Group Dynamics: Theory, Research, and Practice © 2014 American Psychological Association 2014, Vol. 18, No. 1, 69–86 1089-2699/14/$12.00 DOI: 10.1037/a0035161 Exploring the Benefits and Boundaries of Transactive Memory Systems in Adapting to Team Member Loss

Jessica Siegel Christian, Aleksander P. J. Ellis Matthew J. Pearsall, University of Arizona and Michael S. Christian University of North Carolina

This study examines how teams respond to unplanned member loss. We draw on theory of team compilation and adaptation to suggest that teams with well-developed trans- active memory systems (TMS) will be better equipped to withstand the loss of a member. Then, based on role criticality theories, we argue that those effects depend on which member is absent, such that when a more critical member is lost, the perfor- mance benefits of a TMS are reduced. Finally, we reason that this interactive effect is because of the team’s ability to engage in plan formulation. We tested and found support for our hypotheses using 78 four-member teams engaged in a command-and- control simulation. TMS positively affected team performance following the loss of a member, but the benefits of the TMS were reduced following the loss of a critical member because teams had more difficulty engaging in plan formulation. We discuss how the results of this study add to our understanding of the precursors of successful team adaptation.

Keywords: team adaptation, transactive memory, critical team member

Team-based organizations often operate in thieu, 2000; Waller, 1999). In organizations, dynamic environments where survival depends teams must adapt to a wide-ranging set of cir- on the ability to successfully adapt to changing cumstances, including external contingencies circumstances (Kozlowski, Gully, Nason, & such as communication equipment failures Smith, 1999). Increasing competition, global- (LePine, 2003, 2005) or unfamiliar performance ization, and technological changes have created contexts (Marks et al., 2000) and internal struc- a need for more flexible and adaptive responses tural contingencies such as planned downsizing (Kozlowski & Bell, 2003; Volberda, 1996). (DeRue et al., 2008) or member replacement Teams are thought to have adaptive advantages (e.g., Moreland, 1999). The nature of the con- over individuals (Kozlowski et al., 1999), lead- tingency is important because it determines ing researchers to turn their focus toward team both the extent to which specific team processes adaptation as an essential performance criterion are disrupted and the team’s ensuing response (e.g., Burke, Stagl, Salas, Pierce, & Kendall, (Burke et al., 2006; Kozlowski & Klein, 2000). 2006; Chen, Thomas, & Wallace, 2005; DeRue, The first purpose of this study is to expand Hollenbeck, Johnson, Ilgen, & Jundt, 2008; our understanding of the range of contingencies This document is copyrighted by the American Psychological Association or one of its allied publishers. LePine, 2003, 2005; Marks, Zaccaro, & Ma- teams face to include situations where a team This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. member is unexpectedly lost or absent and not replaced. This situation is commonplace. In ac- tion teams such as military units or firefighters, Jessica Siegel Christian, Matthew J. Pearsall, and Mi- chael S. Christian, Kenan-Flagler Business School, Univer- members are sometimes injured or incapaci- sity of North Carolina; Aleksander P. J. Ellis, The Eller tated, leaving the team short on resources and College of Management, University of Arizona. skills. In project, decision-making, and cus- Correspondence concerning this article should be ad- tomer service teams, members may unexpect- dressed to Jessica Siegel Christian, University of North Carolina, Kenan-Flagler Business School, McColl Building edly need to travel, fall ill, or be called away on 4734, Chapel Hill, NC 27599. E-mail: jessica_christian@ other assignments (see Sundstrom, 1999). For unc.edu these types of teams with specialized expertise,

69 70 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

losing a member can be particularly detrimental important predictor of adaptive success (i.e., because of the concurrent loss of the member’s team performance) in situations where teams unique knowledge and skills. Despite this, no lose a member. research has examined factors that predict suc- However, the potential benefits of a TMS cessful performance in the situation where a may depend on the role played by the lost team member is unexpectedly lost. Research suggests member. Not all positions within a team are that adjusting to a change in the task or envi- equally important to the team’s workflow (Ar- ronmental structure requires a series of contin- row & McGrath, 1993; Kozlowski et al., 1999). uous developmental processes that compile According to Humphrey, Morgeson, and Man- over time (Burke et al., 2006). The unexpected nor’s (2009) strategic core theory, team mem- loss of a member represents a different situation bers in highly critical positions are more “core” and requires the team to quickly find new ways to the team. We argue that teams with a well- to redirect the remaining members’ distributed developed TMS will be able to withstand the knowledge to perform at a high level. We are loss of a less critical team member, but will interested in teams’ initial response to the loss have more trouble compensating for the loss of of a member and how members quickly adjust a highly critical member. their behaviors to meet this unexpected chal- Our final objective is to uncover the mecha- lenge. nism underlying this conditional relationship. Our second purpose is to uncover factors that Kozlowski et al. (1999) argued that adaptive help and hinder adaptation to member loss. In teams must have the ability to choose a plan for terms of successful adaptation, we examine action. In an effort to respond effectively to new team performance following the loss of a mem- demands, teams must quickly devise a strategy ber. Kozlowski et al.’s (1999) compilation the- to address the change so that the lost members’ ory suggests that team members are better responsibilities are not neglected (Burke et al., equipped to deal with losing a team member 2006; Stout & Salas, 1993). We argue that the when they understand the roles and responsibil- loss of critical nodes within the TMS will ham- ities of their teammates and how they connect per the availability of information required for together. Therefore, we suggest that a team’s the team to effectively formulate adaptive plans. transactive memory system (TMS)—an orga- Therefore, we expect plan formulation to rep- nized store of knowledge that is contained en- resent a key mechanism explaining why losing tirely in the individual memory systems of a critical member reduces the effects of TMS on group members; a set of knowledge-relevant team performance following member loss. transactive processes that occur among group With this research, we contribute to the team members (Wegner, Giuliano, & Hertel, 1985)— adaptation literature in several ways. First, we plays an important role in how teams perform expand the typology of adaptive contexts by following the loss of a member. examining team performance following the loss Our study is not the first to examine TMS in of a team member. Second, based on the unique adaptive team contexts. Previous research has requirements of this form of adaptation, we provided valuable insight into how a TMS im- identify a team’s TMS as a key determinant of pacts team effectiveness when new members success. Third, we establish criticality as an are added to a team (e.g., Levine & Choi, 2004; important boundary condition that determines Lewis, Belliveau, Herndon, & Keller, 2007) and the degree to which performance is influenced This document is copyrighted by the American Psychological Association or one of its allied publishers. how the benefits of a TMS are reduced when by TMS. Finally, we uncover and examine plan This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. team membership is scrambled after training formulation as a mechanism underlying the in- (Moreland, 1999; Moreland & Argote, 2003; teractive effects of TMS and criticality on team Moreland, Argote, & Krishnan, 1996; Moreland performance. & Myaskovsky, 2000). However, these studies have generally viewed TMS as an outcome vari- Transactive Memory Systems and Team able that is negatively affected by the addition Performance Following Member Loss of newcomers, or as a mediator of the effects of team training strategies. We take the literature Transactive memory was originally con- on TMS and team adaptation in a different ceived by Wegner and colleagues (Wegner, direction, suggesting that TMS represents an 1987; Wegner et al., 1985) as a way to explain TRANSACTIVE MEMORY AND TEAM MEMBER LOSS 71

the behavior of individuals in close relation- that reflect emergent cognitive aspects of trans- ships. Wegner noticed that couples tend to have active memory: memory differentiation (devel- a system for processing information that allows opment and recognition of specialized areas of them to do more without exhausting their cog- expertise within the team), task coordination nitive resources, which he labeled transactive (effective combination of knowledge), and task memory. Each member of a couple develops a credibility (trust in one another’s competence specific domain of knowledge and expertise, and expertise; Liang, Moreland, & Argote, taking responsibility for any relevant informa- 1995). TM has been conceptualized as “a com- tion within that domain. By doing so, they es- bination of knowledge possessed by each indi- tablish information exchange patterns between vidual and a collective awareness of who knows differentiated memory banks that allows them what,” (Austin, 2003, p. 866). Lewis (2003) to more efficiently and effectively complete provided additional refinement of these dimen- tasks. For example, a wife may remember fam- sions, which resulted in a widely used scale ily birthdays better than her husband and, there- measuring specialization (the level of memory fore, will be recognized as the “expert” in birth- differentiation within a team), coordination (the day scheduling. This frees the husband from ability of members to work together effec- having to encode and store this information in tively), and credibility (members’ beliefs about his own memory because he knows that he can the reliability of other members’ knowledge). access the information regarding birthdays from Social psychological researchers have often her at any time. focused on the communications associated with At the team level, a TMS operates much like these emergent cognitions, primarily following a computer network, in which each of the group Wegner’s (1987) conceptualization of the con- members are information nodes that their team- struct, focusing on directory updating, informa- mates can use to access specific information or tion allocation, and retrieval coordination. This to send information to be stored for later re- approach focuses on the “transactions” between trieval by other members of the team (Wegner, team members—encoding and retrieval process 1995). These systems involve the knowledge of (Hollingshead, 1998b, 1998c; Wegner, 1995)— individual group members, combined with a that reflect the emergence and development of shared awareness of who knows what (More- the underlying transactive memory system. Di- land & Myaskovsky, 2000). Wegner et al. rectory updating refers to learning what others (1985) defined transactive memory as having know, information allocation refers to commu- two components: an organized store of knowl- nicating new information to a member whose edge contained within the individual memories expertise will facilitate its storage, and retrieval of group members, and a set of knowledge- coordination refers to retrieving information on relevant transactive processes that occur across a needed topic from a member with expertise in members. By distributing responsibility of dif- the domain of interest. Although this work be- ferent domains of expertise across the team, gan with examinations of dyads (Hollingshead, transactive memory provides an information 1998a, 1998c; Wegner, Erber, & Raymond, processing system that enables teams to develop 1991), it has been extended to work teams (see deep, discrete domains of expertise, reducing Ellis, 2006). It is important that the underlying their cognitive load while providing access to cognitive structure of the TMS precedes the their combined knowledge to all team members transactions (see Ellis, Porter, & Wolverton, This document is copyrighted by the American Psychological Association or one of its allied publishers. (see Hinsz, Tindale, & Vollrath, 1997). There- 2008). Thus, research on transactive memory This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. fore, it enables “quick and coordinated access to communications (i.e., transactions) assumes deep, specialized knowledge, so that a greater that a TMS structure is already in place and is amount of task-relevant expertise can efficiently smoothly operating. be brought to bear on team tasks” (Lewis, 2003, Both approaches represent viable operation- p. 587). alization of transitive memory in studies of Two different yet compatible approaches to teams. In our research, we are particularly in- studying transactive memory have emerged in terested in team processes and thus focus on the extant organizational and social psycholog- transactive memory communications because ical literatures. Organizational researchers typ- (a) we are interested in how teams behave and ically break the construct into three dimensions communicate in adaptive situations, (b) the 72 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

teams examined in this study were formed for a 2004; Liang et al., 1995; Moreland & Myask- short period of time which makes the credibility ovsky, 2000; Ren & Argote, 2011). dimension less applicable, and (c) our study When faced with the loss of a member, these involved “preprogramed” levels of distributed benefits translate directly into preserving the expertise. team’s ability to function. Compilation theory To develop hypotheses regarding the effect suggests that adapting to such a change requires of transactive memory following member loss, team members to have a well-developed under- we use Kozlowski et al.’s (1999) compilation standing of the relationships among members theory. Compilation theory argues that teams and the ability to coordinate knowledge ex- develop adaptive capabilities over time by shift- change. When a team member is lost, the team ing their focal level and paying attention to must act quickly to account for the increased different content, processes, and outcomes in workload and the loss of that member’s knowl- different stages. In the first phase, team forma- edge. We suggest that a TMS provides these tion, team members seek basic information benefits, although this notion has yet to be em- about the nature of the team and their place in it. pirically tested. Previous work in this area has In the second phase, task compilation, members examined membership change and how the focus on demonstrating their task competency TMS and knowledge structure is affected by and begin to monitor and regulate performance newcomers or complete membership turnover. on their individual tasks. In the third phase, role Lewis et al. (2007) examined how membership compilation, members explore dyadic relation- change impacted TMS, finding that partially ships with other members—they begin to un- intact groups relied on an outdated TMS struc- derstand how their task outputs affect other ture in a second experimental session, which members with whom they directly interact. The hindered team performance compared with in- final phase, team compilation, involves a focus tact groups and completely reconstituted groups. Baumann (2001) compared TMS devel- on the entire team. In this phase, members de- opment in intact groups to development in velop an understanding of the role linkages newly composed groups and found that when among the entire team, and they learn how to members were assigned to work with a new improve their network of roles. group with the same role structure to their pre- According to compilation theory, TMSs de- vious group, they could still develop a TMS. velop through dyadic role-based exchanges in Other work has addressed membership the role compilation phase of team development changes in teams but has not specifically ad- (also see Pearsall, Ellis, & Bell, 2010). Through dressed the role of TMSs. For example, Kane, these role-based exchanges team members learn Argote, and Levine (2005) examined how “who knows what” within the team and begin to knowledge was transferred when individuals coordinate the flow of information with team- were rotated among groups, and Levine, Choi, mates who are responsible for storing or apply- and Moreland (2003) examined conditions that ing it. During this phase of team development, result in newcomer knowledge transfer and per- members engage in a “mutual process of role suasive abilities. Summers, Humphrey, and Fer- negotiation,” (Kozlowski et al., 1999, p. 266) ris (2012) showed the importance of flux, a where they learn how to tailor their interactions critical emergent state during member change and coordinate with other members. The more and how it is impacted by which team member This document is copyrighted by the American Psychological Association or one of its allied publishers. these activities take place, the more well- is replaced and the newcomer’s relative cogni- This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. developed the TMS (Palazzolo, Serb, She, Su, tive ability. Although these works have pro- & Contractor, 2006) and the greater the amount vided valuable insight into how the TMS func- of resources that can be exchanged among tions when a new member is brought into the members (Balkundi & Harrison, 2006). TMSs team (Lewis et al., 2007) and general member- therefore allow for greater performance effi- ship change, they have not addressed the role of ciencies as members are required to handle only the TMS in adapting to losing a member. A key the information they need while receiving the differentiating factor is that in studies of mem- task-related information necessary to promptly bership change, teams still have a full comple- execute their tasks, improving team coordina- ment of knowledge, skills, and manpower to tion, performance, and satisfaction (e.g., Lewis, accomplish their work. When teams lose a TRANSACTIVE MEMORY AND TEAM MEMBER LOSS 73

member—with no replacement—remaining turning their attention to the realignment of members must shift responsibilities among work, which is most critical in the short term. themselves without having a full set of re- Thus, we are specifically interested in how the sources. original TMS functions when a node is re- A TMS should help the team survive the loss moved. We hypothesize the following: of a member by providing members with an understanding of the patterns of knowledge ex- Hypothesis 1: Transactive memory will change and the relationships between members, positively affect team performance follow- which promotes adaptability by allowing them ing the loss of a member. to rapidly respond with appropriate behaviors (Kozlowski et al., 1999). Teams with a func- tioning TMS will be better able to actively The Effects of Lost Member Criticality reassign task duties based on team members’ existing knowledge domains and are “more Certain positions within the team’s workflow likely to fully utilize members’ expertise and are likely to have a greater influence on perfor- realize the value of embedded team knowledge” mance than others. Such positions are consid- (Lewis, 2004, p. 1519). Understanding connec- ered critical in nature, given that their removal tions between members should help the team “breaks the workflow chain” (Brass, 1984,p. “reconfigure the network to meet immediate 522) and impair the team’s ability to distribute internal or external contingencies” (Kozlowski and process information. A critical position is et al., 1999, p. 254). As team members are defined by the number of irreplaceable connec- accustomed to frequently searching for and dis- tions a team member has within the relationship tributing information within the team based on network (Pearsall & Ellis, 2006). That is, the their teammates’ knowledge domains, TMS relative criticality of a role is determined by the communication patterns within the team will number of other team members that can perform help reduce the amount of information and ca- that person’s function with the actions of less pacity that is lost along with the missing mem- critical members more easily replaced in the ber and provide performance efficiencies that team (Brass, 1984, 1985). Criticality therefore minimize any loss in team performance. A refers to substitutability; if other positions working TMS creates a shared understanding of within the workflow network cannot easily re- member–expertise associations, which helps place or substitute for a focal position, the po- members to anticipate how other team members sition is high in criticality. Team members who will behave and react (Cannon-Bowers, Salas, occupy critical roles perform important func- & Converse, 1993; Lewis, 2004). In a recent tions; if these duties are not completed, other review article, Zajac, Gregory, Bedwell, members cannot complete their tasks. Thus, a Kramer, and Salas (2013) argued that for teams critical member possesses unique knowledge to successfully adapt in an ill-defined problem and information (Girvan & Newman, 2002), situation, members must know who holds what making this role more crucial for the achieve- knowledge and how to draw on that knowledge. ment of performance goals. This shared understanding should positively af- Teams are often designed to include a critical fect member behavior following a loss, as mem- member such that “one role is more tightly bers will implicitly understand which of the linked to the overall performance of the team This document is copyrighted by the American Psychological Association or one of its allied publishers. remaining members can fill in for the missing than are other roles” (Humphrey et al., 2009,p. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. member. 49). Such roles are more “core” to the team As discussed earlier, a critical question for because, according to the theory of the strategic teams in member loss situations is how they use core, these individuals (a) encounter more prob- their existing TMS to respond in the short term. lems that must be overcome in the team, (b) We are interested in unexpected and typically have greater exposure to team tasks, and (c) are short-term losses as opposed to permanent more central to the workflow of the team. Core downsizing that would likely necessitate large- roles are more critical in terms of resources and scale modifications to the TMS. When members workflow and therefore more critical to TMS, view losing a member as a temporary situation, given that a core member completes tasks that they will respond with work-around behavior, cannot be completed by other members. This 74 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

argument is consistent with work on structural to transform the current environmental state holes in networks; individuals who occupy core into the desired or goal state (Burke et al., 2006; roles have greater access to information within Rosen et al., 2011), and plans function as “a the network and greater control over the net- guiding framework, aiding team members with work (Burt, 1992, 1995). their interpretation and reaction to the environ- In essence, the criticality of a missing mem- ment” (Rosen et al., 2011, p. 110). ber determines the size of the “hole” in the This type of planning helps teams following a team’s TMS left by his or her absence. When a nonroutine event. For example, Waller (1999) more critical team member is lost, the size of found that teams that quickly engaged in adap- the “hole” is substantially larger than if a less tive planning behaviors following a nonroutine critical team member is missing. Although each event outperformed teams that did not engage in possesses unique skills and expertise, the re- such behaviors or began engaging in such be- sponsibilities of the less critical role are more haviors at a much later stage, and Zajac et al. easily filled by one or more of their teammates. (2013) discussed the crucial role of plan formu- Further, because of their importance to the pat- lation in helping teams to adapt to ill-defined terns of interaction within the team the loss of a environments. Within the context of adapting to more critical role is most likely to affect the member loss, we define plan formulation as whole team (Kozlowski et al., 1999), whereas communications that serve to “cover” the miss- the loss of a noncritical member has less impact ing members’ responsibilities, such as redistrib- on knowledge sharing and coordination within uting task duties or reprioritizing tasks to ac- the team. When a critical member is lost, it is count for the loss. To effectively respond to an more likely that workflow systems that the team emergent demand such as the loss of a team previously relied on will no longer function member, teams must quickly engage in these properly, causing a breakdown in the TMS. As types of communications so that the lost mem- such, the benefits of TM during team perfor- ber’s responsibilities are not neglected (Burke et mance following the loss of a member will al., 2006; Stout & Salas, 1993). diminish as the criticality of the lost member We argue that although plan formulation is increases. As a result, we hypothesize: crucial for adaptive performance in general Hypothesis 2: The effects of transactive (e.g., Burke et al., 2006; Waller, 1999; Zajac et memory on team performance will be al., 2013), the degree of its occurrence follow- moderated by the criticality of the lost ing member loss (and resultant impact on team team member, such that the benefits of performance) is driven by the interaction be- transactive memory will be reduced when tween transactive memory and lost member crit- a team loses a more critical member. icality. Plan formulation is influenced by the combination of transactive memory and lost The Mediating Role of Plan Formulation member criticality, such that when either of these factors is altered, plan formulation will also Compilation theory (Kozlowski et al., 1999) be more or less likely to occur (see Figure 1). This and other theories of adaptation (Burke et al., is because when a team loses a more peripheral 2006) suggest that plan formulation plays a member, the team members are able to use their significant role in the relationship between TM, TMS to work around the absence and formulate a This document is copyrighted by the American Psychological Association or one of its allied publishers. criticality, and team performance in member plan for dealing with the loss. However, when This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. loss situations. Plan formulation is defined by teams lose a more critical member, the benefits of Burke et al. (2006) as “deciding on a course of a TMS will be reduced because it will be more action, setting goals, clarifying member roles difficult for team members to come up with a plan and responsibilities with the context of a course of attack in dealing with the loss. In essence, more of action, discussing relevant environmental “nodes” will need to be reconfigured, communi- characteristics and constraints, prioritizing cation channels will need to be restructured, tasks, clarifying performance expectations, and knowledge will need to be absorbed into team sharing information related to task require- members’ knowledge domains, and workload will ments” (p. 1194). During plan formulation, need to be redistributed within the team. Summers teams generate a sequence of actions designed et al. (2012) argued that teams experiencing mem- TRANSACTIVE MEMORY AND TEAM MEMBER LOSS 75

Lost Member Criticality

Transactive Memory Plan Formulation Team Performance System

Figure 1. Hypothesized model.

ber change to more strategically core roles should destroying any enemy targets as quickly as pos- have more difficulty developing new patterns of sible. interaction and routines; an idea which was sup- Geographic region. The geographic region ported by their data. Team members will need to is partitioned into four quadrants of equal size, establish new patterns of interactions to accom- one for each team member. In the center of the plish all of their tasks, given that a “resource hub” square is a 4 ϫ 4 square titled the “highly is missing. Therefore, we hypothesize the follow- restricted zone.” This square is nested in a ing: larger 12 ϫ 12 square titled the “restricted zone.” Outside the restricted zone is neutral Hypothesis 3: Plan formulation will mediate the interactive effects of lost member criti- space. Bases and vehicles. cality and transactive memory on team per- In terms of monitoring formance following the loss of a team the restricted zone, team members have a home member. base in the center of their quadrant. Each quad- rant has a detection ring around it allowing team members to detect the presence or absence of Method targets within their base. To detect targets that are not covered by team members’ detection Participants ring, they must request aid from teammates or Participants included 312 undergraduate use the vehicles stationed at their base. Team students from a large Southwestern United members were assigned to have four of a certain States university who were arrayed into 78 type of vehicle stationed at each base. There are four-person teams. Of the 312 participants, four different types of vehicles: (a) AWACS 59.4% were female, 67.9% were White, and (surveillance planes), (b) tanks, (c) helicopters, their average age was 21 years. Participants and (d) jets. Vehicle assets vary along five ca- received class extra credit in exchange for pabilities: range of vision, speed, duration of their participation and were also eligible to operability, identification ability, and power. receive $40 each ($160 per team) based on Assets are distributed such that each vehicle has their team’s performance. strengths and weaknesses. For example, AWACS has the greatest range of vision but no This document is copyrighted by the American Psychological Association or one of its allied publishers. power to destroy enemy targets. Tanks have the

This article is intended solely for the personal use of the individualTask user and is not to be disseminated broadly. most power but are the slowest vehicles and Participants engaged in a modified version of have a very limited range of vision. Although distributed dynamic decision-making (DDD) all vehicles can detect targets, only AWACS simulation (see Miller, Young, Kleinman, & has the ability to identify targets as friendly or Serfaty, 1998). The DDD is a computerized, enemy and share the information with the rest of dynamic command and control simulation re- the team. quiring participants to monitor and defend a Targets. When targets enter any team geographic zone against enemy targets. The ob- member’s detection ring, they show up as un- jective of the task is to protect the region by identified. Once the target is identified by the identifying targets as friendly or enemy and AWACS plane, any other team member can 76 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

engage it with a tank, helicopter, or jet, depend- After the training task, teams performed a ing on the strength of the particular target. If the 40-min developmental task, in which each vehicle has the correct level of power (greater member had a specific area of expertise. Team than or equal to the target), then the target can members were given a “specialty sheet” that be successfully destroyed. In this study, teams illustrated their role for the game, which they faced four different types of targets: E, F, G, were able to keep during the entire task. For and H. The E targets are friendly with a power example, DM1’s specialty sheet stated that he of 0, and all other targets are unfriendly (enemy or she would play the role of AWACS and targets). The H targets have a power of 1, the G that E targets had a power level of 0 (friend- targets have a power of 3, and the F targets have ly). The purpose of the developmental task a power of 5. was to simulate the role compilation phase of Team member roles. During the develop- team development, allowing team members to mental and experimental tasks, team members begin interacting together and providing them were given specific areas of responsibility and an opportunity to develop their transactive expertise. These areas of expertise were created memory systems (see Kozlowski et al., 1999; by giving each member control over only one Pearsall et al., 2010). During this task, teams type of vehicle and dividing the knowledge played the simulation and worked together as a about the target power levels among the team- team to identify targets and destroy un- mates. During these tasks, DM1 had four friendly targets in the restricted zone. During AWACS planes and knew that E targets were this task, the researchers coded transactive Power Level 0 (friendly), DM2 had four tanks memory. At the end of 40 min, the experi- and knew that F targets were Power Level 5, menter ended the simulation. DM3 had four helicopters and knew that G At this point, teams experienced the lost targets were Power Level 3, and DM4 had member criticality manipulation (see Lost four jets and knew that H targets were Power Member Criticality Manipulation section). Level 1. Teams were given no time to strategize or plan; instead, the three remaining members immedi- Procedure ately began performing the 10-min experimen- tal task, during which plan formulation behav- Only one team was run through the exper- ior was coded. The experimental task was not iment at a time. Upon entering the laboratory, qualitatively different from the developmen- participants were randomly assigned to one of tal task in any way—teams still had the same the four stations (DM1, DM2, DM3, or DM4) responsibilities of identifying targets and de- as part of a four-person team. Teams were stroying unfriendly targets. The only differ- randomly assigned to one of two conditions ence was the lost member. Team performance (more critical team member loss vs. less crit- was recorded at the end of the experimental ical team member loss). In the laboratory, the task. four stations were situated close enough so that team members could comfortably hear Lost Member Criticality Manipulation one another verbally, but not closely enough that they could view the computer screens of Immediately prior to the 10-min experimen- their teammates. Throughout the experiment tal task, one member was removed from the This document is copyrighted by the American Psychological Association or one of its allied publishers. team members communicated with each other team. Criticality is based on the number of This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. verbally. After being seated at their stations, alternative workflow channels remaining if a participants were trained on the declarative team member is removed from a team; the and procedural knowledge necessary to com- higher the number of routes, the less critical the plete the tasks through audio training for ap- team member (Brass, 1984, 1985). Following proximately 15 min. Following the audio Ellis, Bell, Ployhart, Hollenbeck, and Ilgen training, participants engaged in a 30-min (2005) and Pearsall and Ellis (2006), the criti- training task where they learned how to cality of all the roles was created by distributing launch vehicles, identify and engage enemy assets unequally between team members. We targets, and began working as a team without then manipulated the relative criticality of the a specific area of expertise. removed member such that teams in the high- TRANSACTIVE MEMORY AND TEAM MEMBER LOSS 77

criticality condition had a more critical member mation to the team member with the correct removed (DM3), whereas teams in the low- specialty to handle the piece of information criticality condition lost a less critical member (e.g., “DM2, there’s an F target in the top left (DM4). In this task, whereas DM3 (who con- corner of my zone”). Retrieval coordination trols the helicopters) can destroy enemy targets occurred when members asked for informa- of both Power Levels 1 and 3, DM4 (who tion known to be part of another team mem- controls the jets) can only destroy enemy targets ber’s specialized information (e.g., “DM4, with Power Level 1. When DM4 is removed, what power are H targets again?”). Teams therefore both DM2 and DM3 can assume the engaging in frequent TM communications, responsibilities of the missing member and de- therefore, received higher TM scores than stroy the enemy targets for which DM4 was teams engaging in fewer transactions. This mainly responsible: Power Level 1 targets. variable was centered prior to analysis (see However, when DM3 is removed, only DM2 Aiken & West, 1991). can assume the duties of the missing member We chose the coding approach described (destroying Power Level 3 enemies), making above because in the laboratory, TMS behav- DM3’s role more critical to the team’s work- iors are observable, which is in line with flow (and less substitutable) than DM4’s. The Wegner, Giuliano, and Hertel’s (1985) con- team was simply told that at this point in the ceptualization. A strength of this approach is game, DM3 (DM4) was removed from every that it shows evidence of an effective and team and that DM3 (DM4) was now free to functioning TMS, which can be observed by leave. an outsider. According to Ren and Argote (2011), our approach represents the transac- Measures tive processes that occur during encoding, storage, and retrieval of information in a Transactive memory. Transactive mem- group’s collective memory, which goes be- ory was coded by the researchers during the yond the simple presence of knowledge developmental task. On the basis of the con- within the team. ceptualization of TMS by Hollingshead (Hol- Plan formulation. Plan formulation was lingshead, 1998b, 1998c, 2000), Ellis (2006), coded by the researchers during the experi- and Ellis et al. (2008), we operationalized mental task, following the removal of one TMS based on the number of times that team team member from each group. Following members engaged in directory updating, in- LePine (2003) and Waller (1999) and as rec- formation allocation, or retrieval coordination ommended by Rosen et al. (2011), we opera- based on their areas of expertise. Directory tionalized plan formulation as any communi- updating refers to members sharing or re- cation between team members within the first questing information concerning one anoth- 5 min that involved specifying the order of er’s expertise. Information allocation refers to importance of tasks or redistributing task du- team members directing the information to ties among members to respond to the loss of the member with the correct area of expertise. a specific member (Burke et al., 2006). Ex- Retrieval coordination refers to team mem- amples include “Let’s remember to send ve- bers using the knowledge of who knows what hicles into DM4’s (the missing member’s) to request information from the correct expert quadrant first,” or “DM2 you will need to take This document is copyrighted by the American Psychological Association or one of its allied publishers. (Wegner, 1995). These communications were out all G targets since we don’t have DM3.” This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. coded as the number of occurrences; each For both transactive memory and plan for- time a communication occurred, the coder mulation, three experimenters were in charge marked a “1” to create a count measure. An of observing teams and coding specific com- additive index (i.e., sum) represented the munications as they occurred. All three ex- number of behaviors (see Ellis, 2006). Direc- perimenters participated in training sessions tory updating occurred when members shared where (a) they were fully trained on the DDD their expertise or requested specific informa- task so that they would understand which TM tion from a teammate (e.g., I’m DM3 and I and plan formulation behaviors were accu- have helicopters”). Information allocation oc- rate, and (b) they were trained to identify curred when members communicated infor- transactive memory and plan formulation be- 78 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

haviors. During training, the experimenters Results discussed and reviewed construct definitions for each behavior, were given a list of exam- All analyses presented were conducted at the ple behaviors, and coded several practice team level. Intercorrelations and descriptive sta- teams to gain consensus and resolve disagree- tistics are presented in Table 1.1 ments. To establish interrater reliability, the exper- Hypotheses Testing imenters coded 18 of the teams in pairs. Spe- cifically, Experimenters 1 and 2 coded six Hypotheses were tested using hierarchical re- teams together, Experimenters 2 and 3 coded gression analyses. To test our hypotheses, we six teams together, and Experimenters 1 and 3 first entered team performance from the earlier coded six teams together. Fleiss’s (1971) ␬ 40-min developmental task in Step 1 of our provides an index of interrater agreement for regression to control for between-teams per- three or more raters. In this study, ␬ϭ.74 for formance differences and overall ability prior transactive memory, and ␬ϭ.79 for plan to the loss manipulation (e.g., DeRue et al., formulation behaviors, indicating acceptable 2008). Hypothesis 1 predicted that Team levels of agreement for both (see Landis & TMS would be positively related to team per- Koch, 1977). Coding disagreements were re- formance following the loss of a team member. solved by averaging together each coder’s TMS was positively related to team perfor- ␤ϭ ϭ ratings. Following these results, the remain- mance ( .23, p .05), providing support for ing 60 teams were divided equally between Hypothesis 1. the experimenters. The remaining hypothesis tests are shown in The coders were not blind to the general Table 2. Hypothesis 2 predicted that the criti- hypotheses that the focal behaviors were cality of the lost member would moderate the beneficial but had no way of determining relationship between TMS and team perfor- performance for teams in each experimental mance, such that the benefits of the TMS would condition (and would therefore be unable to diminish when a team loses a more critical inadvertently code more instances of plan for- member. As shown in Table 2 (column 2), the mulation behaviors for higher performing interaction between lost member criticality and teams). TMS was significantly related to team perfor- ␤ϭϪ Ͻ Team performance. The team perfor- mance ( .31, p .05). Further, as shown mance measure focused on the team’s main in Figure 2, teams with a well-developed TMS objective—maximizing the number of points in that lost a less critical member outperformed both their offensive and defensive scores (see teams with a less well-developed TMS. How- Ellis et al., 2003; Pearsall & Ellis, 2006). A ever, teams with a well-developed TMS that team’s offensive score went up by 5 points each lost a more critical member performed signifi- time an enemy target was destroyed in the re- cantly worse than those who lost a less critical stricted zone. However, if any team member member, at levels similar to teams with a less destroyed a friendly target or any target outside well-developed TMS. A simple slopes test in- the restricted zone, the team offensive score dicated that TMS was positively related to team dropped by 25 points. The team lost 1 point per performance when teams lost a less critical ␤ϭ ϭ Ͻ second on their defensive score for each enemy member ( .45, t 3.19, p .01), but it had This document is copyrighted by the American Psychological Association or one of its allied publishers. that resided within the restricted zone and 2 no significant effect on team performance when This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ␤ϭ ϭ points per second for each enemy that resided in teams lost a more critical member ( .03, t the highly restricted zone. Team performance 0.21, ns). In other words, the positive effects of was measured by standardizing and combining the team’s offensive and defensive scores dur- 1 Consistent with previous research, transactive memory ing the 10-min experimental task. Although was positively related to task performance prior to the loss teams were aware of their nominal score, they (r ϭ .29, p Ͻ .01). Prior performance and performance received no feedback concerning their perfor- following the loss of a member were positively, though not significantly, related (r ϭ .15). The relatively weak corre- mance relative to other teams and therefore had lation is likely because of the ineffectiveness of prior team no means of assessing their performance rela- routines and processes when faced with an adaptive shock tive to other teams. to the system. TRANSACTIVE MEMORY AND TEAM MEMBER LOSS 79

Table 1 Means, Standard Deviations, and Intercorrelations Among Variables of Interest

Variable MSD12 3 45 1. Prior performance (preloss) 0.05 0.77 — — ءءTMS 17.60 5.08 .29 .2 3. Lost member criticality 0.50 0.50 .05 .00 — — 19. ءPlan formulation 0.72 0.53 .03 .25 .4 — ءء30. ءءϪ.35 ءTeam performance 0.00 0.68 .15 .26 .5 Note. N ϭ 78 teams. Criticality is coded as 1 ϭ more critical member lost, 0 ϭ less critical member lost. Team performance refers to performance after the member loss manipulation. TMS ϭ transactive memory systems. .p Ͻ .01 ءء .p Ͻ .05 ء

the TMS on team performance remained strong see Table 2 (column 1). The third criterion, for following the loss of a less critical team mem- the mediator to significantly predict the depen- ber, but losing a more critical member reduced dent variable when controlling for the interac- the team’s ability to use its TMS to successfully tions between the moderator and the indepen- adapt. Therefore, Hypothesis 2 was supported. dent variable and between the moderator and Hypothesis 3 proposed that plan formulation the mediator, was also met. As shown in Table behavior would mediate the interactive effects 2 (column 3), plan formulation significantly of criticality and transactive memory on team predicted team performance when controlling performance following the loss of a team mem- for the two interaction terms (␤ϭ.39, p Ͻ .01). ber. To test Hypothesis 3, we followed the pro- Finally, the relationship between the indepen- cedures for mediated moderation outlined by dent variable and the dependent variable de- Muller, Judd, and Yzerbyt (2005). The first cri- creased significantly after entering the mediator, terion, for the interaction between the indepen- as the relationship between the TMS-lost mem- dent variable and the moderator to significantly ber criticality interaction and team performance predict the dependent variable, was evidenced decreased from ␤ϭϪ.31, p Ͻ .05 to ␤ϭ by the support for Hypothesis 2. The second Ϫ.19, ns (see Table 2). We then tested the criterion, for the interaction between the mod- significance of the indirect path using Preacher erator and the independent variable to signifi- and Hayes’s (2004) bootstrapping mediation cantly predict the mediator, was also met. The technique with 2,000 resamples. In the presence interaction between lost member criticality and of plan formulation behavior, the effects of the transactive memory was significantly related to interaction between transactive memory and plan formulation behavior (␤ϭϪ.32, p Ͻ .05), lost member criticality on team performance

Table 2 Regression Results of the Effects of Transactive Memory and Lost Member Criticality on Plan Formulation and Team Performance

Team performance This document is copyrighted by the American Psychological Association or one of its allied publishers. Plan formulation Step 1 Step 2 This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Variable ␤ t ␤ t ␤ t Prior performance (preloss) Ϫ.03 Ϫ0.26 .14 1.33 .17 1.62 1.58 24. 3.02 ءء43. 3.25 ءءTransactive memory (TM) .48 Ϫ3.45 Ϫ.21 Ϫ1.08 ءءLost member criticality (Crit) .19 1.83 Ϫ.35 Ϫ2.22 Ϫ.19 Ϫ1.35 ءϪ2.16 Ϫ.31 ءTM ϫ Crit Ϫ.32 3.07 ءءPlan formulation (PF) .39 PF ϫ Crit Ϫ.26 Ϫ1.25 R2 .16 .25 .33 Note. N ϭ 78 teams. Team performance refers to performance after the member loss manipulation. .p Ͻ .01 ءء .p Ͻ .05 ء 80 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

More Crical Member Lost Less Crical Member Lost

0.6

0.4

0.2

0

-0.2

-0.4 Low TM High TM

Figure 2. The effects of the interaction between transactive memory and lost member criticality on team performance. Graph depicts standardized mean for team performance.

were significantly reduced (95% confidence in- through which the team can marshal its skills terval [CI] Ϫ.042 to Ϫ.002). Therefore, plan and expertise to successfully adapt to this loss. formulation mediated the relationship between Then, using Humphrey et al.’s (2009) theory of the TMS-lost member criticality interaction and the strategic core, we identified an important team performance, providing support for Hy- boundary condition by showing that the advan- pothesis 3. tages of TMS are moderated by the member criticality. Specifically, teams with a well- Discussion developed TMS can survive the loss of a pe- ripheral member, but the benefits of the TMS Keeping teams together at all times is nearly are significantly reduced following the loss of a impossible in modern workplaces, where teams more critical member. Finally, we were inter- operate lean and members are often temporarily ested in identifying the mechanism underlying absent because of responsibilities on other proj- the interactive effects of TMS and criticality on ects and other teams, illness, travel, family team performance following member loss. We emergencies, or other commitments. We be- found that the effects of this interaction occur in lieve it is important to expand our conceptual- part because teams lose the ability to engage in This document is copyrighted by the American Psychological Association or one of its allied publishers. ization of adaptive contexts to include member plan formulation following the loss of a more This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. loss to ensure that we fully understand how critical member. teams react to changes in their environment. Therefore, the purpose of this study was to Theoretical Implications focus specifically on loss adaptation and to the- oretically identify predictors of team perfor- With this research, we are able to extend our mance in situations where teams suddenly need understanding of team adaptive demands to in- to operate without their full complement of clude adaptation to member loss. Similarly to knowledge and skill. On the basis of Kozlowski Pulakos, Arad, Donovan, and Plamondon et al.’s (1999) compilation theory, we found that (2000), who expanded adaptive typologies at TM represents an important emergent resource the individual level, we help to broaden our TRANSACTIVE MEMORY AND TEAM MEMBER LOSS 81

understanding of the adaptive performance do- ine, 2003; Waller, 1999), our results indicate main at the team level. This is an important that the team’s ability to quickly analyze and issue given that team members are often sud- respond to emergent demands on or within the denly absent because of travel, illness, and ro- team represents a crucial step in successful team tation among work teams (Barry, Kemerer, & adaptation, particularly when teams must recon- Slaughter, 2006; Mathieu, Maynard, Rapp, & figure their network structure and redistribute Gilson, 2008). Although researchers have be- task responsibilities. gun to examine changes in team structure (De- Finally, this study contributes to our under- Rue et al., 2008; Lewis et al., 2007; Moreland, standing of role distribution in teams. We ex- 1999), their focus has been on strategic planned tend previous work identifying the importance reorganizations and realignment of resources or of “core” team members (Humphrey et al., on the effects of a newcomer, rather than the 2009) and team member criticality (Pearsall & sudden removal of a member for which teams Ellis, 2006) to examine how the role of the lost are not formally briefed or given the opportu- member influences team interaction and perfor- nity to plan. For example, Lewis et al. (2007) mance. This study provides further support for compared the performance of intact groups with the idea that some role positions are more crit- that of reconstituted groups with varying num- ical than others within a team and that the bers of newcomers, finding that a TMS actually internal distribution of role knowledge and re- hindered the performance of partially intact sponsibilities has implications for how teams groups because such groups relied on an out- use information to actively adapt to member dated TMS structure. In contrast, our study in- loss. Although adopting a distributed, func- dicates that the effects of a TMS are largely tional structure allows teams to take advantage dependent on the adaptive context. When a of the diverse knowledge of skilled team mem- team member is abruptly removed from the bers (Cohen & Bailey, 1997; Smith-Jentsch, team, a TMS can actually help the team adjust Mathieu, & Kraiger, 2005), it leaves them vul- to the loss. Our study helps to differentiate nerable to the unexpected loss of a member and between adaptive contexts, indicating that a re- all the skills and expertise associated with that liance on inaccurate cognitions represents a hin- role. drance when newcomers join a previously intact group, yet a TMS can be beneficial when groups Practical Implications must “cover” for a missing member. We are also able to extend the literature ex- Our findings suggest that should a loss occur amining the benefits and boundaries of TMS. team members and managers need to first de- Although TMS has been shown to positively termine how critical the missing member was influence learning, performance and satisfaction within the network; a process that can be expe- for intact teams (Lewis, Lange, & Gillis, 2005; dited if managers identify criticality levels dur- Liang et al., 1995; Moreland & Myaskovsky, ing team formation. Although the team can sur- 2000) across a variety of task environments vive the loss of a less critical member, if a more (e.g., Lewis, 2003, 2004; Majchrzak, Jarvenpaa, critical member is unavailable the team needs to & Hollingshead, 2007), we show that develop- be made aware of the issue, and managers may ing a TMS is one way that teams can attempt to need to intervene to formulate a plan for re- mitigate the impact of the loss of a member on sponding to the temporary loss. One way of This document is copyrighted by the American Psychological Association or one of its allied publishers. performance. Therefore, we help to answer re- intervening is to hold a “plan formulation” This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. cent calls for research on transactive memory in meeting where the team is forced to come up adaptive situations (Zajac et al., 2013). with a strategic response before reconvening This research also contributes to the under- and attempting to complete the task. standing of mediating factors that help to ex- Another option for managers would be to plain the negative impact of losing a critical have a critical team member loss contingency team member. The breakdown occurs in part plan already in place. Perhaps this plan involves because teams that lose a more critical member replacing the lost member with someone else in can no longer engage in plan formulation, a the organization with the same set of knowledge vital adaptive behavior (Burke et al., 2006). and skills that is aware that they are on standby Similar to other team adaptation research (LeP- in case a problem occurs. Further, because the 82 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

quick recognition of the loss is so important, Limitations and Future Directions more critical members can be held responsible for alerting the team of their absence as early as As with any study, this study had several possible. limitations, including that this research focused Managers should also act preemptively be- on one type of team. In this study, we examined fore a loss occurs and actively encourage team teams that closely resemble action teams and members to engage in extensive role-based thus only represent one distinct type of work communications during the role compilation team. Action teams work together closely, usu- phase of team development (Kozlowski et al., ally for short periods of time, and are composed 1999; Pearsall et al., 2010), through which they of members who possess specific areas of ex- can develop shared cognition. However, al- pertise. Although other types of teams may also though these early interactions are critical for experience member loss, the current research establishing a base knowledge of “who knows might not completely generalize to teams with what” within the team, communications aimed longer tenures or less reliance on specialized at a deeper understanding of the breadth and information. For example, manufacturing or as- boundaries of teammates’ domains must con- sembly teams tend to have longer life cycles tinue to occur over time to strengthen and main- than action or project teams (Sundstrom, 1999). tain the network. Lewis et al. (2005) referred to In such teams, the loss of a critical member may these exchanges as higher order communica- have less impact as the team may have devel- tions, through which members more fully com- oped functional behaviors for dealing with sit- uations they have faced previously. In contrast, prehend their teammates’ roles and expertise established teams who rarely deal with un- and the interdependencies within the team. planned changes may fall into patterned behav- Managers should also encourage team members ioral routines and take much longer to respond to engage in reflection concerning the team’s to the loss of a member (Gersick & Hackman, TMS structure and relevance. In a supplemental 1990). Or, we can imagine that there are some study, Lewis et al. (2007) found that an interven- situations where knowledge distributed among tion focusing team members on the requirements of team members is so specialized that even with a their specific role, and another teammate’s spe- transactive memory system and plan formula- cific role allowed for greater flexibility and ef- tion behavior, members are not able to cover for ficiency in TMS processes and higher perfor- one another. Therefore, future research should mance when integrating a newcomer into the address how teams’ respond to the loss of a team. Although we did not directly test this idea member in different contexts and at more ad- as applicable to member loss, these findings vanced developmental stages. concerning reflection may generalize to mem- Second, because this study was conducted in ber loss situations. a laboratory context, future research needs to For many teams, it might be just as effective examine the external validity of these results. to eliminate the need for a well-developed TMS Although the nature of the team task did not altogether by shifting the structure of the team exactly mirror those of real organizational as- from one that is more functional or specialized signments, features of this task provided “mun- to one that is divisional, where members evi- dane realism” (Berkowitz & Donnerstein, dence a high degree of overlap in knowledge 1982). As well, participants were engaged in a This document is copyrighted by the American Psychological Association or one of its allied publishers. and skill. Researchers have shown that divi- psychologically absorbing task with the prom- This article is intended solely for the personal use of the individualsional user and is not to be disseminated broadly. structures can be particularly effective in ise of cash bonuses for top performing teams, unstable task environments (see Hollenbeck et this study also provided “psychological real- al., 2002), and teams are very amenable to shift- ism” (Berkowitz & Donnerstein, 1982). Hum- ing from functional to divisional structures phrey, Hollenbeck, Ilgen, and Moon (2004) where they can operate more autonomously noted that simulations such as the DDD “bridge within their environment (Johnson et al., 2006). the gap between field operations and university- However, although structural shifts may be fea- based theoretical research” (p. 201) and “allow sible for some types of teams (e.g., sales teams), for an increase in the level of mundane realism they may be problematic in others (e.g., auto- while increasing the level of experimental motive design teams). rigor” (p. 202). Further, given that we are inter- TRANSACTIVE MEMORY AND TEAM MEMBER LOSS 83

ested in teams’ initial reactions to member loss, formulation as defined in our study represents a we examined adaptation to member loss during qualitatively different adaptive mechanism than a short time period. Therefore, our results may the behaviors examined by DeRue et al. (2008). not generalize to the time period following the Perhaps the relationships differ with the removal initial response to member loss. of a team leader versus a team member, which is Third, we examined the effects of member a question that warrants investigation in future loss in relatively small teams. In larger teams research. the loss of a single member may be easier to Finally, researchers should examine other po- overcome, as the potential impact of a missing tential moderators for the relationship between member will necessitate less of a shift from transactive memory and team performance fol- existing, functional behaviors. Or, as the num- lowing member loss. For instance, teams with ber of members increases, the missing mem- more defined leaders may have a buffer against ber’s absence may be less noticeable, and other the detrimental effects of losing a more critical team members may fail to cover that person’s member. Formal team leaders may have more responsibilities. Future research on team adap- knowledge concerning the missing member’s tation to member loss in larger teams would role responsibilities and be more willing and allow for a better understanding of this issue. able to redistribute tasks among members, ame- Fourth, we conceptualized our measure of liorating the confusion resulting from the plan formulation as any communication be- change in role structure. Although there is cer- tween team members that involved specifying tainly room for future work in this area, this the order of importance of tasks or redistribut- research has provided an important step in ex- ing task duties among members regardless of panding the definition of team performance and the plan’s quality. It is possible that following illuminating a factor that benefits teams in the loss, team members may have developed a member loss situations. poor or unproductive plan for how to respond to the loss. However, Waller (1999) found that teams that engaged in any type of plan formu- References lation immediately following a nonroutine Aiken, L. S., & West, S. G. (1991). Multiple regres- event outperformed teams that failed to respond sion: Testing and interpreting interactions. New- to the nonroutine event and plan any type of bury Park, CA: Sage. action. In effect, the implementation of any plan Arrow, H., & McGrath, J. E. (1993). Membership led to higher performance than no plan at all. matters: How member change and continuity af- However, future work may find that teams with fect small group structure, process, and perfor- well-developed transactive memory systems mance. Small Group Research, 24, 334–361. doi: consistently develop high-quality plans com- 10.1177/1046496493243004 pared with teams with less well-developed sys- Austin, J. R. (2003). Transactive memory in organi- tems. In terms of our coding procedures, we zational groups: The effects of content, consensus, also coded team interactions live as they oc- specialization, and accuracy on group perfor- mance. Journal of Applied Psychology, 88, 866– curred, which we acknowledge is not consistent 878. doi:10.1037/0021-9010.88.5.866 with traditional communications coding proce- Balkundi, P., & Harrison, D. (2006). Ties, leaders, dures that use transcriptions (e.g., Hollingshead, and time in teams: Strong inference about network 1996). structure’s effects on team viability and perfor- This document is copyrighted by the American Psychological Association or one of its allied publishers. Future research should also examine forms of mance. Academy of Management Journal, 49, 49– This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. adaptive behavior other than plan formulation, 68. doi:10.5465/AMJ.2006.20785500 and how these behaviors are affected by the Barry, E., Kemerer, C., & Slaughter, S. (2006). En- criticality of a lost member. For example, De- vironmental volatility, development decisions, and Rue et al. (2008) examined two different types software volatility: A longitudinal analysis. Man- of adaptive behaviors—quantitative and quali- agement Science, 52, 448–464. doi:10.1287/mnsc .1050.0463 tative. They found that teams where hierarchy was Baumann, M. R. (2001). The effects of manipulating eliminated by the removal of a team leader— salience of expertise and membership change on a “shock” to the system—engaged in both types transactive memory. Unpublished doctoral disser- of behaviors. These findings contrast with the re- tation, University of Illinois at Urbana-Cham- sults found in our study. It is possible that plan paign. 84 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

Berkowitz, L., & Donnerstein, E. (1982). External learning: Collectively connecting the dots. Journal validity is more than skin deep: Some answers to of Applied Psychology, 88, 821–835. doi:10.1037/ criticisms of laboratory experiments. American 0021-9010.88.5.821 Psychologist, 37, 245–257. doi:10.1037/0003- Ellis, A. P. J., Porter, C. O. L. H., & Wolverton, S. A. 066X.37.3.245 (2008). Learning to work together: An examina- Brass, D. (1984). Being in the right place: A struc- tion of transactive memory system development in tural analysis of individual influence in an organi- teams. In V. I. Sessa & M. London (Eds.), Work zation. Administrative Science Quarterly, 29, 518– group learning: Understanding, improving, and 539. doi:10.2307/2392937 assessing how groups learn in organizations (pp. Brass, D. (1985). Men’s and women’s networks: A 91–115). New York, NY: Taylor & Francis. study of interaction patterns and influence in an Fleiss, J. (1971). Measuring nominal scale agreement organization. Academy of Management Journal, among many raters. Psychological Bulletin, 76, 28, 327–343. doi:10.2307/256204 378–382. doi:10.1037/h0031619 Burke, C. S., Stagl, K., Salas, E., Pierce, L., & Gersick, C. J. G., & Hackman, J. (1990). Habitual Kendall, D. (2006). Understanding team adapta- routines in task-performing groups. Organiza- tion: A conceptual analysis and model. Journal of tional Behavior and Human Decision Processes, Applied Psychology, 91, 1189–1207. doi:10.1037/ 47, 65–97. doi:10.1016/0749-5978(90)90047-D 0021-9010.91.6.1189 Girvan, M., & Newman, M. (2002). Community Burt, R. S. (1992). Structural holes. Cambridge, MA: structure in social and biological networks. Pro- Harvard University Press. ceedings of the National Academy of Sciences of Burt, R. S. (1995). Structural holes: The social struc- the United States of America, 99, 7821–7826. doi: ture of competition. Cambridge, MA: Harvard 10.1073/pnas.122653799 University Press. Hinsz, V. B., Tindale, R., & Vollrath, D. (1997). The Cannon-Bowers, J. A., Salas, E., & Converse, S. A. emerging conceptualization of groups as informa- (1993). Shared mental models in expert decision tion processors. Psychological Bulletin, 121, 43– making teams. In N. J. Castellan Jr. (Ed.), Current 64. doi:10.1037/0033-2909.121.1.43 issues in individual and group decision making Hollenbeck, J. R., Moon, H., Ellis, A. P. J., West, B. J., (pp. 221–246). Hillsale, NJ: Erlbaum. Ilgen, D. R., Sheppard, L.,...Wagner, J. A. (2002). Chen, G., Thomas, B., & Wallace, J. C. (2005). A Structural contingency theory and individual differ- multilevel examination of the relationships among ences: Examination of external and internal person- training outcomes, mediating regulatory processes, team fit. Journal of Applied Psychology, 87, 599– and adaptive performance. Journal of Applied Psy- chology, 90, 827–840. doi:10.1037/0021-9010.90.5 606. doi:10.1037/0021-9010.87.3.599 .827 Hollingshead, A. B. (1996). Information suppression Cohen, S. G., & Bailey, D. E. (1997). What makes and status persistence in group decision making: teams work: Group effectiveness research from The effects of communication media. Human the shop floor to the executive suite. Journal of Communication Research, 23, 193–219. doi: Management, 23, 239–290. doi:10.1177/ 10.1111/j.1468-2958.1996.tb00392.x 014920639702300303 Hollingshead, A. B. (1998a). Communication, learn- DeRue, D. S., Hollenbeck, J. R., Johnson, M. D., ing, and retrieval in transactive memory systems. Ilgen, D. R., & Jundt, D. K. (2008). How different Journal of Experimental Social Psychology, 34, team downsizing approaches influence team-level 423–442. doi:10.1006/jesp.1998.1358 adaptation and performance. Academy of Manage- Hollingshead, A. B. (1998b). Distributed expertise ment Journal, 51, 182–196. doi:10.5465/AMJ and communication processes in groups. In E. .2008.30776769 Mannix, M. A. Neale, & D. H. Gruenfeld (Eds.), Ellis, A. P. J. (2006). System breakdown: The role of Research on managing groups and teams (Vol. 1, This document is copyrighted by the American Psychological Association or one of its allied publishers. mental models and transactive memory in the re- pp. 105–125). Greenwich, CT: JAI Press. This article is intended solely for the personal use of the individual user and is notlationship to be disseminated broadly. between acute stress and team perfor- Hollingshead, A. B. (1998c). Retrieval processes in mance. Academy of Management Journal, 49, transactive memory systems. Journal of Personal- 576–589. doi:10.5465/AMJ.2006.21794674 ity and Social Psychology, 74, 659–671. doi: Ellis, A. P. J., Bell, B. S., Ployhart, R. E., Hollen- 10.1037/0022-3514.74.3.659 beck, J. R., & Ilgen, D. R. (2005). An evaluation of Hollingshead, A. B. (2000). Perceptions of expertise generic teamwork skills training with action teams: and transactive memory in work relationships. Effects on cognitive and skill-based outcomes. Group Processes & Intergroup Relations, 3, 257– Personnel Psychology, 58, 641–672. doi:10.1111/ 267. doi:10.1177/1368430200033002 j.1744-6570.2005.00617.x Humphrey, S. E., Hollenbeck, J. R., Ilgen, D. R., & Ellis, A. P. J., Hollenbeck, J. R., Ilgen, D. R., Porter, Moon, H. (2004). The changing shape of large- C. O. L. H., West, B. J., & Moon, H. (2003). Team scale programs of research: MSU-DDD as an il- TRANSACTIVE MEMORY AND TEAM MEMBER LOSS 85

lustrative example. In S. Schiflett, L. Elliott, E. process and performance (pp. 153–176). Washing- Salas, & M. Coovert (Eds.), Scaled worlds: Devel- ton, DC: American Psychological Association. opment, validation and applications (pp. 200– doi:10.1037/10690-008 219). Burlington, VT: Ashgate. Levine, J. M., Choi, H. S., & Moreland, R. L. (2003). Humphrey, S. E., Morgeson, F. P., & Mannor, M. J. Newcomer innovation in work teams. In P. Paulus (2009). Developing a theory of the strategic core of & B. Nijstad (Eds.), Group creativity: Innovation teams: A role composition model of team perfor- through collaboration (pp. 202–224). New York, mance. Journal of Applied Psychology, 94, 48–61. NY: Oxford University Press. doi:10.1093/acprof: doi:10.1037/a0012997 oso/9780195147308.003.0010 Johnson, M. D., Hollenbeck, J. R., Humphrey, S. E., Lewis, K. (2003). Measuring transactive memory Ilgen, D. R., Jundt, D. K., & Meyer, C. J. (2006). systems in the field: Scale development and vali- Cutthroat cooperation: Asymmetrical adaptation to dation. Journal of Applied Psychology, 88, 587– changes in team reward structures. Academy of 604. doi:10.1037/0021-9010.88.4.587 Management Journal, 49, 103–119. doi:10.5465/ Lewis, K. (2004). Knowledge and performance in AMJ.2006.20785533 knowledge-worker teams: A longitudinal study of Kane, A. A., Argote, L., & Levine, J. M. (2005). transactive memory systems. Management Science, Knowledge transfer between groups via personnel 50, 1519–1533. doi:10.1287/mnsc.1040.0257 rotation: Effects of social identity and knowledge Lewis, K., Belliveau, M., Herndon, B., & Keller, J. quality. Organizational Behavior and Human De- (2007). Group cognition, membership change, and cision Processes, 96, 56–71. doi:10.1016/j.obhdp performance: Investigating the benefits and detri- .2004.09.002 ments of collective knowledge. Organizational Kozlowski, S. W. J., & Bell, B. S. (2003). Work Behavior and Human Decision Processes, 103, groups and teams in organizations. In W. Borman, 159–178. doi:10.1016/j.obhdp.2007.01.005 D. R. Ilgen, & R. Klimoski (Eds.), Handbook of Lewis, K., Lange, D., & Gillis, L. (2005). Transactive psychology (Vol. 12, pp. 333–375). Hoboken, NJ: memory systems, learning, and learning transfer. Wiley. doi:10.1002/0471264385.wei1214 Organization Science, 16, 581–598. doi:10.1287/ Kozlowski, S. W. J., Gully, S. M., Nason, E. R., & orsc.1050.0143 Smith, E. M. (1999). Developing adaptive teams: Liang, D., Moreland, R., & Argote, L. (1995). Group A theory of compilation and performance across versus individual training and group performance: levels and time. In D. Ilgen & E. Pulakos (Eds.), The mediating role of transactive memory. Per- The changing nature of work performance: Impli- sonality and Social Psychology Bulletin, 21, 384– cations for staffing, personnel actions, and devel- opment (pp. 240–292). San Francisco, CA: Jossey- 393. doi:10.1177/0146167295214009 Bass. Majchrzak, A., Jarvenpaa, S. L., & Hollingshead, Kozlowski, S. W. J., & Klein, K. J. (2000). A mul- A. B. (2007). Coordinating expertise among emer- tilevel approach to theory and research in organi- gent groups responding to disasters. Organization zations: Contextual, temporal, and emergent pro- Science, 18, 147–161. doi:10.1287/orsc.1060.0228 cesses. In K. J. Klein & S. W. J. Kozlowski (Eds.), Marks, M. A., Zaccaro, S. J., & Mathieu, J. E. (2000). Multilevel theory, research, and methods in orga- Performance implications of leader briefings and nizations: Foundations, extensions, and new direc- team-interaction training for team adaptation to novel tions (pp. 3–90). San Francisco, CA: Jossey-Bass. environments. Journal of Applied Psychology, 85, Landis, J. R., & Koch, G. (1977). The measurement 971–986. doi:10.1037/0021-9010.85.6.971 of observer agreement for categorical data. Bio- Mathieu, J., Maynard, M., Rapp, T., & Gilson, L. metrics, 33, 159–174. doi:10.2307/2529310 (2008). Team effectiveness 1997–2007: A review LePine, J. A. (2003). Team adaptation and of recent advancements and a glimpse into the postchange performance: Effects of team compo- future. Journal of Management, 34, 410–476. doi: This document is copyrighted by the American Psychological Association or one of its allied publishers. sition in terms of members’ cognitive ability and 10.1177/0149206308316061 This article is intended solely for the personal use of the individual user and is notpersonality. to be disseminated broadly. Journal of Applied Psychology, 88, Miller, D. L., Young, P., Kleinman, D., & Serfaty, D. 27–39. doi:10.1037/0021-9010.88.1.27 (1998). Distributed dynamic decision-making sim- LePine, J. A. (2005). Adaptation of teams in response ulation: Phase I. Release notes and user’s manual. to unforeseen change: Effects of goal difficulty and Woburn, MA: Aptima. team composition in terms of cognitive ability and Moreland, R. L. (1999). Transactive memory: Learn- goal orientation. Journal of Applied Psychology, ing who knows what in work groups and organi- 90, 1153–1167. doi:10.1037/0021-9010.90.6.1153 zations. In L. Thompson, D. Messick, & J. M. Levine, J. M., & Choi, H. S. (2004). Impact of Levine (Eds.), Shared cognition in organzations: personnel turnover on team performance and cog- The management of knowledge (pp. 3–31). Mah- nition. In E. Salas & S. M. Fiore (Eds.), Team wah, NJ: Erlbaum. cognition: Understanding the factors that drive Moreland, R. L., & Argote, L. (2003). Transactive 86 SIEGEL CHRISTIAN, PEARSALL, CHRISTIAN, AND ELLIS

memory in dynamic organizations. In R. Peterson & Smith-Jentsch, K. A., Mathieu, J., & Kraiger, K. E. Mannix (Eds.), Understanding the dynamic or- (2005). Investigating linear and interactive effects ganization (pp. 135–162). Mahwah, NJ: Erlbaum. of shared mental models on safety and efficiency Moreland, R. L., Argote, L., & Krishnan, R. (1996). in a field setting. Journal of Applied Psychology, Socially shared cognition at work: Transactive 90, 523–535. doi:10.1037/0021-9010.90.3.523 memory and group performance. In J. L. Nye & Stout, R., & Salas, E. (1993). The role of planning in A. M. Brower (Eds.), What’s so social about social coordinated team decision making: Implications cognition? Social cognition research in small for training. Paper presented at the Human Factors groups (pp. 57–84). Thousand Oaks, CA: Sage. and Ergonomics Society, Santa Monica, CA. Moreland, R. L., & Myaskovsky, L. (2000). Exploring Summers, J. K., Humphrey, S. E., & Ferris, G. R. the performance benefits of group training: Transac- (2012). Team member change, flux in coordina- tive memory or improved communication? Organi- tion, and performance: Effects of strategic core zational Behavior and Human Decision Processes, roles, information transfer, and cognitive ability. 82, 117–133. doi:10.1006/obhd.2000.2891 Academy of Management Journal, 55, 314–338. Muller, D., Judd, C. M., & Yzerbyt, V. Y. (2005). doi:10.5465/amj.2010.0175 When moderation is mediated and mediation is mod- Sundstrom, E. (1999). Supporting work team effec- erated. Journal of Personality and Social Psychol- tiveness: Best management practices for fostering ogy, 89, 852–863. doi:10.1037/0022-3514.89.6.852 high performance. San Francisco, CA: Jossey- Palazzolo, E., Serb, D., She, Y., Su, C., & Contractor, Bass. N. (2006). Coevolution of communication and Volberda, H. (1996). Toward the flexible form: How knowledge networks in transactive memory sys- to remain vital in hypercompetitive environments. tems: Using computational models for theoretical Organization Science, 7, 359–374. doi:10.1287/ development. Communication Theory, 16, 223– orsc.7.4.359 250. doi:10.1111/j.1468-2885.2006.00269.x Waller, M. J. (1999). The timing of adaptive group Pearsall, M. J., & Ellis, A. P. J. (2006). The effects of responses to nonroutine events. Academy of Man- critical team member assertiveness on team perfor- agement Journal, 42, 127–137. doi:10.2307/ mance and satisfaction. Journal of Management, 32, 257088 575–594. doi:10.1177/0149206306289099 Wegner, D. M. (1987). Transactive memory: A con- Pearsall, M. J., Ellis, A. P. J., & Bell, B. S. (2010). temporary analysis of the group mind. In G. Mullen Building the infrastructure: The effects of role & G. Goethals (Eds.), Theories of group behavior identification behaviors on team cognition devel- (Vol. 185, pp. 185–208). New York, NY: Springer- opment and performance. Journal of Applied Psy- Verlag. doi:10.1007/978-1-4612-4634-3_9 chology, 95, 192–200. doi:10.1037/a0017781 Wegner, D. M. (1995). A computer network model of Preacher, K., & Hayes, A. (2004). SPSS and SAS human transactive memory. Social Cognition, 13, procedures for estimating indirect effects in simple 319–339. doi:10.1521/soco.1995.13.3.319 mediation models. Behavior Research Methods, Wegner, D. M., Erber, R., & Raymond, P. (1991). Instruments, & Computers, 36, 717–731. doi: Transactive memory in close relationships. Jour- 10.3758/BF03206553 nal of Personality and Social Psychology, 61, 923– Pulakos, E. D., Arad, S., Donovan, M. A., & Plam- 929. doi:10.1037/0022-3514.61.6.923 ondon, K. E. (2000). Adaptability in the work- Wegner, D. M., Giuliano, T., & Hertel, P. (1985). place: Development of a taxonomy of adaptive Cognitive interdependencies in close relationships. performance. Journal of Applied Psychology, 85, In W. J. Ickes (Ed.), Compatible and incompatible 612–624. doi:10.1037/0021-9010.85.4.612 relationships (pp. 253–276). New York, NY: Ren, Y., & Argote, L. (2011). Transactive memory Springer-Verlag. doi:10.1007/978-1-4612-5044- systems 1985–2010: An integrative framework of 9_12 key dimensions, antecedents, and consequences. Zajac, S., Gregory, M. E., Bedwell, W. L., Kramer, This document is copyrighted by the American Psychological Association or one of its allied publishers. The Academy of Management Annals, 5, 189–229. W. S., & Salas, E. (2013). The cognitive under- This article is intended solely for the personal use of the individual user and is notdoi:10.1080/19416520.2011.590300 to be disseminated broadly. pinnings of adaptive team performance in ill- Rosen, M. A., Bedwell, W. L., Wildman, J. L., Fritz- defined task situations: A closer look at team cog- sche, B. A., Salas, E., & Burke, C. S. (2011). nition. Organizational Psychology Review, 1–25. Managing adaptive performance in teams: Guiding principles and behavioral markers for measure- Received July 17, 2012 ment. Human Resource Management Review, 21, Revision received August 26, 2013 107–122. doi:10.1016/j.hrmr.2010.09.003 Accepted October 19, 2013 Ⅲ