NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Committee on the Science of Team Science; Board on Behavioral, Cognitive, and Sensory Sciences; Division of Behavioral and Social Sciences and Education; National Research Council; Cooke NJ, Hilton ML, editors. Enhancing the Effectiveness of Team Science. Washington (DC): National Academies Press (US); 2015 Jul 15.

Cover of Enhancing the Effectiveness of Team Science

Enhancing the Effectiveness of Team Science.

Show details

4Team Composition and Assembly

Together, team composition and assembly make up one of the aspects of a team identified in Chapter 3 that can be manipulated to support team science. Team composition and assembly involve putting together the right set of individuals with relevant expertise to accomplish the team goals and tasks and to maximize team effectiveness.

The first section of the chapter discusses research on team and group composition that can be used to inform strategies for optimizing composition and enhancing effectiveness. Much of this research focuses on how individual characteristics of team or group members are related to performance. However, team composition is more complex than staffing individual positions because the members must collaborate well if the team is to be effective (Klimoski and Jones, 1995). This line of research provides robust evidence based on meta-analyses of empirical work on teams. The second section of the chapter reviews an emerging strand of research—team assembly—that takes a broader focus, examining how both individual characteristics and team processes (including the process of assembling the team or group) are related to team effectiveness. The third section of the chapter discusses tools and methods to facilitate composition and assembly of science teams and larger groups. The fourth section discusses the role of team composition and assembly in addressing the seven features that create challenges for team science outlined in Chapter 1. The chapter ends with conclusions and a recommendation.


Researchers have found that various individual characteristics are important considerations when composing teams or larger groups, both in science and in other contexts. Perhaps the most important individual characteristic to consider when composing a team science project is scientific, technical, or stakeholder expertise. As discussed in previous chapters, one of the key features that creates challenges for team science is high diversity of membership, as it may be necessary to include experts from multiple disciplines and professions to accomplish scientific or translational goals. For example, macrosystems ecology addresses ecological questions and environmental problems at the scale of regions to continents, linking these broad scales to local scales across space and time (Heffernan et al., 2014). Research in this field demands diverse expertise, including information scientists as well as ecologists (Heffernan et al., 2014).

One recent study provides evidence that high diversity of disciplines can improve scientific outcomes: Stvilia et al. (2010) studied 1,415 experiments that were conducted by teams at the national High Magnetic Field laboratory from 2005 to 2008. The authors' analysis of internal documents found that increased disciplinary diversity of the experimental teams was related to increased research productivity, as measured by publications.

Another study, however, illuminates the challenges as well as the benefits of highly diverse membership. In a longitudinal study of more than 500 National Science Foundation-funded research groups, Cummings et al. (2013) found that as the size of research groups increased, research productivity as measured by publications also increased. However, the marginal productivity of the larger groups declined as they became more heterogeneous, either by including experts from more disciplines or from more institutions (Cummings et al., 2013).

Other individual characteristics, including personality traits, may influence team science effectiveness. Feist (2011) has found that the characteristics of eminent, highly creative scientists include not only openness to experience and flexible thinking, but also dominance, arrogance, hostility, and introversion—personality traits that are not associated with being a good team player. Several studies have found that higher intelligence among team members, as measured by a team's mean level of general cognitive ability, is positively related to goal achievement, and the effect sizes are fairly large (e.g., ρ = .29 in Devine and Philips, 2001; ρ =.40 in Stewart, 2006). Higher conscientiousness, measured as a team's mean conscientiousness, is also positively related to team performance, although the relationship is stronger for performance and planning tasks than it is for creative and decision-making tasks that are similar to those carried out by science teams (Koslowski and Bell, 2003).

Extroverts who can easily monitor and respond appropriately to actions and attitudes of others (McCrae and Costa, 1999) may work more effectively in science teams or larger groups than introverts who are less attuned to teammates' actions and attitudes (Olson and Olson, 2014). Some evidence supports this theory, indicating that teams with higher mean levels of extroversion are more effective than teams with lower levels of this personality trait1 (Kozlowski and Bell, 2003). Woolley et al. (2010) recently identified a new individual construct related to extroversion that they refer to as “social sensitivity,” as well as a team-level construct called “collective intelligence.” In two studies of nearly 700 people working in small groups, the authors found evidence of the team-level collective intelligence factor and showed that it was related to group performance on a variety of tasks. The new factor was not strongly correlated with the mean level of intelligence within a group, but it was significantly correlated with the mean level of social sensitivity, the level of equality in taking turns during group discussions, and the proportion of females in the group. Individual social sensitivity was measured using a test requiring participants to “read” the mental states of others from looking at their eyes. In a follow-up study focusing on online groups, Engel et al. (2014) again found that a group's level of general collective intelligence was related to performance across a variety of tasks and that social sensitivity of group members was significantly related to collective intelligence. The result was surprising because social sensitivity was measured using the same test of one's ability to discern another's mental states by looking at their eyes and face, although the members of the online groups never saw each other at all. This suggests that the test measures a deeper aspect of an individual's ability to discern the mental states of others, beyond what the individual can “read” from another's eyes and facial expressions.

Based on data collected using unobtrusive badges to record team member interactions, Pentland (2012) also found that the level of equality in taking turns when speaking was related to team performance. He proposed that the most valuable team members are “charismatic connectors,” who circulate among all team members and spend equal amounts of time listening and speaking, while also seeking ideas outside the team; in a study of business leaders attending an executive education program, he found that the more charismatic connectors were included in a team, the more successful the team was. Finally, another related construct—the disposition to forge connections and share information among groups and individuals—was studied in the engineering division of an auto manufacturing firm. Individuals with this disposition were more frequently involved in innovation than other individuals (Obstfeld, 2005).

Although the finding that a high level of general cognitive ability enhances team effectiveness might suggest that science teams and groups should be composed entirely of individuals with this characteristic, a balance of characteristics may most benefit team effectiveness (Kozlowski and Ilgen, 2006). For example, a team composed entirely of extroverts might focus more on socializing than completing tasks while a team of highly conscientious individuals might be so task focused that the members do not collaborate well. Little research has tested this theory; however, one study of 41 teams in a research and development firm used an assessment to assign the team members into one of three cognitive styles: creativity, conformity to rules, and attention to detail (Miron-spektor, Erez, and Naveh, 2011). The authors found that including a balance of both creative and conformist members on a team enhanced its radical innovation (characterized as developing something completely new), whereas including a higher proportion of attentive-to-detail members hindered radical innovation (Miron-spektor et al., 2011). More recently, Swaab et al. (2014) found that basketball and soccer teams (which require highly interdependent actions by teammates) with the highest proportion of the most talented athletes performed worse than teams with more moderate proportions of the most talented athletes.

Other individual differences on dimensions such as gender, ethnicity, age, and specialized knowledge and abilities have been shown to exert both positive and negative influences on group processes and effectiveness. However, it is important to note that, in general, these other individual differences show smaller effects than do those discussed above (average level of cognitive ability, conscientiousness) (Bell et al., 2011). Prior studies that have examined the influence of individual differences and team diversity on team functioning generally have focused on one characteristic (or very few) at a time. However, each individual brings multiple characteristics to the team, making it difficult to prescribe individual factors for ideal team composition. By contrast, an emerging line of research on group faultlines (defined and discussed further below) takes into account the interplay among diverse individual characteristics and has made substantial progress in the past decade (Lau and Murningham, 1998; Chao and Moon, 2005; Thatcher and Patel, 2011; Carton and Cummings, 2012, 2013; Mathieu et al., 2014).

Here we highlight general findings for team composition based on team diversity, group faultlines, team subgroups, and changing team membership—factors that have clear implications for team science effectiveness.

Team Diversity

Diversity is at the heart of being a team, as teams have been defined as groups of individuals with different roles who work interdependently (Swezey and Salas, 1992). Indeed, interdisciplinary science teams and groups can be characterized this way (Fiore, 2008), making diversity the rule, not the exception. Research in this area has generally been conducted under the theoretical assumption that greater heterogeneity is associated with more diverse perspectives and, hence, better quality outcomes for diverse groups (Jackson, May, and Whitney, 1995; Mannix and Neale, 2005). However, support for this optimistic view has proven to be elusive and mixed at best, with findings supporting positive (Gladstein, 1984), negative (Wiersema and Bird, 1993), and no relationships (Campion, Medsker, and Higgs, 1993).

In their narrative review, Mannix and Neale (2005) concluded that demographic heterogeneity (based on easily recognizable surface features of an individual, such as gender, race, or age) tends to impede the ability of group members to collaborate effectively, whereas heterogeneity of knowledge and personality types—that is more task relevant—is more often associated with positive outcomes, but only when group processes are appropriately aligned with the task. A meta-analysis of the team diversity literature by Horwitz and Horwitz (2007) found no relationship between demographic diversity and the quality and quantity of team outcomes and small but statistically significant positive relationships between task-related diversity and the quality (ρ = .13) and quantity (ρ = .07) of team outcomes. A subsequent and larger meta-analysis by Joshi and Roh (2009) examined how contextual factors influenced the relationship between task-related diversity, demographic diversity, and team effectiveness. The authors found that contextual factors such as team interdependence and occupational setting influenced the direction and level of the relationships. For example, gender diversity had a significant negative effect on team performance in male-dominated occupational settings but a significant positive effect on team performance in gender-balanced occupational settings.

In light of the small and mixed effect sizes in previous studies of the relationship between diversity and team performance, Bell et al. (2011) conducted a new meta-analysis. The authors distinguished between the various conceptualizations of “diversity” used in previous studies, including diversity variety (multiple sources of expertise or knowledge that may contribute to team effectiveness), diversity separation (similarities or differences among team members that may lead to subgroups and negatively affect performance), and disparity (inequality within the team, such as the inclusion of one very senior member and many newcomers that may affect performance). They examined specific variables, rather than clusters of “job-related” (i.e., task-related) and “demographic” or “less job-related” variables, and considered how different performance measures and team types influenced the relationship. Significantly for team science, the performance measures included innovation or creativity, as well as general performance, and the team types included design teams charged with creating and designing new products.2 The authors found that only one type of task-related diversity—functional background diversity (i.e., the organizational division or profession of the team members)—had a small positive relationship with general team performance (ρ = .11). This relationship was larger when the performance measure was innovation or creativity (ρ = .18) and for design teams compared with teams in general (ρ = .16). In contrast, race variety diversity and gender diversity were negatively related to team performance (ρ = –.13 and –.09, respectively). Age diversity was unrelated to team performance.

In contrast to these meta-analytic findings, two recent studies focusing specifically on science found positive relationships between demographic and national diversity and the effectiveness of science teams or groups. First, Freeman and Huang (2014) studied the citation rates of more than 1.5 million scientific papers, finding that persons of similar ethnicity coauthor together more frequently than can be explained by chance given their proportions in the population of authors and that this homogeneity in authoring teams or groups is associated with weaker scientific contributions (as measured by citations). Papers produced by authors of diverse ethnicities are cited more frequently than those produced by authors of similar ethnicity. Freeman and Huang (2014a) proposed that ethnic diversity reflects idea diversity and, thus, better science is produced when collaborators bring different ideas and ways of thinking to the effort. They found the same positive effect on citations when researchers from geographically diverse universities collaborated. In a further analysis including 2.5 million papers, Freeman and Huang (2014b) again found that papers produced by authors of diverse ethnicities are cited more frequently than those produced by authors of similar ethnicity. Second, Smith et al. (2014) analyzed all papers published between 1996 and 2012 in eight disciplines, finding that those with more countries in their affiliations performed better in journal placement and citation performance than those whose authors came from fewer countries.

Other studies suggest that gender diversity can be beneficial for team science, showing that women tend to collaborate more than men do in academic science (Bozeman and Gaughan, 2011; Rijnsoever and Hessles, 2011). As noted above, Woolley et al. (2010) found that the proportion of women in a group was related to the group's collective intelligence, or ability to perform a variety of tasks. Bear and Woolley (2011) reported that the presence of women on teams is associated with improved collaborative processes. These processes have been shown to increase team effectiveness, as discussed in Chapter 3.

Overall, the research findings on the facilitative or inhibiting aspects of team diversity are mixed, although the meta-analytic evidence clarifies the picture somewhat. Further research is needed to explore how various forms of diversity are related to team performance. Following Bell et al. (2011), it will be important to carefully articulate the theoretical connection between the specific variable, the conceptualization of diversity, and team performance.

Group Faultlines

Faultlines are hypothetical divisions within a team based on team composition (e.g., two biologists and two physicists in a team form a possible faultline based on discipline). When compositional differences among members are made salient, such as when the team has to decide how to allocate resources or how to divide up the work, faultlines are said to be “activated” and subgroups are formed, raising potential for conflict (Bezrukova, 2013). For example, if a science team including two biologists and two physicists has enough funding to hire only one doctoral student, then faultlines may be activated as each disciplinary group wants to hire a student within its discipline.

Although the faultline concept is relatively new to the literature, it has stimulated a substantial amount of research, enabling an integrative and informative meta-analytic review by Thatcher and Patel (2011). Essentially, research in this area supports the differential effects of task-relevant and demographic diversity on team effectiveness: demographic diversity (factors such as gender, race, age, and tenure) is related to faultline strength, whereas task-related diversity in factors, such as educational level and experience with a team function are as well, but less so. Faultline strength contributes to weakened team relationships and task conflict that, in turn, inhibit team member satisfaction and performance. However, managers can address this problem by fostering identification with the larger team and developing shared goals (Bezrukova, 2009, 2013; see further discussion in Chapter 6).

Subgroups in Teams

Going beyond the faultline concept, Carton and Cummings (2012, 2013) have developed an alternative conceptualization of subgroup formation. Subgroups are subsets of team members who are uniquely interdependent in some way, such as those members who develop friendships with each other or who choose to collaborate. Prior empirical work has highlighted some of the benefits and costs of subgroup formation in teams. For example, in a study of 156 teams in pharmaceutical and medical products firms, Gibson and Vermeulen (2003) found that subgroup strength (i.e., the extent to which members in a subgroup overlapped on attributes, such as age, gender, ethnicity, function, and tenure) facilitated team learning behaviors. Teams with subgroups who had more in common were better able to come up with new ideas, communicate with each other, and document what they learned. However, when Polzer et al. (2006) examined the impact of subgroups within geographically dispersed teams, they found that teams including subgroups based on geography experienced higher conflict and lower trust. In particular, conflict was highest and trust was lowest when there were two equally sized subgroups each in a different country.

Other research findings have also illustrated the challenges of communicating across subgroups when faultlines are stronger, when subgroup distance is greater (e.g., subgroups based on very different ages; Bezrukova et al., 2009), and when subgroup size is imbalanced (e.g., six members in one subgroup and two members in another subgroup; O'Leary and Mortensen, 2010). A recent study by Carton and Cummings (2013) begins to reconcile some of the different results around the impact of subgroups in teams. They show that having more balanced subgroups can be better for team performance if the subgroups are knowledge-based (e.g., members with the same business unit and reporting channel in the organization) but worse for team performance if the subgroups are based on demographic characteristics, such as the same age and gender. On the one hand, in the case of knowledge-based subgroups, having an equal representation of knowledge sources on the team can be beneficial for integrating what is known (van Knippenberg, De Dreu, and Homan, 2004). On the other hand, having two subgroups composed of members with the same demographic characteristics can be costly when members get locked into in-group/outgroup differences (Tajfel and Turner, 1986).

Recent research provides insights on how to manage subgroups, whether based on knowledge or demographic characteristics. For example, Sonnenwald (2007) discussed some of the issues that can arise, such as mistrust, misunderstanding, and conflict, when ethnic minorities and minority-serving institutions participate in team science. He reported on strategies to address these issues, which include conducting extensive outreach to all participants early in the research planning, convening facilitated discussions with community authorities (e.g., religious leaders, tribal leaders), and using focus groups to elicit the community's concerns and priorities related to the research. DeChurch and Zaccaro (2013) identified leadership strategies to mitigate competition between teams within a larger multi-team system (similar to subgroups within a team) and foster shared identification with high-level goals (see Chapter 6 for further discussion). Structured discussions can be used to foster communication across subgroups based on discipline (O'Rourke and Crowley, 2013; see Chapter 5).

Changing Team Membership

Recent empirical work on teams, though not supported by meta-analytic findings, nonetheless suggests that changing team membership can enhance team performance. Gorman and Cooke (2011) found that in a three-person military command and control task, changing team members in a second session resulted in teams that were more adaptive in that they could better respond to novel events. In another study (Fouse et al., 2011), it was found that simply changing the location of team members doing a military planning task around a table resulted in a superior plan score compared to teams whose members stayed in the same location. Gorman and Cooke (2011) hypothesized that changes in team membership provide a chance for team members to experience more diversity in process behaviors, which is useful when the team faces challenges requiring different approaches. Similarly, changes in group membership associated with members leaving a group for another and then returning have been associated with increased creative ideas in essay writing (Gruenfeld, Martorana, and Fan, 2000). There seems to be some evidence for the positive influence of changing team membership from studies conducted outside of the laboratory. Kahn (1993) described the value of adjusting the composition of interdisciplinary science teams over the life cycle of a research network supported by the MacArthur Foundation.

Changing team composition through membership changes, often considered detrimental to team effectiveness, seems in some instances to have a positive effect and might be a useful intervention. In particular, faultlines that have formed may be disrupted by changing membership and collaboration dynamics that may be dysfunctional to team effectiveness may be pushed off their trajectory, resulting in positive process change.


Science teams and larger groups may be assembled by individual scientists, university research administrators (who sometimes function as matchmakers; see Murphy, 2013), funding agencies, or other groups or individuals. To guide the assembly process, individuals or organizations may rely on information about potential teammates based on prior relationships, consultations with experts in relevant areas, or more structured information sources. A new strand of research, known as team assembly, examines not only the composition of the team but also these processes.

Research on team assembly examines team composition at the team level (including the fit between team and task), the relational level within the team (e.g., individuals' prior relationships with each other), and the ecosystem surrounding the team (National Research Council, 2013). The goal is to understand how these multiple levels influence team performance. Here, we briefly discuss some of the findings from this new strand of research.

Guimera et al. (2005) studied science team formation, composition, and performance based on the analysis of teams in another domain—the universe of creative artist teams that made Broadway musicals from 1950 to 1995. Both Broadway and scientific teams aim to advance novel ideas and be creative (Uzzi et al., 2013). The authors found that Broadway teams were composed of two fundamental types of teammates: newcomers and experienced incumbents. They then defined the relationships within the team as newcomer-newcomer, newcomer-incumbent, incumbent-incumbent, and incumbent-repeated ties, finding that musical teams including a mix of all four types of relationships were most successful.

Guimera et al. (2005) applied this framework to science teams in four academic disciplines: astronomy, ecology, economics, and social psychology. Data on team composition were derived from authorship data from the five to seven top journals in each field, circa 1955–2004, as recorded in the Web of Science. They found that science team performance, as measured by the average citations accumulated by a paper (i.e., the journal impact factor), was positively associated with the probability of incumbents on the team, but only if the team had diversity, including newcomers and repeated ties among incumbents on the team. It is important to note that the model is predictive first and foremost of the population's performance level, not individual team-level performance. Consequently, any one team can be an exception in the short run, whereas the long-run systemic network within which teams in a field are embedded predicts average team performance in that field.

Contractor et al. (2014) conducted a study of student teams focusing on how they were assembled. First, students could either be assigned to teams or they could self-organize. Second, students could either use unstructured information about the other individuals to select teammates or use a team-builder tool populated with data provided by students with information about their attributes, social networks, and the sorts of people they would like on their team. The researchers found that teams that had used the team-builder tool were more homogeneous in age and cultural sensitivity, but more heterogeneous by sex. Not surprisingly, the self-organized teams (whether or not they used the builder) were more likely to contain members who had previously worked together than the teams that were assigned randomly. Analysis of surveys conducted 4 weeks after team formation showed that teams whose members all played a role in their organization (whether by using the builder or simply choosing their friends) communicated more and were more confident in their ability to work together effectively than teams with any members who were assigned.

Findings such as these raise questions about funding requirements that mandate inclusion of certain individuals, scientific disciplines, or institutions, within a team or larger group, rather than allowing teams or groups to self-organize. On the other hand, self-organizing teams or groups may be composed primarily of individuals with prior collaborative relationships, missing the benefits of newcomers with innovative ideas.


When the general focus of a research and/or translational problem has been established, team assembly can be guided using a “person-task fit” approach, or matching characteristics of individuals with characteristics of the research and/or translational task (National Research Council, 2013). Fields such as human factors (Wickens et al., 1997) and cognitive engineering (Lee and Kirlik, 2013) have contributed a number of methods for analyzing tasks that can guide team assembly. Task analysis involves the systematic decomposition of the behavior required of a task in order to understand the human performance requirements (Kirwan and Ainsworth, 1992). When composing a science team or group, it may be important to understand the tasks involved in operating scientific tools or equipment that will likely require specific technical competencies of one or more team members.

Assembly of science teams and groups may also benefit from cognitive engineering methods. Cognitive architectures, such as ACT-R, social network models, and agent-based modeling, have been used to understand and improve team effectiveness in highly cognitive tasks and can also be used to guide team assembly (Kozlowski and Ilgen, 2006). In addition, task analytic methods such as Cognitive Work Analysis (Vincente, 1999) have been used to design teams for first-of-a-kind work systems (Naikar et al., 2003). The fact that these complex systems are first of a kind makes the early analysis challenging, but in essence, the task model is developed alongside requirements for the team. This method takes advantages of constraints in the work environment that influence behavior. It involves detailed observations of work in context, accompanied by interviews at various levels of the organizational hierarchy to develop an understanding of the task or work in context. This approach has been applied to complex sociotechnical systems in which there are many people working with complex technology. Some science teams and groups work in similar environments, where they collaborate in designing and operating large and complex scientific equipment that is shared (e.g., the Large Hadron Collider). There are no data on the effectiveness of teams designed using this approach; however, it provides an analytic way of decomposing a task and work environment that may suggest team design needs that would otherwise be missed. These cognitive engineering approaches provide a systematic way of determining team requirements in terms of knowledge, skills, and abilities that can be used to guide team composition and assembly.

In other cases, however, the problems to be addressed using a team science approach are not clearly defined. As noted in the previous chapter, a team science project may begin when a group of scientists and/or stakeholders comes together to explore a problem or issue and the first phases may involve clarifying the focus and delineating research questions (Hall et al., 2012b; Huutoniemi and Tapio, 2014). In these cases, information on the larger ecosystem—the network of scientists and stakeholders with relevant interests and knowledge—may be helpful for team assembly.

Surveys have found that scientists, university administrators, and others involved in assembling science teams need a variety of information about potential collaborators, including not only publications, but also research interests, grant topics, and patents (Obeid et al., 2014). Such information is available from research networking systems that use data mining and social network approaches to create large, easily searchable databases, facilitating the search for scientific collaborators. These systems enable users to discover research expertise across multiple disciplines; identify potential collaborators, mentors, or expert reviewers; and assemble science teams based on publication history, grants, and/or biographical data (Obeid et al., 2014).

Many research networking tools are available, including Biomed Experts;3 Elsevier's SciVal© Experts and Pure Experts Portal;4 Harvard Catalyst Profiles;5 DIRECT: Distributed Interoperable Research Experts Collaboration Tool;6 and VIVO (Börner et al., 2012). VIVO, for example, is a free, open-source web application developed with support from the National Institutes of Health that facilitates search of researchers by publications, research, teaching, and professional affiliations across institutional boundaries (Börner et al., 2012). My Dream Team Assembler builds upon VIVO to incorporate social network analysis and modeling of the seeker to make recommendations of potential scientific collaborators (Contractor, 2013). An evaluation guide7 to research networking systems is available to assist institutions as they consider adopting these new tools.

Recent surveys suggest that research universities, especially academic medical centers, are increasingly adopting research networking systems (Murphy et al., 2012; Obeid et al., 2014), and many plan to share data on research expertise at their institutions using linked open data, allowing it to be widely accessed and analyzed. These publicly available data show promise for use in assessing cross-institution research collaborations in future team science research (Obeid et al., 2014). A recent study of implementation at the University of California at San Francisco (Kahlon et al., 2014) found that the research networking system was attracting an increasingly large pool of visitors whose behavior suggested they were using the tool to identify new collaborators or research topics. In response to an online survey, users identified a range of benefits to using the system to support research and clinical work. With the exception of this one study, however, there is little evidence to date that using the tools to guide team assembly results in teams or groups that are more effective than other teams or groups. The committee suggests that practitioners who choose to try one or more of these tools track the tools' usefulness and usability in assembling teams and collaborate with researchers to assess their impact on scientific outcomes.


How does the research on team composition and assembly speak to each of the seven features that create challenges for team science?

High diversity of membership (feature #1) is directly addressed by the research in team composition, faultlines, and subgroups summarized above. The finding that task-related diversity is associated with more effective teams is a promising finding for team science projects, which are composed primarily on the basis of task diversity.

Deep knowledge integration (feature #2) is actually a result of team composition, given that team science projects often require the integration of knowledge from multiple disciplines and stakeholders. Some of the tools discussed above such as the research networking systems, can potentially help mitigate the communication challenges resulting from this feature by making it possible to learn more about potential teammates in advance of team or group formation.

Large size (feature #3) is moderated by the heterogeneity of team or group members such that larger groups have been found to be more productive, but this advantage over smaller teams declines with increased heterogeneity in the disciplines and institutions represented (Cummings et al., 2013). Using methods such as cognitive work analysis to carefully analyze the tasks and requirements for team or group members of varying disciplines would help avoid unnecessary challenges of size and diversity.

The challenges emerging from goal misalignment with other teams (feature #4) are consistent with the concept of faultlines and subgroups that can be avoided by careful attention to team or group composition. However, science leaders or funding agencies sometimes place additional constraints on composition by requiring that a team or group include certain types of individuals, scientific disciplines, or institutions. Such constraints can inadvertently bring together subteams with multiple and sometimes conflicting goals. In these cases, it may be difficult to avoid the development of subgroups, and leadership and professional development interventions can be directed toward increasing the alignment of all subgroups with the high-level goals of the larger group.

Permeable team and group boundaries (feature #5) have been addressed only recently by research on dynamic team membership that acknowledges that modern teams tend to have fluid boundaries (Mathieu et al., 2014). Tannenbaum et al. (2012) observed as well that because organizations often need to rapidly reconfigure teams, individuals increasingly participate simultaneously in multiple teams. They noted that membership fluidity has been found to have both positive and negative effects on team performance, facilitating knowledge transfer on one hand, yet potentially reducing team members' bonds of affiliation on the other hand. To address these challenges, the authors suggested using team assembly tools, increasing role clarity, developing transportable team competencies, and focusing on team handoffs and transitions. At the same time, team processes may in fact be strengthened by changes in team membership as a result of increased team flexibility and adaptivity (Gorman and Cooke, 2011), increased unique ideas (Gruenfeld, Martorana, and Fan, 2000), and improved transfer of knowledge and alignment of member knowledge, skills, and abilities with task demands (Tannenbaum et al., 2012). Some research has found that acquaintance among team members and the trust it engenders facilitates effectiveness in cross-institutional teams or groups (Gulati, 1995; Shrum, Genuth, and Chompalov, 2007; Cummings and Kiesler, 2008). But, as discussed earlier, other studies suggest that membership changes and inclusion of members who are not prior acquaintances can improve the effectiveness of science teams or larger groups (Pelz and Andrews, 1976; Kahn and Prager, 1994; Guimera et al., 2005).

Geographic dispersion (feature #6) is known to create challenges for team success. Polzer et al. (2006) found that having subgroups based on geography was associated with higher conflict and lower trust. Geographically dispersed science team or groups are more likely to be successful if they are assembled so as to avoid faultlines and subgroups known to be problematic. However, if the scientific problem demands inclusion of members who may potentially divide along faultlines, interventions such as those described in Chapter 7 may be warranted.

Finally, high task interdependence (feature #7), a feature of many science teams and larger groups, can generate challenges when interdependence is required across subgroups or faultlines. Balancing teams at assembly to avoid such faultlines or counteracting them via leadership or other interventions will help facilitate interdependent work.


Most of the studies of the relationship between team composition and team effectiveness have yielded conflicting or weak effects. However, task-relevant heterogeneity does seem to be related to team effectiveness with important implications for science teams or groups including multiple disciplines. Further research on faultlines and the subgroups that can result from them corroborate the positive influence of task-related heterogeneity and the need to carefully manage demographic heterogeneity. At the same time, emerging research suggests that demographic heterogeneity can sometimes support scientific productivity.

The recent research on team assembly is beginning to offer insights into how the process of assembling the team or group and the prior relationships between the members affects the scientific and translational outcomes of team science. Research networking systems show promise for helping individual scientists, university research administrators, funders, and others identify potential team members. Further research on team assembly would be valuable at a time of rapid growth in team science.

The committee views this body of work as preliminary evidence that team composition and assembly matter and require careful management to facilitate effectiveness (Fiore, 2008). It is important to recognize that assembling and composing the team provides the raw building material for an effective team and therefore is a critical step, but it is only the first step toward an effective group or team (Hackman, 2012). Ployhart and Moliterno (2011) pointed out that human capital originates in the knowledge, skills, abilities, and other characteristics of individuals, but is transformed into a team resource through interpersonal processes such as those described in Chapter 3. Interventions in other aspects of a teams or groups, beyond composition and assembly, are important to support positive team processes and effectiveness, and we discuss these other aspects in the following chapters.

CONCLUSION. Research to date in non-science contexts has found that team composition influences team effectiveness, and this relationship depends on the complexity of the task, the degree of interdependence among team members, and how long the team is together. Task-relevant diversity is critical and has a positive influence on team effectiveness.

CONCLUSION. Task analytic methods developed in non-science contexts and research networking tools developed in science contexts allow practitioners to consider team composition systematically.

RECOMMENDATION 1: Team science leaders and others involved in assembling science teams and larger groups should consider making use of task analytic methods (e.g., task analysis, cognitive modeling, job analysis, cognitive work analysis) and tools that help identify the knowledge, skills, and attitudes required for effective performance of the project so that task-related diversity among team or group members can best match project needs. They should also consider applying tools such as research networking systems designed to facilitate assembly of science teams and partner with researchers to evaluate and refine these tools and task analytic methods.



When considering potential members for a team or larger group, it is important to recognize that individuals lacking in a beneficial characteristic (e.g., social or communication skills related to extroversion) may develop it through education or professional development, as discussed in Chapter 5.


As discussed in Chapter 1, new product development teams experience many of the same challenges as science teams.


Biomed Experts, see http://www​ [April 2015].


Elsevier's SciVal© Experts and Pure Experts Portal, see http://www​​/onlinetools/research-intelligence​/products-and-services​/pure [April 2015].


Harvard Catalyst Profiles, see https://connects​.catalyst​ [May 2015].


DIRECT: Distributed Interoperable Research Experts Collaboration Tool, see http:​// [April 2015].

Copyright 2015 by the National Academy of Sciences. All rights reserved.
Bookshelf ID: NBK310388


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (2.1M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...