Citation
Rupert, J. (2010, April 14). Diversity faultlines and team learning. Kurt Lewin Institute Dissertation Series. Retrieved from
https://hdl.handle.net/1887/15223
Version: Not Applicable (or Unknown)
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the
University of Leiden
Downloaded from: https://hdl.handle.net/1887/15223
Note: To cite this publication please use the final published version (if applicable).
Being
Different, Yet Similar
How Faultline Strength and Distance Affect Team Learning and Performance
11This chapter is based on Rupert, J. & Jehn, K.A. (2009c) and is therefore written in the first-person plural.
W
ork group diversity has been shown to have an important impact on team processes, learning, and performance (Mannix & Neale, 2006; VanKnippenberg & Schippers, 2007; Williams & O’ Reilly, 1998). Especially informational diversity, which reflects differences in knowledge, perspectives and expertise, has been shown to enhance team functioning and performance (e..g. Argote, Gruenfeld & Naquin, 2000; Jehn, Northcraft, & Neale, 1999; Hinsz, Tindale & Vollrath, 1997). Therefore, organizations increasingly rely on teamwork, bringing members with different expertise together to stimulate team learning, innovation, and performance (Wilson, Goodman & Cronin, 2008).
While some teams manage to break routines and generate new solutions, other teams get stuck in previously adopted routines, unable to develop and change their coordination in fundamentally different ways (Argyris & Schon, 1978;
Edmondson, 1999; 2002; Gibson & Vermeulen, 2003). Similarly, as much diversity research has shown (e.g. Stewart, 2006; Van Knippenberg & Schippers, 2007; Webber & Donahue, 2001), effects of diversity are mixed and reaping the benefits of diversity remains a difficult challenge. As a response to these mixed results, Lau and Murnighan (1998) introduced the faultline perspective, focusing on the demographic alignment of members’ characteristics in the group (e.g., all accountants in a group are junior consultants, while all business analysts are senior consultants), creating subgroups in a team which may disrupt group functioning and performance.
In line with what the faultline perspective suggests, some studies have shown that subgroups can have negative effects, increasing levels of intra- group conflict (e.g., Lau & Murnighan, 2005, Li & Hambrick, 2005; Molleman, 2005), and decreasing levels of information sharing and elaboration (e.g.
Homan, Van Knippenberg, Van Kleef & De Dreu, 2007a; 2008; Philips, Mannix, Neale, & Gruenfeld, 2004; Sawyer, Houlette & Yeagley, 2003), cohesion (e.g., Li &
Hambrick, 2005, Sani, 2005), and performance (e.g., Bezrukova, Jehn, Zanutto, &
Thatcher, 2009; Homan et al., 2008; Rico, Molleman, Sánchez-Manzanares & Van der Vegt, 2007; Sawyer et al., 2005). In contrast, other studies have shown that subgroups can act as healthy divides, stimulating information elaboration, team learning, and performance under specific circumstances (e.g. Gibson &
Vermeulen, 2003; Homan, Van Knippenberg, Van Kleef, & De Dreu, 2007b;
Thatcher, Jehn & Zanutto, 2003). One of the criticisms of this past work on faultlines is that the theoretical focus has been largely drawn to the concept of faultline strength, reflecting the extent to which demographics align in a group, creating homogeneous subgroups. Yet, another aspect of faultlines, which is the distance between demographic subgroups (e.g., two members with 5 years
work experience are closer to another subset of team members with 8 years experience than to two members with 25 years work experience) has been largely ignored in past faultline research, while it may have an unique effect on team functioning (cf. Bezrukova et al., 2009). Therefore, in this study we extend past faultline research by examining the interplay between faultline strength and distance on team learning and performance. We focus on informational faultlines (e.g., educational level and work experience) in particular, since the combination of different levels of expertise is likely to influence team learning (cf. Van der Vegt, Bunderson & Oosterhof, 2006). As past research has shown, informational diversity can have potentially beneficial effects by stimulating team learning, innovation, and performance (e.g., Argote et al., 2000; Hinsz et al., 1997; Jehn et al., 1999).
A second contribution of this study is that we propose a theoretical framework (see Figure 1) in which we test two underlying mechanisms that can help to explain the effects of informational faultlines on team learning:
psychological safety and transactive memory. Psychological safety is defined as the ‘shared belief that the team is safe for interpersonal risk-taking’
(Edmondson, 1999: p. 354). This factor originates from human interactions characterized by interpersonal trust and mutual respect, and therefore is related to how interpersonal relations are being managed. The second mechanism that we study, transactive memory, is focused on how task-relevant expertise is being managed in the team. Transactive memory can be defined as the degree to which team members are aware of each other’s areas of expertise (Wegner, 1987), and results from the exchange of task relevant knowledge and
information team member posses. Research suggests that both mechanisms – interpersonal relations and task expertise – are likely to influence team learning (e.g., Edmondson, 1999; 2002; Lewis, Lange & Gillis, 2005) and to be influenced by a team’s group composition (e.g., Lau & Murnighan, 2005; Hollingshead &
Fraidin, 2003; Moreland, 1999). However, the link between group faultlines, interpersonal relations, task-expertise, and team learning has not yet been thoroughly examined.
Finally, we contribute to recent literature on team learning (Sessa &
London, 2007) that has suggested that different types of team learning can be distinguished based on the topic of team learning. According to this literature, two work-related types of team learning can be distinguished (Jehn & Rupert, 2007): task and process learning. In this paper, we empirically test a model linking informational faultlines to these types of team learning in health care teams in two large organizations, thereby advancing our understanding of
these different types and their relationship with other concepts. In addition, we explore the link between these types of team learning and team performance, and test whether task and process learning mediate the relationship between psychological safety, transactive memory, and team performance (see Figure 1).
Faultline Distance
In past diversity literature, the concept of “distance” often refers to people’s perceptions of differences. For instance, in research on multiform heterogeneity (Blau, 1977) distance refers to the idea that people make inferences based on demographic characteristics of others, which exacerbates differences between people, creating social distinctions and social barriers. In line with this, cross-cultural researchers refer to the concept of cultural distance which indicates the perceptions of differences between members of different countries (Leong & Ward, 2000), which is particularly relevant for research in the area of transnational and international joint venture teams (e.g., Earley &
Mosakowski, 2000; Hambrick et al., 2001). Distance is also implicit in the concepts of separation and disparity (Harrison & Klein, 2007), distinguishing different types of heterogeneity, with separation referring to horizontal differences in position or opinion among unit members (opposition being the most extreme case) and disparity referring to vertical differences between unit members that have to do with status, hierarchy, and pay (having a leader with
+ + +
+ +
Faultline Distance
Psychological Safety Transactive
Memory
Faultline
Strength Team
Performance Task Learning
Process Learning
Figure 1. Theoretical Model Linking Faultlines to Team Processes and Team Outcomes
followers being the most extreme case). In a similar vein, the relational demography literature refers to the concept of demographic distance, which is the degree of isolation of an individual from a group (e.g., Tsui, Egan, & O’Reilly, 1992; Wagner et al., 1984). This past diversity research indicates that when studying diversity, one should not only consider the amount of differences between people but also how distant people are based on these differences.
Despite this prior research on distance, early work on the effects of faultlines on group functioning has largely focused on the concept of faultline strength, neglecting to consider the distance between subgroups (cf. Bezrukova et al., 2009; for exceptions see Bezrukova, Thatcher, & Jehn, 2007; Molleman, 2005). The concept of faultline strength reflects the extent to which
demographic characteristics align with each other, creating two relatively homogeneous subgroups (Lau & Murnighan, 1998; 2005; Thatcher et al., 2003).
For instance, a group with a strong faultline would be a group consisting of two junior nurses who just graduated from school and two senior behavioral therapists with 25 years work experience (e.g., a faultline based on job function and work experience). However, this group would get the same faultline strength score as a group consisting of two nurses having 15 years of work experience and two senior behavioral therapists with 25 years of work
experience, while these groups are likely to have very different group dynamics (Bezrukova et al., 2009). This is captured by the concept of faultline distance, which refers to the distance between two subgroups based on specific demographic attributes. In this study, we test a mediated moderation model, examining the effect of faultline strength on psychological safety, transactive memory, and team learning, moderated by faultline distance (see Figure 1).
Additionally, we examine the impact of psychological safety, transactive memory, and team learning on performance.
Theoretical Model and Hypotheses
The Effects of Faultline Strength and Distance on Team Learning
Much past faultline research has focused on the negative aspects of subgroup formation, creating negative conflict, lowering behavioral integration and cohesion, and decreasing performance (e.g., Bezrukova et al., 2009; Homan et al., 2007a; 2008; Li & Hambrick, 2005; Molleman, 2005; Philips et al., 2004).
However, other research indicates that faultlines can be positive for information sharing, team learning, decision making, and performance (e.g., Gibson &
Vermeulen, 2003; Homan et al., 2007b; Larson, Sargis, & Bauman, 2004; Thatcher et al., 2003). In this study, we try to reconcile these inconclusive findings by considering the moderating role of faultline distance. We argue that the effects of faultline strength may vary according to the level of distance between the subgroups. In particular, we draw on research on cohort formation and shared and unshared information in groups to argue that the smaller the distance is between subgroups – hence being different yet similar, the more likely it is that subgroups will foster team learning.
In teams with faultlines based on educational level and work experience, team members with different work experiences may also hold different educational levels. This can lead to cohort formation (Asch, 1952;
Walsh, 1988), creating subgroups of team members with similar educational levels and work experiences that often share similar views. These “supportive cohorts” (Gibson & Vermeulen, 2003) stimulate team members to put forth their knowledge and views on issues in the team as a whole, which can stimulate team learning under certain circumstances. For instance, the classic Asch studies (1952, 1956) demonstrate that a team member’s opinion is more likely to be expressed and listened to when at least one other team member supports this view (Azzi, 1993; Wittenbaum & Stasser, 1996). In line with this, research on shared and unshared information shows that shared information is mentioned more often and team members are more likely to consider divergent opinions or countervailing information when it is held by multiple people (e.g. Azzi, 1993;
Brodbeck, Kerschreiter, Mojzisch, Frey & Schulz-Hardt, 2002; Christenson et al., 2000; Larson et al., 2004, Sargis & Larson, 2002; Stasser, Taylor, & Hanna, 1989;
Stasser & Titus, 1985, 1987). Team members who share a viewpoint or
information are apt to be perceived as more reliable sources of information (e.g.
Kameda, Ohtsubo & Takezawa, 1997; Wittenbaum, Hubbel, & Zuckerman, 1999).
Thus, subgroup formation along informational lines may support team members in expressing knowledge and divergent viewpoints and to consider, explore, and reflect upon different ideas put forth by other team members (e.g.
Gibson & Vermeulen, 2003; Larson et al., 2004). However, we propose that this will only be the case when faultline distance is small.
The potential beneficial effects of cohort and coalition formation can become disruptive and antagonistic when subgroups are too far apart (Jetten, Spears, & Manstead, 1998). When faultline distance is high, this may stimulate subgroup members to remain “psychologically located” within their subgroup
consisting of member’s with similar views (Nesdale & Mak, 2003), resulting in opposing fractions holding divergent viewpoints. Recent research (Bezrukova et al., 2009) indeed shows that faultline distance can exacerbate disruptive
processes in faultline groups, resulting in lower levels of performance. Similarly, in a study on top management teams, Wagner et al., (1984) found that higher demographic distances among vice-presidents further increased negative team dynamics.
When subgroups become too distant along informational lines this may result in team members speaking “different languages”, which interferes with knowledge and information sharing within the group (Weber & Camerer, 2003). This can impair task learning, since the availability of task-relevant knowledge will be reduced (Friedman & Podolny, 1992), and team members are less likely to consider divergent viewpoints from team members with a different informational background. Moreover, we expect that strong and distant informational faultlines will also impede learning about processes. Due to lowered levels of process reflection in the team as a whole, the team is more likely to make shallow assessments of the teams’ own performance and solutions may fail to be implemented (e.g. Ancona & Caldwell, 1992; Miller, Burke, & Glick, 1998; Sutfcliffe, 1994). In addition, the work is more likely to be divided according to salient divisions, rather than based on a thorough assessment of the match between members’ expertise and task requirements.
In sum, the potential positive effects of informational faultlines are more likely to flourish and stimulate task and process learning when distance is small. We propose that:
Hypothesis 1. Faultline distance moderates the relationship between faultline strength and task and process learning, that is, when distance is small, faultline strength is more likely to result in result in higher levels of task and process learning than when distance is large.
In sum, we discussed the moderating role of faultline distance and argued that when distance is low, subgroups can stimulate task and process learning. We now discuss two possible mediating mechanisms based on relational and task processes in groups, that can help explain the interplay between faultline strength and distance on task and process learning:
psychological safety and transactive memory.
Faultlines and Psychological Safety
The first mediating mechanism that may help explain the relationship between faultlines and team learning is psychological safety. Psychological safety focuses on how interpersonal relationships are managed (see Figure 1), and entails a shared belief that a team is safe for interpersonal risk-taking. It evolves from interpersonal relationships based on mutual respect and trust (Edmondson, 1999; 2002). In a team that is psychologically safe, team members feel confident that their team members will not embarrass or reject each other for bringing up mistakes or difficult issues.
A team’s psychological safety is likely to be influenced by a team’s composition. Although in faultline research faultlines have been argued that to negatively impact psychological safety, (e.g., Gibson & Vermeulen, 2003; Rupert
& Jehn, 2008), a recent study by Lau and Murnighan (2005) has shown that subgroups can positively influence psychological safety. In line with this, research on cohort and coalition formation has found that team members with similar backgrounds are more likely to share a common language and
understand each other’s viewpoint (e.g., Crott & Werner, 1994; Katz, 1982;
O’Reilly, Caldwell, & Barnett, 1989; Schein, 1985; Walsh, 1988; Zenger &
Lawrence, 1989). Thus, subgroups can make it safer for team members to express their viewpoints, because they feel supported by their fellow subgroup members.
However, and in line with what has been previously argued, these arguments are likely to hold particularly in the situation of small distance between subgroups. When members of different subgroups are too distant, they are less likely to feel safe in the team as a whole and are more likely to remain “psychologically located” within their subgroup of members with a similar background (Nesdale & Mak, 2003). As a result, there will be less
exchange of relevant knowledge, information, and ideas, and team learning will suffer. In contrast, when distance is small and subgroup members align based on informational characteristics, they are more likely to support each other based on common attitudes towards issues (Murnighan & Brass, 1991). This can make it “psychologically safer” for the individual to express knowledge and ideas in their subgroup with similar others (Asch, 1952; Crott & Werner, 1994;
Edmondson, 1999; Gibson & Vermeulen, 2003), enhancing team members’ self- efficacy (Bandura, 1997). As a result, team members will feel encouraged to also share their viewpoints and information in the team as a whole when they know their fellow subgroup members will support their view (e.g. Kramer, 1990;
Brewer, 1991). Thus, team members in teams with informational faultlines will also perceive higher levels of psychological safety in the team level as a whole.
Therefore, subgroups can make team member’s feel more psychologically safe, especially when between-subgroup distance is low.
We propose:
Hypothesis 2a. Faultline distance moderates the relationship between informational faultline strength and psychological safety, that is, when distance is small, faultline strength is more likely to result in higher levels of psychological safety than when distance is large.
Psychological Safety as a Mediator
When team members perceive the team as psychologically safe, as we propose is the case in faultline teams with a small distance between subgroups, the team is more likely to learn. In a team that is psychologically safe, members are more likely to speak up, share task-relevant knowledge and information, bring up tough problems and discuss mistakes, and value each other’s skills and differences, which can all foster team learning (Edmondson, 1999; 2002). For instance, admitting and discussing an error that occurred can be of incredible importance for making changes in the future and therefore for the team to learn (Homsma, Van Dyck, De Gilder, Koopman & Elfring, 2009; Van Dyck, Frese, Baer,
& Sonnentag, 2005). However, team members may be unwilling to bring this error up because they are concerned about being seen as incompetent. A psychologically safe climate alleviates this concern of other’s reactions to actions that are potentially embarrassing or threatening, which learning behaviors often are (Edmondson, 1999).
Research has indeed shown that a psychological safe climate is an important condition that facilitates learning behavior (e.g., Carmeli, 2007;
Edmondson, 1999; Nembhard & Edmondson, 2006; Tjosvold, Yu, & Hui, 2004).
When team members feel that the group is a safe place to express ideas and opinions and members will not be punished or rejected for making mistakes, team members will feel more at ease and will be more likely to present new ideas (Edmondson, 1996; 1999). In contrast, the willingness of employees to participate in problem-solving activities diminishes significantly when they perceive the team as hostile (e.g., Dutton, 1993; MacDuffie, 1997) and this may diminish learning behaviors (Argyris & Schön, 1978). David H. Smith, head of
knowledge development for Unilever, explained why there are so many
examples of repeated mistakes in organizations: ‘‘Fear is the No. 1 reason: fear of being embarrassed, chewed out or worse. Many people and companies are so busy trying to hide bones (from the boss, from stock analysts, from customers and competitors) that they tuck away the learning along with the evidence’’
(Stewart, 1997: 159). We therefore expect that psychological safety is likely to foster both task and process learning. When the team is perceived as
psychologically safe, team members are more likely to share and reflect upon task-relevant information and generate better task relevant solutions, resulting in higher levels of task learning. Similarly, psychological safety is also likely to foster process learning, as team members are more likely to ask each other critical questions, give feedback on each other’s performance, and to discuss errors and unexpected actions when the team is perceived as psychologically safe (Edmondson, 1999). As a result, they are more likely to adjust processes when they are ineffective and monitor their own performance as a team, resulting in higher levels of process learning. We propose the following mediated moderation hypothesis:
Hypothesis 2b. Psychological safety mediates the relationship between the interaction of informational faultline strength and distance and task and process learning.
In line with this, research has shown that when organizations provide a non-threatening environment they are more successful in terms of firm goal achievement and return on assets (Baer & Frese, 2003). Also, team members will exert more effort to accomplish team goals (Brown & Leigh, 1996). For instance, a study by Nembhard and Edmondson (2006) shows that psychologically safe teams engage more in quality improvement efforts. Therefore, psychologically safe teams are more likely to take appropriate actions to accomplish their work by team learning, thereby fostering performance (Edmondson, 1999). We therefore propose the following additional mediating hypothesis (see Figure 1):
Hypothesis 2c. Task and process learning mediate the relationship between psychological safety and performance.
Thus, we argue that how interpersonal relations are being managed, which is reflected in a team’s psychological safety, will help explain the relationship between faultline strength and task and process learning, under
varying circumstances of faultline distance. We now consider an alternative mediating mechanism focused on task-relevant expertise that may also mediate the relationship between faultlines and task and process learning: a team’s transactive memory.
Faultlines and Transactive Memory
Another mechanism that we propose may mediate the relationship between faultlines and team learning focuses on how task-relevant expertise is managed in the team: a teams’ transactive memory (see Figure1). A transactive memory can be defined as a system in which the knowledge and expertise that team members possess is combined with an awareness of who knows what (Liang, Moreland, & Argote, 1995; Wegner, 1987). For instance, the health care teams in this study provide care for patients who are mentally and/or physically handicapped. These teams often consist of nurses with particular knowledge about the specific disorders, combined with trainers, educators, doctors and/or behavioral therapists, who have knowledge about how to deal with these patients and their handicaps. When a patient has physical problems, for instance, it is important that team members know who to ask for help with medication, and when there are emotional or behavioral problems which team members have the ability to help the patient. Through the transactive memory system, members will know who to consult for what and how to take advantage of the informational sources that are available in the team in order to come to an optimal solution (Liang et al., 1995).
A prerequisite of transactive memory is that there is a certain composition of expertise present, which represent the “intellectual capital”
or “knowledge assets” available to the team (Marquardt, 1996) and will foster
“distributed” expertise sharing (Mohammed & Dumville, 2001; Rau, 2005).
Therefore, a group’s composition based on informational differences will influence the development of a transactive memory. In line with what was argued before, we expect that faultlines can have beneficial effects for the development of a team’s transactive memory, particularly when faultline distance is small. In strong but close subgroups, team members with similar informational backgrounds are more likely to share knowledge and expertise (e.g., Azzi, 1993; Wittenbaum & Stasser, 1996). This shared knowledge is more likely to be brought up and considered in the team as a whole and to be seen
as credible since it comes from more than one person (e.g., Kameda et al., 1997; Wittenbaum et al., 1999). Credibility is an important dimension of transactive memory as the extent to which team members will rely on other member’s expertise depends on their beliefs about the reliability of this knowledge (Liang et al., 1995; Lewis, 2003). In sum, information is more likely to be stored, encoded, and retrieved in a teams’ transactive memory when this information is 1) shared and 2) when it is reliable. The supportive cohorts in teams with strong but close faultlines will stimulate the sharing of
knowledge within and across subgroups and increase the reliability of knowledge within the group. We therefore propose that these teams are more likely to develop a transactive memory. Thus,
Hypothesis 3a. Faultline distance moderates the relationship between informational faultline strength and transactive memory, that is, when distance is small, faultline strength is more likely to result in higher levels of transactive memory than when distance is large.
Transactive Memory as a Mediator
Transactive memory in turn is likely to influence team learning for different reasons. As the task knowledge and work expertise of group members is especially important to the development of the transactive memory system (Moreland, 1999), we expect transactive memory to be positively related to task learning. Recent research indicates that transactive memory affects members’
ability to apply prior learning and to develop a shared understanding of the task (Lewis et al., 2005). Moreover, when a group has a transactive memory system, team members are more likely to rely on accurate information, since a
transactive memory can provide team members with more and better
information than each individual team member can remember alone (Liang et al., 1995; Littlepage, Robison, & Reddington, 1997). By knowing each other’s expertise areas well, new knowledge can be directed more easily to the relevant experts, allowing group members to acquire and store knowledge more efficiently as a whole than as individuals (cf. Ren, Carley & Argote, 2006), which promotes learning about the task.
Transactive memory is also likely to influence process learning. When a transactive memory exists, coordination is likely to improve since team
members can better anticipate each other’s behavior and plan their work more accordingly, for instance by dividing the work according to members’ expertise (Moreland & Levine, 1992; Murnighan & Conlon, 1991). In line with this, research has indeed shown that when a transactive memory exists team members correctly identify expertise in their team and delegate tasks more easily according to team members’ expertise and experiences, thereby enhancing coordination processes (e.g. Hollingshead, 1998; Liang et al., 1995; Littlepage et al, 1997; Moreland, 1999; Moreland & Myaskovsky, 2000; Stasser, Stewart, &
Wittenbaum, 1995). To summarize, teams with strong but close faultlines are more likely to develop a transactive memory system, which will foster task and process learning (see Figure 1). Thus, we propose:
Hypothesis 3b. Transactive memory mediates the relationship between the interaction of informational faultline strength and distance and task and process learning.
Additionally, past research has demonstrated a positive relationship between transactive memory and group effectiveness in groups composed of co-workers (e.g. Hollingshead 2000), and work groups in laboratory settings (e.g. Liang et al., 1995; Hollingshead, 1998; Moreland, Argote, & Krishnan, 1998, Moreland & Myaskovsky, 2000). Several studies showed that the group’s performance and decision making quality depended on the extent to which the group recognizes and utilizes the knowledge of its member, by correctly identifying their experts and delegating tasks according to members’ expertise (e.g., Ellis, 2006; Faraj & Sproull, 2000; Hollenbeck, et al., 1995; Hollingshead 1998; Lewis, 2000; Littlepage et al., 1997; Stasser et al., 1995). We propose that the relationship between transactive memory and performance is mediated by higher levels of task and process learning.
Hypothesis 3c. Task and process learning mediate the relationship between transactive memory and performance.
Method
Sample and Procedure
We conducted a field study in two health care organizations for mentally and/or physically disabled people. We collected survey data on 67 teams (response rate 84%), with 503 respondents. Additionally, we collected supervisors ratings of team performance (97% response rate), and archival data to complete the demographic information of the teams. For calculating faultlines, we used the 100% decision rule for work-group diversity (Allen, Stanley, Williams, & Ross, 2007), which prescribes that only teams with full demographic information will be included. This allowed us to include 54 teams in our analyses with faultline scores. We had almost full response on our supervisor ratings of team performance and the average response rate to the team survey was 82% (teams with less than 50% were discarded from the analyses). The average age of the team members was 36, 80% were female and 96% were Dutch. The average group size was 10 members, who worked together for 5 years, on average. Participants represented different levels of education (secondary school [17%], lower vocational education [63,6 %], and higher vocational education and university [19,5 %]).
To increase the response rate for each team, agreements were made with the two organizations that we would visit team meetings to ask team members to fill in the survey during the team meeting. The HR managers of the two organizations asked the team leaders to announce our visit two weeks in advance. During our visit to the team meeting we explained the purpose and importance of the research and guaranteed team member’s anonymity. Team members who were not present at the meeting received a survey in their mailbox and were urged to send this survey back. Supervisors were asked to rate team performance.
Measures
Faultlines. For measuring faultline strength, we used the faultline algorithm developed by Thatcher et al., (2003) and used in faultline research (e.g., Lau & Murnighan 2005, Molleman 2005). The algorithm calculates the
percent of total variation in overall group characteristics explained by the strongest group split by calculating the proportion of the between-group sum of squares compared to the total sum of squares (faultline strength can vary between zero and one, with larger values indicating greater strength). In this study, we calculated overall faultline scores based on educational level and prior work experience, since we were interested in informational faultlines (Bezrukova et al., 2009; Jehn et al., 1999; Tsui et al., 1992). The values of faultline strength in our dataset ranged from .37 to .96, which is an appropriate range to determine faultline effects (e.g., Bezrukova et al., 2007; 2009).
To calculate faultline distance we used the measure developed by Bezrukova et al. (2009). This formula reflecting how far apart the subgroups are from each other based on demographic characteristics (e.g., a group of two 20- years old females and two 50-years old males get a larger distance score than a group of two 35-year old females and two 50-years old males). The score is calculated as the distance between the faultline variable centroids for the subgroups (the Euclidean distance between the two sets of averages):
1 2
2
1
j j
p g
j
D
¦
x x , where the centroid (vector of means of each variable)for subgroup 1 = (x11,x21,x31, , x p1)and the centroid for group 2 =
12 22 32 2
(x ,x ,x , , xp ). Faultline distance can take on values between zero and ∞, with larger values indicating larger distances between demographic subgroups (Bezrukova et al., 2009). Values of faultline distance in our sample ranged from 1.56 to 3.65 (M = 2.42, SD = .37).
Team Member and Supervisory Surveys
We used existing measurement scales to measure our dependent and mediating variables in an employee survey, with 1-7 Likert scale items. The factor analysis of all survey items is displayed in Table 1. All constructs load on different factors, showing that the constructs are distinct. We assessed team performance in a supervisor survey.
Psychological safety. We measured psychological safety using Edmondson's (1999) scale consisting of seven items. Sample items were: “It is safe to take a risk on this team” and “Members of this team are able to bring up problems and tough issues”. As the factor analysis in Table 1 indicates, the
reversed coded items of this construct loaded on a different factor than the straight items, but both factors were still distinct from the other survey variables. For theoretical reasons we decided to combine all psychological safety items into one scale, after rescaling the reversed coded items. The scale had a Cronbach’s alpha reliability coefficient of .72, an intraclass correlation ICC[1] of .22, an ICC[2] of .71, and rwg = 83. A significant F-test indicated that aggregation was appropriate (Klein & Kozlowski, 2000).
Transactive memory. We used an adaptation and extension of the transactive memory subscale of specialization by Lewis (2003), which measures the extent to which team members have specialized knowledge. However, the definition of transactive memory does not only capture the extent to which there is specialized knowledge, but also the extent to which members know who knows what (Liang et al., 1995; Wegner, 1987). As the specialization subdimension has only one item measuring the extent to which team members know who knows what (“I know which team members have expertise in specific areas”), we extended this scale by adding a few questions measuring this meta- knowledge dimensions of transactive memory (Liang et al., 1995). The final scale consisted of five 1-7 likert scale items with sample items “If I need to know about a particular topic, I know who to go to”, and “As a team we know each other’s expertise well” (see Table 1 for the factor analysis with all items). The internal consistency of the scale was good (α = .89), with ICC[1] of .09, an ICC[2]
of .47, and rwg = .73. A significant F-test indicated that aggregation was appropriate (Klein & Kozlowski, 2000).
Task Learning. We used an adaptation of the task learning scale developed by Rupert and Jehn (2008; 2009b). The scale consisted of five items, which measured the extent to which team members felt that they shared and reflected upon information, knowledge, and ideas about the task at hand. We also asked to what extent this learning about the task improved team
performance. Sample items were “By reflecting upon knowledge about the task we improve our performance” and “We improve our performance on the task by sharing task-related knowledge with each other”. The internal consistency of the scale was good (α = .88), and the aggregation measures were acceptable with ICC[1] = .05, and ICC[2] = .28, and an rwg = .80 (LeBreton & Senter, 2008).
Process Learning. We adapted the process learning scale developed by Rupert and Jehn (2008; 2009a), measuring the extent to which team members thought their team learned about work processes and routines and adjusted
1 2 3 4 5 Task Learning
TL1 As a team we improve our performance by learning about the
task. .02 -.81 .06 -.02 .02
TL2 By working together as a team, we learn more about the content
of the task. -.05 -.80 -.02 .03 .01
TL3 We improve our performance on the task by sharing task-related
knowledge with each other. .06 -.80 -.03 -.02 .05
TL4 Through interaction with each other we increase our potential to
perform the task. .08 -.78 .05 -.02 .06
TL5 As a team, we learn about the task at hand. -.01 -.77 .02 -.07 -.03 Proces Learning
PL1 In our team, we learn about different ways to do our work. -.01 .00 .02 -.88 -.07 PL2 We regularly reflect on our work procedures to see how we can
improve them. .02 .03 .05 -.85 -.06
PL3 As a team, we develop work routines that help us to improve the
performance of our work. .02 .01 -.05 -.80 .07
PL4 We improve our performance by reflecting upon the way we do
our work .01 -.27 -.06 -.63 .09
Psychological safety
PS1 People on this team sometimes reject others for being
different.(R) -.01 .04 .86 . 00 -.05
PS2 It is safe to take a risk on this team. .02 -.02 .00 . 02 .74 PS3 It is difficult to ask other members of this team for help. (R ) .08 -.15 .73 . 01 -.09 PS4 Members of this team are able to bring up problems and tough
issues. .13 -.05 -.11 .02 .70
PS5 If you make a mistake on this team, it is often held against you. (R) -.03 . 06 .67 -.03 .28 PS6 No one on this team would deliberately act in a way that
undermines my efforts. -.15 -.07 .10 -.01 .57
PS7 Working with members of this team, my unique skills and talents
are valued and utilized. .20 .04 .09 -.12 .57
Transactive memory
TM1 My team members know what I am good at. .91 .06 -.04 .01 . 03 TM2 As a team we know who is good in what. .88 -.02 .05 -.00 .03 TM3 As a team we know each other's expertise well. .85 -.02 .03 -.06 .09 TM4 If I want to know something about a particular topic I know
exactly who to go to. .68 -.16 -.04 .04 -.10
TM5 We work efficiently because we know what everyone is good at. .67 .07 .16 -.21 .11 Note. Extraction method: Principal Component Analysis. Rotation method: Oblimin with Kaiser Normalization
Table 1. Factor Loadings of Constructs
these processes when they were no longer effective (e.g., “We regularly reflect on our work procedures to see how we can improve them”, and “In our team, we learn about different ways to do our work”). The scale consisted of four items. The scale had a mean Cronbach’s alpha reliability coefficient of .84, with an ICC[1] of .09, an ICC[2] of .46, and rwg = 75. A significant F-test indicated that aggregation was appropriate (Klein & Kozlowski, 2000).
Team Performance. We asked supervisors to rate team performance on a two item scale (Jehn et al., 1999): “How well do you think this team performs?” and “How effective is this team?” The internal consistency of the scale was good (α = .88).
Controls. In our analyses we controlled for organization, group size, team tenure, and team heterogeneity. Since organization, group size, and team tenure were not associated with the other variables in our study, we only controlled for team heterogeneity in our final regression analyses.
Results
The means, standard deviations, and correlations among the variables in our study are displayed in Table 2. To test our hypotheses, we conducted hierarchical linear group-level regression analyses (see Table 3) and centralized our variables, as recommended by Aiken and West (1991).
Table 2. Means, Standard Deviations, and Correlations among the Variables
M SD N 1 2 3 4 5 6 7 8
1. Team Heterogeneity¹ .62 .18 52 - 2. Faultline Strength .63 .14 52 -.27* - 3. Faultline Distance 2.42 .37 52 .35** .21 - 4. Psychological Safety 5.00 .60 67 -.01 .15 -.18 - 5. Transactive Memory 5.18 .54 67 .14 .08 -.06 .47*** - 6. Task Learning 5.63 .40 67 .05 .25 -.12 . 48*** .51*** - 7. Process Learning 4.84 .53 67 -.14 .17 -.31* .52*** .61*** .61*** - 8. Team Performance 5.58 .85 65 .20 .02 -.08 .23† .25* .18 .24* - Note. ¹ Based on educational level and prior work experience (years)
*** p <.001,** p <.01, * p <.05, † p <.10
In hypothesis 1 we proposed that faultline distance moderates the relationship between faultline strength and task learning that is, when distance is small, faultline strength is more likely to result in higher levels of task learning than when distance is large. The results of our regression analysis supported this hypothesis, with a significant interaction between faultline strength and distance on task learning (β = -.30, p < .05; see Table 3). In line with what was hypothesized, faultline strength was positively associated with task learning, but only when distance was small (see Figure 2). We found a similar pattern for process learning, further supporting hypothesis 1. Regression analysis showed a significant interaction (β = -.27, p < .05), indicating that faultline strength was positively associated with process learning when faultline distance was small (see Figure 3).
4,30 4,50 4,70 4,90 5,10 5,30
low med high
Faultline Strength
Process Learning
Faultline Distance high
low
Figure 3. The Moderating Role of Faultline Distance on Faultline Strength to Figure 2. The Moderating Role of Faultline Distance on Faultline Strength to Task Learning
4,9 5,1 5,3 5,5 5,7 5,9
low med high
Faultline Strength
Task Learning Faultline Distance
high low
MEDIATORS DEPENDENT VARIABLES Psychological Transactive Safety Memory Task Learning Process Learning H2a H3a H1 H2b H3b H1 H2b H3b Team Heterogeneity¹ .15 .25 .25 .20 .15 .06 .00 -.08 Faultline Strength (Fau Str) .28* .22 .40** .30* .31* .29* .18 .16 Faultline Distance (Fau Dist) -.25† -.14 -.25† -.17 -.19-.36* -.26† -.28** Fau Str x Fau Dist -.36** -.39** -.30* -.17 .14 -.27* -.11 -.03 Psychological Safety .34* .41** Transactive Memory .41** .58*** Change in R² .09 .13 .13 .27 F Change 6.21* 9.22*** 9.18** 23.87*** R² .22 .21 .22 .32 .35 .22 .35 .49 Adjusted R² .24 .28 .28 .43 F 3.22* 3.14* 3.37* 4.24** 5.02*** 3.69* 5.00*** 8.78*** Note. ¹ N = ranges from 52-67, ¹ Based on educational level and prior work experience (years) *** p < .001, ** p <.01, p <.05, † p <.10
Table 3. Hierarchical Linear Regression Results
We found support for our hypothesis 2a, indicating that faultline strength was positively associated with psychological safety when faultline distance was small (β = -.36, p < .01; see Figure 4). To test whether psychological safety mediated the relationship between the interaction of faultline strength and distance and task and process learning (hypothesis 2b), three conditions must be met (Baron & Kenny; 1986; Muller, Judd, & Yzerbyt, 2005). First, there must be a significant relationship between the interaction term and the dependent variable. Second, there must be a significant relationship between the interaction term and the mediator. Finally, when including the interaction term and the mediator in a regression analysis, the mediator should remain significant and the relationship between the interaction term and the
dependent variable should disappear or be suppressed. We found support for the first two conditions in our testing of hypothesis 1a, 1b, and 2a. Psychological safety indeed mediated the relationship between the interaction between faultline strength and distance and task and process learning, providing support for hypothesis 2b. When the interaction term was included in the regression analysis together with the mediator, the mediator remained significant for both task (β = -.34, p < .05) and process learning (β = .41, p < .01), while the
interaction term became non-significant (see Table 3). The Sobel test showed that the mediation was significant for process learning (z = -2.04, p < .05) and nearly significant for task learning (z =-1.83, p = .06). Therefore, hypothesis 2a and 2b were supported. Groups with strong but close subgroups displayed higher levels of task and process learning and this was due to higher levels of psychological safety.
4,20 4,40 4,60 4,80 5,00 5,20
low med high
Faultline Strength
Psychological Safety Faultline Distance
high low
Figure 4. The Moderating Role of Faultline Distance on Faultline Strength to Psychological Safety
In hypothesis 2c we proposed that task and process learning, in turn, would mediate the relationship between psychological safety and performance.
According to the procedure described by Baron and Kenny (1986), we tested whether there was a significant relationship between the independent variable and the dependent variable. Regression analysis showed a marginal
relationship between psychological safety and team performance β = .23, p = .07), indicating that the first mediation step was only marginally supported.
Therefore, we did not proceed with further mediation steps.
In line with what was expected, we found support for hypothesis 3a proposing that faultline distance would moderate the relationship between faultline strength and transactive memory. A significant interaction between faultline strength and distance and transactive memory (β = -.39 p < .01) indicated that faultline strength was associated with higher levels of transactive memory when distance was small (see Figure 5).
The first two steps of the mediation as proposed in hypothesis 3b were confirmed in our tests of hypothesis 1a, 1b, and 3a. To test whether transactive memory would be a significant mediator in the relationship between the interaction of faultline strength and distance on task and process learning we conducted two regression analyses in which we included the interaction term together with the mediator. Transactive memory was a significant mediator in the relationship between the interaction of faultline strength and distance and task learning (β = .41, p < .01), and process learning (β = .58, p < .001). The Sobel tests were significant for both task learning (z = -2.15, p < .05) and process
4,20 4,40 4,60 4,80 5,00 5,20
low med high
Faultline Strength
Transactive Memory
Faultline Distance high
low
Figure 5. The Moderating Role of Faultline Distance on Faultline Strength to Transactive Memory
learning (z = -2.38, p = .05). Therefore, hypothesis 3a and 2b were supported.
Groups with strong but close subgroups displayed higher levels of task and process learning and this was due to higher levels of transactive memory.
In hypothesis 3c we proposed that task and process learning would mediate the relationship between transactive memory and performance. In the first step mediation step (Baron & Kenny, 1986) we found a significant
relationship between transactive memory and team performance (β = .25, p <
.05). We also found support for the second step, indicating that there was a positive association between transactive memory and task learning (β = .51, p <
.001) and process learning (β = .61, p < .001). However, the final mediation step was not supported.
Discussion
The purpose of this study was to examine the interplay between faultline strength and distance on team learning and performance. Past faultline research has mainly focused on the effects of faultline strength on group processes and performance, largely neglecting the effect of faultline distance (cf. Bezrukova et al., 2009). However, a team with two subgroups in a team that have different job functions and differ between five years of work experience and twenty-five years of work experience are likely to have very different group dynamics. The faultline strength of these groups is the same, but faultline distance can explain why these teams will have different group processes.
In contrast to past research that often has proposed negative faultline effects, we proposed in this study that strong but close subgroups can have positive effects on team learning. Informational differences, such as educational level and prior work experience, can be important for the team to learn, but at the same time similarities are needed to make it safe for team members to bring forward their viewpoints and to have members from different subgroup listen to each other. Hence, being different yet similar is important for group functioning. Based on cohort formation theory (Asch, 1952; Walsh, 1988) we argued that when team members form strong but close informational cohorts, team members are more likely to bring forth their knowledge and opinions in the team and more likely to be listened to. We argued that this will stimulate team learning.
Our results support our predictions, showing that when the distance between subgroups is small, subgroups can stimulate task and process learning.
Additionally, we tested two underlying mechanisms that can help explain these relationships. The first mechanism – psychological safety – focuses on how interpersonal relations are managed in the team and can be defined as the extent to which team members perceived the team as safe for interpersonal risk taking. The second mechanism – transactive memory – was focused on how task-relevant expertise was managed and comprises the extent to which team members know each other’s expertise well. The results showed that both mechanisms acted as mediators in our model and helped to explain the
relationship between faultlines and team learning. Groups with strong but close subgroups were perceived as psychologically safe and developed the best transactive memory systems, which was positively associated with task and process learning. Additional analyses showed that when both mechanisms were taken into account simultaneously, transactive memory appeared to be the strongest mediator in the relationship between faultline strength and distance and team learning. This might be due to the fact that transactive memory is a more task-related factor than psychological safety, which is more important for task and process learning. Moreover, transactive memory was highly relevant for the teams in our sample, since these teams often consisted of members with varying expertise and levels of work experience.
Additionally, we found that transactive memory was positively related to task and process learning and team performance. However, task and process learning did not mediate the relationship between transactive memory and team performance. It could be that a team that manages the available expertise well is just much more efficient and therefore performs well, without team members necessarily learning from each other’s expertise. The team as a whole can learn from a transactive memory system, but it does not necessarily have to be the underlying factor that helps explain why they perform better.
We also contributed to literature on team learning by measuring two types of team learning that have recently been theoretically distinguished (Jehn
& Rupert, 2007): task and process learning. We empirically linked these types of team learning to group composition and factors relevant for team learning, such as psychological safety and transactive memory, and also related them to supervisor ratings of team performance. We found that the interplay between faultline strength and distance directly influenced the types of team learning, with strong but close subgroups displaying the highest levels of task and process learning in our sample. Additionally, and in line with previous research
(e.g., Edmonson, 1999; 2003; Lewis, Lange, & Gillis, 2005) we found that psychological safety and transactive memory were important team learning facilitators. Interestingly, process learning was significantly associated with team performance while task learning was not. Process learning is learning about how to cooperate with each other and to develop and adjust work processes and routines according to what is effective, while task learning is learning about the content of the task. The finding that process learning is more important for team performance than task learning is in line with the notion that teams mostly suffer from process and coordination losses (Steiner, 1972).
Hence, there is much to gain from effective team processes. It might also be the case that the teams in this sample had a relatively well developed mental model of the content of the task to perform and were more focused on coordination and management issues in the teams. Process learning might therefore be of greater importance to these teams than task learning. Future research should conduct research in other settings to see whether process learning is indeed a better predictor of team performance than task learning and see under what circumstances and in which developmental phase teams benefit from different types of learning.
Limitations and Future Research Directions
This study has a number of limitations. First, this study is cross sectional, which limits us in drawing causal conclusions. Future research should therefore collect longitudinal data and test the relationship between faultlines and team learning in an experimental setting in order to facilitate causal inference.
Second, this study has been conducted in two health care organizations with the same types of teams, performing similar tasks. The findings of this study might have limited generalizability to teams in other sectors performing different tasks. Future research should therefore replicate these findings in other settings.
Additionally, future research should consider different demographic attributes that objective faultlines can be based on. In this study we looked at faultlines based on educational level and prior work experience since we were interested in informational characteristics that were relevant for team learning (Argote et al., 2000; Hinsz et al., 1997; Van der Vegt et al., 2006). We found that these informational faultlines explained variance above and beyond general levels of team heterogeneity based on these characteristics, giving support for
the faultline perspective (Lau & Murnighan, 1998). As part of our sample had a relatively high level of relevant work experience gained outside and prior entering the team, faultlines based on prior work experience were much more influential than faultlines calculated based on total work experience (which consists of the sum of work experience prior and after entering the team). The fact that our findings hold for faultlines based on educational level and prior work experience makes sense as work experience gained outside the team represents a form of expertise that is unique and different than work experience gained together with other team members in the team. This expertise is highly relevant for the concepts of transactive memory and team learning that we investigated in this study. Future research should therefore specify what sort of work experience is being studied and consider which demographic
characteristics are most relevant given the constructs under study.
Finally, future research should examine other moderators and
mediators that may specify the conditions under which subgroup formation can boost team learning and disentangle the processes underlying this effect. Prior research, for instance, has indicated that moderators such as diversity beliefs and superordinate identity can help to weaken negative faultline effects (Bezrukova et al., 2009; Homan et al., 2007b). Also, information sharing has been found to be an important mediator in the relationship between faultlines and team performance (Homan et al., 2007b; 2008; Sawyer et al., 2006), but has not yet been related to team learning in faultline groups. These and other variables could help us to uncover potential benefits of faultlines for team learning.
Conclusion and Managerial Implications
In sum, this study illustrates the importance of considering faultline distance in research on faultlines and team learning. We hypothesized and found that when subgroups are strong but close, this can stimulate task and process learning in teams. Additionally, we found two underlying mechanisms that help explain these relationships: psychological safety and transactive memory. In teams with strong but close subgroups team members are more likely to experience the team as psychologically safe. These teams also have well-developed transactive memory systems, which stimulate team members to learn and to efficiently utilize expertise in order to perform well as a team. This last factor, which is focused on the management of expertise, was the strongest
mediator for both task and process learning. Finally, process learning was positively associated with team performance.
For managers, it is important to realize that subgroup formation in teams can be good, as long as distances between subgroups do not become too large. Similarities between members within the same subgroup can help team members to put forth their expertise and views within the team as a whole. When these opportunities exist in the team and members from different subgroups listen to each other, team members are more likely to experience the team as psychologically safe and team members are more likely to have
accurate perceptions of who knows what in the team. In teams in which subgroups are more distant, managers could stimulate team members to exchange knowledge and information with each other, thereby helping the team to build an accurate transactive memory system, which can stimulate team learning and performance. Additionally, in teams with strong but distant faultlines managers should focus on potential process losses and coordination processes that can disrupt performance. In order to benefit from the
informational differences available to the team, team members must experience the similarities between each other – hence being different yet similar.
