A Review of the Team Learning Literature
Research on team learning spans a vast range of organizational settings, research methods, and dependent variables. Although this diversity in scope is reflective of the rich array of learning performed by real teams in organizations, it can also lead to confusing inconsistency in terminology and difficulty in accumulating findings. Indeed, the field of organizational learning is similarly diverse, with a long history of well-studied, but varied theoretical perspectives (for example, Argyris and Schön, 1978; Huber, 1991; Leavitt and March, 1988). Given the strength of each divergent theoretical contribution, we argue that a highly specified definition of team learning would sacrifice breadth for depth. Therefore, similar to definitions of organizational learning, we broadly define team learning as the processes and outcomes that involve positive change as the result of investments in developing shared knowledge or skill (for a discussion of different definitions of team learning, see Edmondson, Dillon, and Roloff, 2007).
To date, the research on team learning falls into three general streams of work: learning curves in operational settings (outcome improvement), psychological experiments on team member coordination (task mastery), and field research on learning processes in teams (group process) (Edmondson, Dillon, and Roloff, 2007). Each of these research traditions provides a different contribution to the understanding of organizational learning. The outcome improvement stream is primarily concerned with issues related to learning measurement. The task mastery stream research is focused on knowledge management. The group process stream examines how to learn. In addition, each of these streams takes a different methodological approach, ranging from small psychological lab groups to large-scale organizational improvement efforts. See Table 12.1 for a summary of the three streams of team learning. By drawing on these three streams of research, we first review team learning from multiple perspectives, then consider how each perspective advances our understanding of organizational learning.
Table 12.1 The three streams of team learning
These three streams also vary with respect to how team learning is operationalized and to disciplinary foundations. Whereas the outcome improvement and the task mastery streams conceptualize team learning as improved task performance, usually related to clearly defined tasks with measurable success, the group process stream is more concerned with team learning as adaptive behaviors with the potential to promote success when tasks, success, and context are less certain. In addition, the task mastery and group process streams draw upon the theoretical foundations of the social psychology of group dynamics as both focus on interpersonal coordination in teams. On the other hand, the learning curve studies largely hail from a more objective conceptualization of efficiency improvement without a focus on the interpersonal behavior related to improvements.
Despite the differences across streams, team learning research has matured from an exploratory, qualitative orientation in the early theory-building phase to a more explanatory, quantitative orientation aimed at fine-tuning or moderating established models of team learning (Edmondson and McManus, 2007). As such, team learning research has begun to investigate the mediators that connect antecedents of team learning to outcomes and moderators that specify the contexts and conditions of team learning. Some moderators tested to date within team learning research include task type (Van der Vegt and Bunderson, 2005) and industry context (Zellmer-Bruhn and Gibson, 2006). In addition, researchers have shown that learning is different for different types of teams, such as the difference between routine production teams versus innovation teams (Wong, 2004).
These three streams of team learning research, with their similarities and important differences, yield a number of insights for future research on how team learning serves as the foundation of organizational learning (Senge, 1990). The learning curve stream shows that practice and experience are important sources of learning for improving efficiency and productivity. Research on task mastery reveals that team learning involves knowledge coordination to effectively leverage team resources and skills in task execution. The group process work demonstrates that how teams learn is just as important as what teams learn, especially in the face of uncertainty.
Next, we review the related literature in each stream, and in the following section we build on this review by outlining how insights from these three streams form a foundation for a new research agenda aimed toward understanding the link between team learning and organizational learning. In particular, we suggest that considering learning across teams, such as through MTM, by leveraging what is known about learning within teams can provide an important link between the team and organizational levels of learning.
Measuring performance improvement: learning curves in teams
Learning curve research at the team level comes from a long and robust history of studying performance improvements with increased experience over time. In these studies, largely initiated in manufacturing settings, researchers documented the link between cumulative experience and improvements in operational performance such as increased productivity, reduced cost, or improved output. Since the early twentieth century, learning curve research has proliferated in the fields of technology management, economics, operations management, and competitive strategy. For the better part of its history, this stream of research has focused on the learning at the organizational level of analysis. However, the appeal of using objective, measurable performance data has attracted attention from team learning scholars, albeit later in the century. Whether the studies were focused on organizations or on teams, the core theme is that learning, in the form of experience over time, results in improved efficiency and performance.
Initial studies on learning curves in teams focused on improvement that is coincident with experience, or, ‘practice makes perfect.’ For example, in a study of thirty-six pizza franchises, Darr, Argote, and Epple (1995) found that unit cost decreased with experience. Similarly, two studies in the medical field found that procedure time decreased and number of procedures increased with experience (Edmondson, Winslow, Bohmer, and Pisano, 2003; Reagans, Argote, and Brooks, 2005). Studies in this stream consider the learning curve to be relevant at the start of a new project (product or process). As the project continues the learning curve reaches a peak level of performance and then plateaus as new learning slows down. Learning curve studies involve regression analysis models of longitudinal, quantitative outcome data from manufacturing or service organizations. The outcome variable of interest is usually related to a measure of efficiency such as cost reduction, increased productivity, or time. Recent research has studied the learning curve using data from multiple groups that implemented the same learning goal, and has stressed the importance of teamwork (such as communication and coordination) in generating improvements (Adler, 1990; Argote, Insko, Yovetich, and Romero, 1995). In general, learning curve research focused on finding and understanding differences in improvement rates across teams with the same task.
Studies of learning curves in teams have identified a number of factors associated with differences in improvement rates. One of the most widely noted factors is team stability (Argote, Beckman, and Epple, 1990; Argote, et al., 1995; Edmondson, et al., 2003). A recent study on self-managed production teams in a manufacturing setting showed that team turnover disrupts important interpersonal processes such as team learning behaviors and task flexibility, which are related to successful self management, in reducing the percentage of defects in production (Van der Vegt, Bunderson, and Kuipers, in press). In essence, teams with unstable membership improve less quickly than teams with stable membership, especially when task improvement involves the acquisition of tacit versus codified knowledge (Edmondson et al., 2003).
One reason team stability promotes learning from experience is simply the fact that team members who are familiar with each other have better coordination and teamwork (Reagans, Argote, and Brooks, 2005). In a study on surgical teams in hospitals, Reagans, Argote, and Brooks (2005) demonstrated that learning by doing increased team members’ ability to coordinate knowledge at the team level and to improve familiarity with the organizational processes linked to the task, especially when teams had membership stability. Similarly, a study of software teams showed that team familiarity—members’ prior experience working together—was a significant determinant of learning and performance in a setting where team membership was multiple and fluid (Huckman and Staats, 2009).
Task stability is another key factor associated with learning curves in teams. A recent field study of forty self-managed production teams in a high technology firm showed that when tasks are stable—are repetitive or continuous—team structure promotes learning (Bunderson and Boumgarden, 2011). The authors argue that when tasks are stable, team members that create clear roles and processes have an increased flow of information among members and reduced instances of conflict. Stability and structure create a safe and predictable environment where experience with tasks can lead to improvements. In an earlier and smaller sample study, Edmondson et al. (2003) showed that operating room teams that kept team members together over successive operations when learning a new cardiac surgery procedure reduced procedure time more quickly than teams with less stable membership.
In general, the results of learning curve studies show that rates of improvement are affected by the way learning is managed within teams. In using field-based research methods to build on the more traditional analytic approach, researchers have built on an established research paradigm on team learning from experience. In particular, the strength of learning curve research is its objective and measurable outcome variables with obvious practical implications. In addition, teams studied in this stream are often learning the same thing simultaneously, offering a view into comparisons across teams while holding the complexity of organizational contexts in which they work constant. Furthermore, this type of learning is likely the most easily translated across different teams and organizational units (Argote and Ingram, 2000; Wong, 2004), and thus more easily propagated in an MTM environment (O’Leary, Mortensen, and Woolley, 2011). On the other hand, the application of this research is limited to teams that perform repetitive, similar tasks and seek incremental improvements in efficiency and performance. Many organizational teams face challenges associated with innovation and radical improvement in highly uncertain contexts and, therefore, this work is limited with respect to insights for these types of teams.
Coordinating Team Knowledge: Task Mastery
A second area of research is focused on understanding how team members learn to master interdependent tasks through knowledge coordination. In this stream, team learning is the outcome of the communication and coordination that results from a shared knowledge base of team member skills, the task, team resources, and the task context. Success is measured in how well a team has learned, and mastered, their tasks.
Task mastery research emphasizes the importance of leveraging team member knowledge, skills, and abilities to increase the resources available to members during task execution. This work is primarily concerned with information processing in teams in the form of encoding, storing, retrieving, and communicating information (Wilson, Goodman, and Cronin, 2007). To say this another way, findings in this stream stress that teams are best able to perform interdependent tasks when members know what each other knows both collectively and individually. According to these researchers, team learning is not explicitly defined, per se, but rather learning is the outcome of improved performance of novel tasks.
Through tightly controlled laboratory studies of primarily student teams, researchers consider team learning from a cognitive perspective. Similar to how individuals develop knowledge, it is suggested that groups develop team-level cognitive systems that categorize and store collective knowledge. Teams of university students are assigned tasks such as simulating flight crews and assembling electronic devices. This type of research design allows for causal inferences about factors of team learning but is limited in terms of external validity. In addition, team members are often unfamiliar with each other before the task and disbanded afterward.
There are many terms for the team-level cognitive constructs studied in this stream including terms such as ‘transactive memory systems’ (Wegner, 1987), ‘shared mental models’ (Cannon-Bowers, Salas, Converse, and Castellan, 1993), and ‘social cognition’ (Larson and Christensen, 1993), among others. Despite different terms, these constructs share in common a characterization of team-level databases that encode, store, retrieve, and communicate knowledge in predicting task performance (for example, Hollingshead, 2001; Wegner, 1987).
These team level databases are thought to provide teams with a mechanism to create and organize a common understanding of team member knowledge, skills, and abilities in order to leverage team member strengths in task execution. The benefit of such a system is to enhance team coordination without the need for discussion. Creating a system of shared knowledge enhances team learning by enabling access to unique individual knowledge, promoting team member specialization, reducing redundancies in knowledge or skill, and creating informal rules about accountability. In general, this type of team learning is particularly important for teams that require diversity in expertise or knowledge to perform a task, such as in the case of product development teams. Developing these types of team-level databases requires effective communication among members, an area that merits further investigation given barriers to streamlined communication in modern, global organizations.
Studies on task mastery have focused on a number of factors associated with performance of novel tasks. First, researchers originally found that training team members on tasks together, rather than individually, was associated with improved task performance (Liang, Moreland, and Argote, 1995). Further research showed that it wasn’t the training, per se, that led to improved performance but rather its effects were mediated by the development of a transactive memory system (Moreland and Myaskovsky, 2000).
The development of transactive memory systems was associated with increased team member complexity, accuracy, and agreement in perceptions of each other’s expertise (Moreland et al., 1998), and a greater degree of tacit knowledge is shared during task execution (Gruenfeld, Mannix, Williams, and Neale, 1996). Quite notably, Stasser, Stewart, and Whittenbaum (1995) suggest that the formation of an effective transactive memory system is not due to simply mentioning differences in task expertise among members, but rather that explicitly recognizing these differences leads to informal schemas of accountability whereby ‘experts’ on the team are responsible for storing and retrieving specialized information.
Task mastery researchers have also suggested that teams develop different types of transactive memory systems related to, for example, task work and team processes (Mathieu, Heffner, Goodwin, Salas, and Cannon-Bowers, 2000). Of course, these various knowledge systems developed by teams are only useful in so far as they are accurate. Certain factors such as team size (Rentsch and Klimoski, 2001), previous task experience (Gino, Argote, Miron-Spektor, and Todorova, 2010), or organizational tenure (Smith-Jentsch, Campbell, Milanovich, and Reynolds, 2001) can inhibit or enhance transactive memory system development.
Furthermore, team processes such as communication are critical for achieving successful transactive memory system development (Lewis, Lange, and Gillis, 2005; Liang et al., 1995; Moreland and Myaskovsky, 2000; Rulke and Rau, 2000; Stasser et al., 1995). In particular, team climates that promote open and honest communication are more effective than climates that are interpersonally threatening. This is critical for diverse teams due to the tendency for team members to stereotype others in terms of the types of knowledge and experience they may bring to bear on the task (Hollingshead and Fraidin, 2003). Candid communication is also helpful for teams to surface unique individual knowledge because team members sometimes have a tendency to discuss shared knowledge as a method of creating social cohesion (Stasser, Stewart, and Whittenbaum, 1995; Whittenbaum, Hubbell, and Zuckerman, 1999). Finally, some have suggested that building team mental models is a highly political process in that team members are particularly concerned about how expertise labels affirm their identity or enhance their self-esteem (London, Polzer, and Omorgie, 2005; Walsh, Henderson, and Deighton, 1988).
In summary, the task mastery stream of research has contributed to an understanding of team learning by characterizing the cognitive dimensions of team level knowledge. That is, ‘knowing who knows what’ is an important feature of coordinating and leveraging team member strengths in learning together to perform novel tasks. Unlike the learning curve conceptualization of team learning, the task mastery form of learning is substantially more likely to be disrupted by MTM, as a larger number of team commitments tend to reduce the amount of time team members spend together which is critical for this type of learning to occur. Furthermore, the studies have predominantly employed laboratory methods, limiting an understanding about how organizational contexts and practices such as MTM affect team learning. Researchers have called for field studies to be performed to test these lab-based findings in real work contexts (Mohammed, Klimoski, and Rentsch, 2000). Such studies would help to expand our understanding about where and when developing coordinated ways of storing knowledge in teams is an essential aspect of the team learning process (for an example, see Lewis, 2004).
Learning How to Learn: Group Process
The third stream of research defines team learning as a group process instead of as a team outcome. Studies in this stream generally employ field research methods on work teams in real organizations. At its foundation, the group process stream originated from the input-process-output (I-P-O) model of team effectiveness, wherein team processes form the link between group inputs such as composition, structure, context, and group outputs such as innovation, quality, and performance (Hackman, 1987; see Ilgen, Hollenbeck, Johnson, and Jundt, 2005, for a review; McGrath, 1984). Whereas the prior two streams of team learning research have focused on inputs and outputs, the group process stream aims to understand the interpersonal processes in teams that constitute team learning. Typically, therefore, these researchers observe and measure the learning process in teams, rather than using performance outcomes as the measure of team learning.
Initially, group process researchers investigated the interpersonal processes in teams that fostered team learning behaviors. Evidence had begun to emerge that when team work is characterized by low quality interpersonal factors, teams are less likely to learn since members are not fully engaged and participating in the work due to fear of ridicule, embarrassment, or other forms of retribution for their actions (Brooks, 1994; Edmondson, 1996). In a study on fifty-three teams in a manufacturing firm, Edmondson (1999) identified psychological safety—a shared belief that it is safe to take interpersonal risks on the team—and demonstrated that it was a significant predictor of learning behavior in work teams. Group process work has shown that team leaders can play an important role in promoting team learning by fostering positive interpersonal climates in teams, involving members in decision making, clarifying team goals, and managing team boundaries with outsiders (Edmondson, 2003; Nembhard and Edmondson, 2006; Sarin and McDermott, 2003). When team leaders create a climate that is safe for risky interpersonal behavior, such as admitting mistakes, teams are more likely to learn. When teams learn, team performance improves as well. In addition to providing a safe climate for speaking up, team leaders can also neutralize power differences among team members (Edmondson, 2003; Nembhard and Edmondson, 2006; Van der Vegt, de Jong, Bunderson, and Molleman, in press).
Team learning scholars have also been interested in investigating the types of behaviors that promote learning first by broadly conceptualizing learning activities such as incremental learning, radical learning, vicarious learning, contextual learning behaviors, and local versus distal learning, among others (Bresman, 2006, 2010; Edmondson, 2002; Wong, 2004). A study of twenty-three process improvement teams in hospital intensive care units found that learn-what (learning behaviors related to acquiring knowledge or know-how) is conceptually distinct from learn-how (learning behaviors related to effective task strategies) in terms of the types of learning behaviors teams engage (Tucker, Nembhard, and Edmondson, 2007). Whereas learn-what behaviors were important for stocking a team’s knowledge base, only learn-how activities were related to team effectiveness. Subsequently, the nature and type of team learning behaviors have been considered in greater detail. Recently, Savelsbergh, van der Heijden, and Poell (2009) developed a team learning behavior instrument containing eight distinct dimensions of team learning (such as reflection, feedback seeking, experimenting) and twenty-eight items.
Indeed new tools to measure team learning reflect the many varied operationalizations of team learning in the group process stream. Edmondson defined team learning as ‘an ongoing process of reflection and action, characterized by asking questions, seeking feedback, experimenting, reflecting on results, and discussing errors or unexpected outcomes of action’ (1999: 353). Wilson, Goodman, and Cronin define team learning as ‘a change in the group’s repertoire of potential behavior’ (2007: 1043). Related constructs such as team reflexivity have also been used to define team learning. Team reflexivity is ‘the extent to which group members overtly reflect upon and communicate about the group’s objectives, strategies, and processes, and adapt them to current or anticipated circumstances’ (West, 2000: 296). Across definitions, team learning involves a change in the way teams operate as a function of noticing and correcting problems. Most notably, team learning is considered to be a verb in this stream. Team learning is also similar to new conceptualizations of adaptation—adjustment to change—in teams (for example, Burke, Stagl, Salas, Pierce, and Kendall, 2006; Woolley, 2009) and also to notions of team reflexivity, or reflection and communication about objectives, strategies, and processes (De Dreu, 2007; Schippers, Den Hartog, Koopman, and Wienk, 2003; Schippers, Homan, van Knippenberg, 2009).
Recent work in this stream has employed a number of dependent variables related to team learning such as effectiveness, performance, creativity (Hirst, Van Knippenberg, and Zhou, 2009), and innovation, among others. Team performance variables have received the most attention, and, whereas the learning curve and task mastery streams operationalize changes in performance as a measure of team learning, the group process stream conceptualizes team performance as an outcome of team learning. Indeed terms such as ‘performance’ and ‘effectiveness’ are ubiquitous in group process research, but operationalizations of these variables lack consistency. However, these differences are necessary for connecting behaviors to outcomes, and, as the team learning literature grows, perhaps models of team learning could become more contingent and precise relative to the types of performance and effectiveness variables of interest.
In addition, models of team learning have begun to detail the antecedents, moderators, and contextual variables associated with team learning processes. As these models are refined, researchers have outlined where, when, and for whom. For example, team learning works best when teams have an interdependent learning goal whereby members must rely on each other to complete teamwork (Bunderson and Sutcliffe, 2003; Ely and Thomas, 2001; De Dreu, 2007; Tjosvold, Yu, and Hui, 2004). In addition, certain types of team tasks require different degrees of learning behavior—routine production teams require less learning than interdisciplinary action teams (Edmondson, 2003). However, most of the task characteristics explored in this stream of research are reflective of the research setting and participants rather than an investigation of task features (for example, Edmondson, 1999, included four types of team tasks and Wong, 2004, measured task routineness).
More recently, researchers have begun to investigate the effects of team composition on team learning and performance. Researchers suggest that moderate levels of team diversity, in terms of demographic identity and functional experience, among others, is optimal for team learning whereas too much or too little diversity can undermine or overburden team learning (Gibson and Vermeulen, 2003; Lau and Murnighan, 2005; Sarin and McDermott, 2003; Van der Vegt and Bunderson, 2005). And team diversity is most beneficial when differences among members are balanced rather than divided into strong subgroups (Lau and Murnighan, 2005). Indeed team composition is a critical area for further investigation as organizations increasingly form diverse teams to leverage unique knowledge and experience held by members in organizational learning efforts.
Furthermore, there are a number of proposed moderators in models of team learning. As discussed, psychological safety—a team climate factor—is an important interpersonal moderator of team learning (Edmondson, 1999). Team identification has been proposed as another influential moderator, especially as a common identity with others on the team can serve to unite diverse individuals and orient them toward a common goal. Van der Vegt and Bunderson (2005) found, in a study of fifty-seven multidisciplinary teams in the oil and gas industry, that collective team identification moderated the relationship between expertise diversity on learning behavior and team performance. When collective identification was high, teams experienced greater learning across expertise faultlines than teams with low collective identification. When teams are diverse in terms of power and hierarchical position, performance feedback moderates the relationship between power asymmetry and team learning (Van der Vegt, de Jong, Bunderson, and Mollerman, in press). In a study of forty-six teams in an industrial setting, Van der Vegt and colleagues (in press) found that when teams receive group performance feedback, as opposed to individual performance feedback, power asymmetry is associated with higher levels of team learning, and thus greater team performance.
Finally, group process researchers have begun to outline which organizational contexts are ideal for team learning. A study of ninety teams from the pharmaceutical industry showed that when teams were interrupted during the course of their work flow, teams had an opportunity to reflect on their current activities (Zellmer-Bruhn, 2003). Further, Zellmer-Bruhn and Gibson (2006) determined that organizations that grant teams greater decision-making autonomy experienced more team learning activities than teams in organizations with strong prescribed practices. Thus, various aspects of organizational contexts can enhance (or inhibit) a team’s opportunities for learning.
This stream provides a detailed look into team learning through the diverse range of field research techniques ranging from rich qualitative data to meticulous quantitative measures. In addition, researchers in this stream acknowledge that individuals are nested within teams and teams are situated within larger organizational contexts through multi-level modeling techniques (for example, Edmondson, 1999; Zellmer-Bruhn and Gibson, 2006). However, the diversity of variables and methods is also limiting in that a comprehensive picture of team learning is difficult. It is difficult to compare results and to build knowledge across disparate terms and measures.
Some of the field-based findings in the team process tradition also suggest that multi-team membership (MTM) environments can be designed in such a way to optimize team learning and, ultimately, organizational learning. As teams come to recognize the potential benefits of the diverse experiences members bring from their work in other teams, they can develop routines for capturing and using that expertise. However, the time intensive nature of this form of learning may also lead it to be swept aside for the sake of efficiency. This underscores the importance of team leader behaviors and other moderating variables (such as team climate) to create the conditions that allow the sharing of new ideas that members may bring in from other teams to become a priority.