < .05 PIM penalties in minutes
Mean
SD
1
2
3
4
5
6
7
0.04 0.31 0.26 0.33 5.49 1.37 0.05 1.47
0.09 0.23 0.11 0.15 11.53 1.74 2.56 1.88
1 −0.05 −0.47* −0.45* −0.05 −0.07 −0.08 0.07
1 0.02 0.15* 0.15* 0.18* 0.27* 0.04
1 0.76* 0.02 0.06 −0.02 −0.03
1 0.10 0.04 0.10 0.03
1 −0.12 0.03 0.05
1 0.49* 0.10
1 0.04
J Bus Psychol Table 2
Means, standard deviations, and intercorrelations among team-level variables Mean
SD
Betweenness 0.29 0.16 Centralization Closeness 0.17 0.08 Centralization Network 0.83 0.13 Density Eigenvector 0.13 0.05 Centralization Degree 0.14 0.05 Centralization Team tournament 0.42 9.29 Plus/minus Team tournament 144.92 36.95 Face-offs won Team tournament 49.67 6.69 Face-offs % Team tournament 2.50 1.93 Wins Team tournament −0.27 1.96 SRS rating Team tournament 7.50 5.54 Points Team tournament 12.17 6.49 Goals Team tournament 19.00 11.91 Assists Team tournament 6.50 3.61 Rank
1
2
3
4
5
6
7
8
9
10
11
12
13
1 −0.22
1
−0.14 −0.12
1
0.23 −0.02 −0.86*
1
0.47 −0.04 −0.38
0.67*
1
−0.40 −0.25
0.62* −0.55
−0.63*
1
−0.43 −0.08
0.64* −0.52
−0.55
0.95*
1
−0.17 −0.17
0.68* −0.57
−0.59*
0.80*
0.85*
1
−0.30 −0.06
0.72* −0.61* −0.57
0.92*
0.97*
0.89*
1
−0.29 −0.26
0.64* −0.47
0.95*
0.93*
0.78*
0.90*
1
−0.27 −0.09
0.68* −0.59* −0.58*
0.93*
0.97*
0.88*
1.00*
0.91*
1
−0.39
0.01
0.35
−0.37
−0.60*
0.85*
0.82*
0.53
0.76*
0.77*
0.79*
1
−0.33
0.03
0.41
−0.34
−0.48
0.83*
0.81*
0.51
0.78*
0.80*
0.80*
0.97*
0.47
0.49
0.29
0.08 −0.61*
−0.43
1
−0.92* −0.97* −0.83* −0.97* −0.92* −0.97* −0.85* −0.85*
N = 12 *p < .05 Face-Offs % percentage of face-offs won, SRS simple rating system
member performance. Focusing on individual performance offers insights into team performance, and therefore offers evidence of the impact of individual-level centrality on team performance. This is evident in that aggregates of individual performance metrics relate to team performance metrics (Table 3). Team-level goals, assists, and plus/minus—direct aggregates of individual player values—correlate significantly with all team-level outcome variables. In short, results support the exploration of the effects of centrality on individual performance as a primary factor influencing team performance. Relations Between Individual Centralities and Individual Performance We first assessed the need for multilevel models to account for dependency in the data. This was done by assessing whether or not a significant amount of variation in the dependent variable exists between the level 2 units (i.e., teams). We compared the fit of a model with a fixed intercept to a model with a random intercept for player points, player plus/minus, and player PIM. When the
latter model fits better than the former, a significant amount of variance between teams in the outcome existed, and a random intercept was modeled. When the former model fits better than the latter, a fixed intercept was modeled (Field et al. 2012). Results suggest that random effect models fit the data better for player points, χ2 (1) = 28.01, p < .001, intraclass correlation (ICC) = .17, and player plus/minus, χ2 (1) = 138.02, p < .001, ICC = .49; a fixed effect model fits the data better for PIM, χ2 (1) < 1.00, n.s., ICC < .10. Therefore, we ran random coefficient regression models for points and plus/minus treating team as a random effect, but ordinary least squares regression for PIM (Field et al. 2012). Model results are presented in Table 4 and suggest that, after controlling for individual player skill (i.e., season plus/minus), closeness centrality is the only measure of centrality that is significantly related to points (Xu’s [2003] R2 = .04) and plus/minus (R2 = .06). No centrality measure, however, was significantly related to PIM (R2 = .01). In short, result partially supports hypothesis 2, but not hypotheses 1, 3, or 4—closeness centrality positively predicts team member points and plus/minus, but not PIM.
J Bus Psychol Table 3 Correlations between aggregate individual-level performance and team outcomes
Team performance
Individual performance Aggregate goals Aggregate assists Aggregate plus/minus
Tournament wins
Tournament points
Tournament f inal rank
Tournament SRS rating
.76*
.79*
.87*
.77*
.78* .92*
.80* .93*
.91* .91*
.80* .95*
N = 12 *p < .05 SRS simple rating system
Relations Between Team-Level Social Network and Team Outcomes Given the aforementioned power issue, we consider these team-level analyses proofs-of-concept of the need for future research on these relations. We expected network level centralization measures to be negatively related to team performance because measures of network-level centralization index variability in the individual-level centrality measures, and this variability has been shown to be negatively related to performance (e.g., Grund 2012). Network density, however, was expected to have a positive relation with team performance. The correlations among team-level network measures and measures of team performance are presented in Table 5. Results largely support our expectations concerning the directionality of these relations. Network centralization measures were negatively related to all team performance, but density was positively related to team performance. Although not all significant, the observed magnitudes of the relations provide some insight concerning the nature of the relations that exist between network structures and team performance. In light of
Table 4
the statistical power, these results generally support for hypotheses 5 and support the need for future research on these relations. Given the low magnitude of the relations between closeness centralization and the outcomes, we do not consider this centralization measure related to team performance.
Discussion Not only is team-based work increasingly becoming the norm in organizations (Hollenbeck et al. 2012), but the use of temporary teams to solve dynamic problems is increasing as well (Contractor 2013). Along with the changing nature of teams, changes must be made to the ways in which team members are assembled. Composition decisions should include consideration of both traditional individual characteristics (e.g., knowledge, skills, and abilities) as well as more relational determinants of team effectiveness (e.g., social capital; Hollenbeck and Jamieson 2015). Whereas ongoing teams have the ability to develop effective team processes through iterative performance episodes (Marks et al. 2001), the ability of temporary teams to develop in this way is meaningfully limited by the
Prediction of individual team-member performance from social network centrality measures when controlling for player plus/minus
Team-member performance metric
Olympic points Olympic plus/minus Olympic penalties in Minutesa
Team-member predictors
Estimate (Standard error) Estimate (Standard error) Estimate (Standard error)
Intercept
Player plus/minus
Betweenness centrality
Eigenvector centrality
Closeness centrality
Degree centrality
0.83 (0.49) −1.03 (0.84) 1.13* (0.44)
−0.02* (0.01) −0.01 (0.01) 0.01 (0.01)
−0.65 (1.31) −0.48 (1.59) 2.03 (1.52)
3.37 (2.06) −2.37 (2.93) −1.26 (1.74)
1.81* (0.91) 3.76* (1.79) 0.17 (0.52)
−2.38 (1.82) 1.86 (2.73) 1.49 (1.22)
N = 262 *p < .05 a
This performance metric was analyzed using ordinary least squares regression results
J Bus Psychol Table 5
Correlations among network level social network variables and team outcomes Network variable
Outcomes Tournament plus/minus
Betweenness centralization −0.40
Closeness centralization −0.25
Eigenvector centralization −0.55
Degree centralization −0.63*
Network density 0.62*
Tournament face-offs Tournament face-offs %
−0.43 −0.17
−0.08 −0.17
−0.52 −0.57
−0.55 −0.59*
0.64* 0.68*
Tournament wins Tournament points
−0.30 −0.27
−0.06 −0.09
−0.61* −0.59*
−0.57 −0.58*
0.72* 0.68*
Tournament team rank Tournament SRS rating
−0.22 −0.29
−0.04 −0.26
−0.46 −0.47
−0.56 −0.43
0.64* 0.64*
N = 12 *p < .05 Face-offs face offs won, Face-offs % percentage of face offs won, SRS simple rating system rating
duration of their life cycle. Understanding if and how temporary team performance may be facilitated through assembly (i.e., input) decisions is, therefore, of particular importance. The purpose of this study was to investigate if preassembly shared work experience among potential team members, as indexed using SNA, predicts performance after team assembly. If so, this information may be valuable to consider when making team composition decisions. Based on the tenets of the IMOI model of team development, pre-assembly shared work experiences were posited to facilitate coordination among team members thereby bolstering individual- and team-level performance. We focus on SNA metrics as they index the amount and strength of shared experience among team members prior to team assembly. We proposed that these shared experiences may help facilitate coordination among team members, thereby improving individual and team performance. To test this, we analyzed men’s Olympic hockey tournament data from the 2014 Winter Olympics. Olympic hockey teams are considered temporary teams in that they are assembled with players from various professional and amateur teams, and the team members first interact as a team about 2 days before their task (i.e., the Olympic tournament) started. Results of these analyses suggest that closeness centrality—sharing high-quality pre-assembly work experiences throughout the team—positively predicted player offensive (i.e., points) and overall (i.e., plus/minus) performance, controlling for individual player skill. We reason that this is likely due to the fact that these individuals can coordinate with other teammates faster because they are more familiar with other teammates work patterns as a result of sharing higher-quality preassembly work experiences. Team-level performance will also improve given that, particularly in this context, individual-level performance is directly related to team performance (see Table 3).
In addition, we provide initial evidence that team-level variants of the individual-level centrality measures (i.e., network centralization and density measures) were related to teamlevel performance. As teams became more centralized around single members, performance decreased. This is likely attributable to the coordination impediments experienced by teams with many members who do not have a familiarity with each other’s work patterns. As team density increased, however, team performance increased—this is likely due to the fact that denser teams have more members sharing higher-quality preassembly work experiences, and can therefore coordinate more effectively. As a whole, our results support the idea that work experiences shared among temporary team members prior to team assembly, as indexed using SNA, provide valuable insights into the social capital of team members (Hollenbeck and Jamieson 2015) and relate to individual and team performance. Not all centrality measures were significantly related to performance. This is likely due to the nature of the criteria investigated in this study. Specifically, the outcomes we used in this study were object performance variables that are (a) multiply determined and (b) distal outcomes of many proximal behaviors and performance episodes. That all centrality measures did not relate to these outcomes is therefore not surprising. We hypothesize, however, that these centrality measures are likely related to other, more proximal outcomes (e.g., passing efficiency), cognitive processes (e.g., team mental models), and attitudinal criteria (e.g., team satisfaction). Although not available here, these criteria should be the focus of future research. Expanding on Coordination We have argued throughout that familiarity with work patterns, obtained through pre-assembly shared work experiences, facilitates coordination (used generally) among
J Bus Psychol
temporary team members. We now expand on some of the emergent states that teams need to develop in order to function successfully, and whose development may be facilitated by pre-assembly shared work experiences. The following are key mediators from the IMOI framework (Ilgen et al. 2005) that influence team coordination. Coordination, as an emergent state, represents the process of orchestrating the sequence and timing of interdependent actions among team members, which develops as team members interact and learn to work together (Argote and McGrath 1993; Brannick et al. 1993; Marks et al. 2001). Coordination is considered vitally important because successful group outcomes are contingent on correct and timely contributions from all group members (Kozlowski and Bell 2013). Team mental models are organized knowledge structures shared among team members that contain mental representations of critical elements in the team’s task environment (e.g., Kozlowski and Bell 2013). These cognitive processes facilitate team performance by helping team members describe, explain, and predict events that occur within their environments (e.g., Mathieu et al. 2000; Mohammed et al. 2010). Team mental models exhibit greater sharedness as team members gain experience with one another (Kozlowski and Bell 2013; Mathieu et al. 2000). Cohesion is a multidimensional construct that is considered an affective team process (Mathieu et al. 2000; Kozlowski and Bell 2013). Cohesive teams share commitments to the task and/or other team members, and familiarity resulting from team members’ prior interactions facilitates cohesiveness (Goodman and Leyden 1991; Harrison et al. 2003). In addition to these, trust (Altschuller and Benbunan-Fich 2010), communication (Sullivan and Feltz 2003), and implicit coordination (Vashdi et al. 2013) are just a subset of the team processes that become entrenched via shared experiences among team members. Although the ability to develop these processes is truncated by the limited number of task cycles performed by temporary teams, sharing experiences prior to assembling as a temporary team should facilitate the development of these key states; future research can explore the relations among pre-assembly shared work experiences and these emergent constructs in different types of teams. Theoretical and Practical Implications The results of this study provide some theoretical and practical implications for our understanding of temporary teams. As noted above, the shared perceptions, normative expectations, and compatible knowledge that result from initial performance episodes (i.e., outcomes) become resources (i.e., inputs) that later serve to shape and refine the team processes employed in subsequent performance episodes (Kozlowski et al. 1999; Marks et al. 2001). Although the interactions addressed by models of team effectiveness tend to focus within the
boundaries of a team’s life cycle, the results of this research suggest that interactions occurring outside a team’s formal lifecycle, in the form of pre-assembly shared work experiences, also have a meaningful influence on performance. Given that the development of temporary teams is limited by the short-term nature of their existence, this finding has important implications for our ability to predict and explain temporary team dynamics. Practically, the results of this study suggest that preassembly shared work experiences are a form of social capital that is an important antecedent of temporary team performance. Temporary team members with higher-quality pre-assembly shared work experiences tended to have greater levels of individual performance than members with lower-quality experiences, and teams with shared work experiences distributed across their networks tended to have greater levels of team performance than teams with shared experiences clustered among subsets of their membership. Therefore, considering the shared work experiences among potential team members along with information about job-related knowledge, skills, and abilities should help composition decisions. Importantly, this study also offers a methodology organizations that can be used to operationalize this form of social capital and examine its relationship with performance criteria. In this way, the application of SNA as done in this study answers the call for evidence-based practices that support the formation, development, and management of temporary teams in today’s workforce (Hollenbeck and Jamieson 2015; Tannenbaum et al. 2012). One final practical implication we offer is a cautionary note. Although the measures of shared work experience used in this study were quantified and significantly related to job performance, assembly decisions should not be made on this information alone. Rather, these measures are intended to be considered in conjunction with information pertaining to the job-relevant knowledge, skills, and abilities of potential team members. As a reviewer of the study rightfully acknowledged, composing teams solely based on members’ shared work experience may restrict diversity in ways that thwart innovation, contribute to adverse impact, and engender negative fairness perceptions (i.e., nepotistic selection practices). In short, we suggest using pre-assembly shared work experiences in conjunction with other assembly criteria to create the necessary combination of shared work experience to facilitate coordination while also maintaining the necessary diversity of thought/ perspectives from new group members (Cummings and Kiesler 2008). Limitations and Future Directions Some limitations to this study should be noted and serve as avenues for future research. One limitation is the use of temporary sports teams as the sample. Although the Olympic
J Bus Psychol
teams are in fact temporary, their task for the tournament is unique. As the purpose of this study was to demonstrate that an effect worthy of further study exists, this sample serves this purpose (Zhu et al. 2015) and is in line with past studies (e.g., Grund 2012; Lusher et al. 2010). Nevertheless, it is important for future research to extend these results to other temporary teams in other organizational contexts to replicate and extend these results. Also, using temporary teams with different tasks may reveal other centrality measures that are important for temporary team member performance. Another limitation of this study is the low power at the team level. The nature of the Olympics only allows for 12 teams to qualify for the tournament—this precluded the possibility of studying team-level outcomes in a more systematic way. Although our analyses do suggest that social network variables at the individual and team levels influence team-level outcomes, future research, with the use of more teams, can more definitively assess these relations. Finally, as noted earlier, this study was limited in the criteria that it could assess; missing from the analyses were other criteria that are not included in box scores. We do not have measures of the emergent states (e.g., coordination and cohesion) thought to facilitate team performance. Likewise, other criteria of interest excluded from box scores include the following: satisfaction with the team, communication, decision-making quality, and willingness to serve again on a team with these individuals. Future research can collect measure of the emergent processes, particularly over time, to determine if pre-assembly shared experiences help these properties emerge faster, and expand the criterion domain. In doing so, researchers may also uncover different relations among the centrality measures and other criteria that were not observed here.
Conclusions Organizations are continuingly turning to work teams to complete tasks. Unfortunately, not all teams have the luxury of repeated task cycles to develop shared experiences or emergent states after they are composed; temporary teams have to assemble and complete a task quickly. This study provides initial evidence that pre-assembly shared work experiences are related to individual and team performance and may be useful to consider when assembling a temporary team. Although future research is needed, when choosing between two individuals to fill the final spot on a temporary team, our results show that, all else equal, choosing the individual with shared work experiences with other team members will result in better performance.
References Abbasi, A., Altmann, J., & Hossain, L. (2011). Identifying the effects of co-authorship networks on the performance of schools: a correlation and regression analysis of performance measures and social network analysis measures. Journal of Informetrics, 5, 594–607. Altschuller, S., & Benbunan-Fich, R. (2010). Trust, performance, and the communication process in ad hoc decision-making virtual teams. Journal of Computer-Mediated Communication, 16, 27–47. Argote, L., & McGrath, J. E. (1993). Group processes in organizations: Continuity and change. In C. L. Cooper & I. T. Robertson (Eds.), International review of industrial organizational psychology (8: 333–389). New York: Wiley. Barth, S., Schraagen, J. M., & Schmettow, M. (2015). Network measures for characterising team adaptation processes. Ergonomics, 58, 1287–1302. Bonacich, P. (2007). Some unique properties of eigenvector centrality. Social Networks, 29, 555–564. Bourbousson, J., R’Kiouak, M., & Eccles, D. W. (2015). The dynamics of team coordination: a social network analysis as a window to shared awareness. European Journal of Work and Organizational Psychology, 24, 742–760. Brannick, M. T., Roach, R. M., & Salas, E. (1993). Understanding team performance: a multimethod study. Human Performance, 6, 287–308. Contractor, N. (2013). Some assembly required: leveraging Web science to understand and enable team assembly. Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 371, 1–14. Carron, A. V., Colman, M. M., Wheeler, J., & Stevens, D. (2002). Cohesion and performance in sport: a meta-analysis. Journal of Sport & Exercise Psychology, 24, 168–188. Cummings, J. N. & Kiesler, S. (2008) Who collaborates successfully? Prior experience reduces collaboration barriers in distributed interdisciplinary research. Proceedings of the ACM conference on Computer-Supported Cooperative Work, 437–446. ACM. Devine, D. J., Clayton, L. D., Philips, J. L., Dunford, B. B., & Melner, S. B. (1999). Teams in organizations: prevalence, characteristics, and effectiveness. Small Group Research, 30, 678–711. Ferriani, S., Cattani, G., & Baden-Fuller, C. (2009). The relational antecedents of project-entrepreneurship: network centrality, team compsotiion, and project performance. Research Policy, 38, 1545– 1558. Field, A., Miles, J., & Field, Z. (2012). Discovering statistics using R. Washington, D.C.: Sage. Fransen, K., Van Puyenbroeck, S., Loughead, T. M., Vanbeselaere, N., De Cuyper, B., Broek, G. V., & Boen, F. (2015). Who takes the lead? Social network analysis as a pioneering tool to investigate shared leadership within sports teams. Social Networks, 43, 28–38. Freeman, L. C. (1978). Centrality in social networks conceptual clarification. Social Networks, 1, 215–239. Gammage, K. L., Carron, A. V., & Estabrooks, P. A. (2001). Team cohesion and individual productivity: the influence of the norm for productivity and the identifiability of individual effort. Small Group Research, 32, 3–18. Goodman, P. S., & Leyden, D. P. (1991). Familiarity and group productivity. Journal of Applied Psychology, 76, 578–586. Grund, T. U. (2012). Network structure and team performance: the case of English Premier League soccer teams. Social Networks, 34, 682– 690. Guimera, R., Uzzi, B., Spiro, J., & Amaral, L. A. N. (2005). Team assembly mechanisms determine collaboration network structure and team performance. Science, 308, 697–702. Harrison, D. A., Mohammed, S., McGrath, J. E., Florey, A. T., & Vanderstoep, S. W. (2003). Time matters in team performance:
J Bus Psychol effects of member familiarity, entrainment, and task discontinuity on speed and quality. Personnel Psychology, 56, 633–669. Hollenbeck, J. R., Beersma, B., & Schouten, M. E. (2012). Beyond team types and taxonomies: a dimensional scaling conceptualization for team description. Academy of Management Review, 37, 82–106. Hollenbeck, J. R., & Jamieson, B. B. (2015). Human capital, social capital, and social network analysis: implications for strategic human resource management. The Academy of Management Perspectives, 29, 370–385. Ilgen, D. R., Hollenbeck, J. R., Johnson, M., & Jundt, D. (2005). Teams in organizations: from input-process-output models to IMOI models. Annual Review of Psychology, 56, 517–543. Iverson, P. (2014). Winter Olympics hockey 2014: Line combinations take shape during men’s practice. Retrieved from http://www. sbnation.com/nhl/2014/2/10/5397838/winter-olympics-hockey2014-line-combinations-practice on August 30, 2015. Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. New York, NY: The Guilford Press. Kilduff, M., & Tsai, W. (2003). Social networks and organizations. Thousand Oaks, CA: Sage. Klein, K. J., Ziegert, J. C., Knight, A. P., & Xiao, Y. (2006). Dynamic delegation: shared, hierarchical, and deindividualized leadership in extreme action teams. Administrative Science Quarterly, 51, 590–621. Kozlowski, S. W., & Bell, B. S. (2013). Work groups and teams in organizations. In N. W. Schmitt & S. Highhouse (Eds.), Handbook of psychology, volume 12: industrial and organizational psychology. San Francisco: Jossey-Bass Kozlowski, S. W., Gully, S. M., Nason, E. R., & Smith, E. M. (1999). Developing adaptive teams: A theory of compilation and performance across levels and time. In E. Pulakos (Ed.), The changing nature of performance: Implications for staffing, motivation, and development (240–292). San Francisco: Jossey-Bass. Lawler, E. E., Mohrman, S. A., & Ledford, G. E. (1995). Creating high performance organizations: practices and results of employee involvement and Total Quality Management in Fortune 1000 companies. San Francisco, CA: Jossey-Bass. Lusher, D., Robins, G., & Kremer, P. (2010). The application of social network analysis to team sports. Measurement in Physical Education and Exercise Science, 14, 211–224. 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. Mathieu, J. E., Heffner, T. S., Goodwin, G. F., Salas, E., & CannonBowers, J. A. (2000). The influence of shared mental models on team process and performance. Journal of Applied Psychology, 85, 273–283.
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. Organizational Science, 17, 64–79. Mohammed, S., Ferzandi, L., & Hamilton, K. (2010). Metaphor no more: a 15-year review on the team mental model construct. Journal of Management, 36, 876–910. O’Leary, M. B., Mortensen, M., & Woolley, A. W. (2011). Multiple team membership: a theoretical model of its effects on productivity and learning for individuals and teams. Academy of Management Review, 36, 461–478. Reagans, R., Zuckerman, E., & McEvily, B. (2004). How to make the team: social networks vs. demography as criteria for designing effective teams. Administrative Science Quarterly, 49, 101–133. Ruhnau, B. (2000). Eigenvector-centrality—a node-centrality? Social Networks, 22, 357–365. STATS Insights (2015). A simple rating system for NHL teams. Retrieved from: http://www.stats.com/insights/nhl/simple-rating-system-nhlteams/on August 29, 2015. Sullivan, P., & Feltz, D. L. (2003). The preliminary development of the scale for effective communication in team sports (SECTS). Journal of Applied Social Psychology, 33, 1693–1715. Tannenbaum, S. I., Mathieu, J. E., Salas, E., & Cohen, D. (2012). Teams are changing: are research and practice evolving fast enough? Industrial and Organizational Psychology, 5, 2–24. U.S. Courts (2015). CJA Study Committee Begins Accepting Comments. Retrieved March 17, 2016, from http://www.uscourts. gov/news/2015/06/08/cja-study-committee-begins-acceptingcomments. Uzzi, B., Mukherjee, S., Stringer, M., & Jones, B. (2013). Atypical combinations and scientific impact. Science, 342, 468–472. Vashdi, D. R., Bamberger, P. A., & Erez, M. (2013). Can surgical teams ever learn? The role of coordination, complexity, and transitivity in action team learning. Academy of Management Journal, 56, 945–971. Wasserman, S., & Faust, K. (1994). Social network analysis: methods and applications (Vol. 8). New York, NY: Cambridge University Press. Wei, W., Pfeffer, J., Reminga, J., & Carley, K. M. (2011). Handling weighted, asymmetric, self-looped, and disconnected networks in ORA (no. CMU-ISR-11-113). Pittsburg, PA: CASOS Carnegie Mellon University. Xu, R. (2003). Measuring explained variation in linear mixed effects models. Statistics in Medicine, 22, 3527–3541. Zhu, X., Barnes-Farrell, J. L., & Dalal, D. K. (2015). Stop apologizing for your samples, start embracing them. Industrial and Organizational Psychology, 8, 228–232.