Getting groups to develop good strategies: Effects of reflexivity interventions on team process, team performance, and shared mental models

Getting groups to develop good strategies: Effects of reflexivity interventions on team process, team performance, and shared mental models

Organizational Behavior and Human Decision Processes 102 (2007) 127–142 www.elsevier.com/locate/obhdp Getting groups to develop good strategies: Effec...

270KB Sizes 5 Downloads 16 Views

Organizational Behavior and Human Decision Processes 102 (2007) 127–142 www.elsevier.com/locate/obhdp

Getting groups to develop good strategies: Effects of reflexivity interventions on team process, team performance, and shared mental models Andrea Gurtner a

a,*

, Franziska Tschan

a,*

, Norbert K. Semmer b, Christof Na¨gele

b

Institute for Work and Organizational Psychology, University of Neuchaˆtel, Fbg. de l’Hoˆpital 106, 2000 Neuchaˆtel, Switzerland b University of Bern, Bern, Switzerland Received 8 June 2005 Available online 11 July 2006

Abstract This study examines the effect of guided reflection on team processes and performance, based on West’s (1996, 2000) concept of reflexivity. Communicating via e-mail, 49 hierarchically structured teams (one commander and two specialists) performed seven 15 min shifts of a simulated team-based military air-surveillance task (TAST) in two meetings, a week apart. At the beginning of the second meeting, teams were assigned either to a reflexivity (individual or group) or to a control condition. Results show that reflexivity enhanced performance, the link between reflexivity and team performance being mediated by communication and implementation of strategies as well as by similarity of mental models. Contrary to expectations, individual reflexivity was superior to group reflexivity. Additional analyses suggested that group reflexivity decreased the commanders’ active behavior and increased discussion of strategies that were too general to be helpful. Results point to the usefulness of reflexivity as a generic intervention but underscore the importance of focusing on strategies that are task-specific. Ó 2006 Elsevier Inc. All rights reserved. Keywords: Reflexivity; Strategy development; Coordination; Shared mental models; Team process; Team performance

Introduction Consider a team that has to take a decision by integrating, weighting, and combining information held by various team members to achieve an optimal assessment of a situation. An example are medical teams, where data about analyses, patient medical history, and current symptoms may be collected by different people and have to be combined and integrated in order to come up with the right diagnosis. Another example is the task used for the research reported in this paper: a team composed of a commander and two specialists *

Corresponding authors. Fax: +41 32 718 1391. E-mail addresses: [email protected] (A. Gurtner), franziska. [email protected] (F. Tschan). 0749-5978/$ - see front matter Ó 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.obhdp.2006.05.002

observes airplanes moving in a military environment. Each of the group members has access to different plane information, and only the correct combination of all the information allows assessment of the threat level of the aircraft (see Choi & Levine, 2004). In the examples cited, the team is characterized by information asymmetry (Stasser & Stewart, 1992; Stasser & Titus, 1985; Wittenbaum, Hubbell, & Zuckerman, 1999), where team members hold different, but crucial information about a problem; the team is hierarchically structured (Sniezek, 1992), and the task is complex. For complex tasks, performance not only depends on whether or not the group members actually share the information (Stasser & Titus, 1985), but it also depends on the extent to which the group is able to develop and implement good strategies (Hackman, 2002; Marks, Mathieu, & Zaccaro,

128

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

2001; Salas, Sims, & Burke, 2005). This process of strategy development and implementation may also be supported by external interventions (Hackman & Wageman, 2005). In this paper, we examine the effects of an intervention based on the work by West (1996, 2000) aimed at guiding teams to reflect on their strategies. We test the effect of a reflexivity intervention on the team process and on team performance in hierarchically structured, computer supported teams working on a simulated team-based military air-surveillance task (TAST). Guided reflexivity Reflexivity and team performance West defined reflexivity as ‘‘the extent to which group members overtly reflect upon the group’s objectives, strategies, and processes and adapt them to current or anticipated endogenous or environmental circumstances’’ (West, 1996, p. 559). Reflexivity in groups compares to reflective thinking and meta-cognition in the learning and performance process of individuals (Ericsson & Lehmann, 1996; Scho¨n, 1983), emphasizing that reflecting on one’s strategies allows for adaptation of strategies and thus is an important basis for high performance in complex tasks. Empirical support for the relationship between high reflexivity levels in groups and performance is provided in several studies. For example, highly reflective BBCTV production teams were assessed by their superiors as performing better (Carter & West, 1998). The link between reflexivity and group performance has been replicated in other studies (Schippers, Den Hartog, Koopman, & Wienk, 2003; Tjosvold, Hui, & Yu, 2003). Higher levels of reflexivity in groups have also been found to be related to higher innovation (De Dreu, 2002; Gevers, van Eerde, & Rutte, 2001; Tjosvold, Tang, & West, 2004) and to higher organizational citizenship behavior (Tjosvold et al., 2003). Reflexivity intervention In the studies cited above, reflexivity was measured as a team’s general working style. In the current study, we used an intervention to induce reflection in groups (guided reflexivity) and we expect that it will help groups to develop more adaptive strategies, and thus foster group performance. The intervention used is based on the three-stage process of reflexivity described by West (2000). We asked teams to reflect on how they have performed so far, to consider potential improvements, and to develop plans how the new strategies should be implemented. Finally, the implementation of the new strategies refers to action or adaptation. Why should reflexivity intervention in groups be necessary and useful? If lay people are asked about the factors influencing team performance, they rate strategy discussion as a critical factor, second only to effort

(Peterson, 1992). This is an indication that the importance of strategy development is widely accepted (Guzzo, Wagner, Maguire, Herr, & Lawley, 1986; Hackman, 2002). Consequently one could expect team members to reflect spontaneously and regularly on their strategies. However, this is often not the case. In a study on the impact of planning for team performance, Hackman, Brousseau, and Weiss (1976) had to force teams to engage in planning, and Weingart (1992) observed that teams faced with a complex task planned less than teams working on a simple task, although it is the complex tasks that require planning. An explanation for the reluctance to deliberately engage in strategy discussion in teams may be found in what has been termed the production paradox (Carroll & Rosson, 1987). Especially under time pressure, people want to get results (produce), not necessarily to learn (Karau & Kelly, 1992). This often implies that they concentrate on what is immediately necessary for production, possibly at the cost of achieving their goals by less than optimal means and of foregoing opportunities to develop a more efficient and elegant way of performing. Another reason a reflexivity intervention may be promising lies in the observation that groups and teams often adopt their strategies early on in the working process and do not spontaneously change or revise them (Argote, 1989; Bettenhausen & Murningham, 1985; Hackman & Wageman, 2005). Hackman and Wageman (2005) therefore suggest that discussion of strategies should be repeated after the group has acquired some experience with the task. Guided reflection may help groups to maintain a higher level of strategy discussion after the phase of initial strategy development and therefore may help groups to profit more from already acquired experience. If teams have difficulties developing task adaptive strategies and rethinking and revising strategies early adopted, interventions to help them overcome these shortcomings are important to foster performance. Teams can be trained in using optimal strategies for a specific task, for example, through lectures, demonstrations and hands-on practice (Cannon-Bowers, Tannenbaum, Salas, & Volpe, 1995). However, such training relates to a given task. This requires extensive preparation with regard to specific training content. Furthermore, it will ensure strategy development with regard to new tasks only to a limited extent. If, therefore, an intervention can be found that induces teams to develop and implement task-adaptive strategies on their own, time and effort could be saved and flexibility of task performance could be increased. Guided reflexivity could be such an intervention. Individual or group reflection Reflection can be conceived as a group discussion, but can also be done individually by each group

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

member. The advantage of a group discussion could be that each group member can contribute ideas and that the group can agree on new strategies and already plan their implementation during the reflexivity discussion. This requires, however, that the group avoids the tendency to focus on common information and to neglect expert information, a tendency often found in teams with distributed expertise (Stasser & Titus, 1985). To the extent that such a bias is operating, individual reflection may have an advantage in that group members with special expertise develop more concrete and better plans. In the research presented here, the team is hierarchically structured and, as often, the team leader has a unique status and holds crucial information (ThomasHunt, Ogden, & Neale, 2003). Therefore, the ability of the team leader to identify suboptimal group processes (Moreland & Levine, 1992), to develop strategies and to foster team coordination will be particularly important (Kozlowski, Gully, Salas, & Cannon-Bowers, 1996; Marks, Zaccaro, & Mathieu, 2000). For such a team structure, individual reflection could be especially useful, as it gives the group leader time to plan his or her interventions. On the other hand, if reflection is done as a group, the leader can learn more about the difficulties the subordinates have, and he or she has more time to communicate strategies already during the reflection period. We expect that both, individual and group reflection positively influence group performance. However, we expect reflection as a group to be superior to individual reflection, because the instruction for reflexivity is designed to focus the group on the decisive issues, thus avoiding the bias of discussing shared information only. Furthermore, in group reflexivity specific insights can be communicated directly and immediately, whereas any insights developed individually had to be communicated later, during a regular shift. This has costs in terms of time and attention. Mediating factors between a reflexivity intervention and performance: Strategy development, strategy implementation, and shared mental models West (1996, 2000) argues that groups with higher reflexivity more often critically revise their working processes, and therefore, are able to develop and implement better task-related strategies. However, to our knowledge, this contention has not yet been assessed empirically. West’s concept of the relationship between group process factors and group performance is in accordance with many models of group performance that emphasize the importance of strategy development and strategy implementation for high group performance (Gladstein, 1984; Hackman, 2002; Tschan, Semmer, Na¨gele, & Gurtner, 2000). Recent theories also emphasize the importance of shared cognition or shared mental models for high team performance.

129

In this paper, we draw on models that focus on the task-related aspects of the group process and their relationship to performance in an input–process–output framework (Espinosa, Lerch, & Kraut, 2004; Marks et al., 2000, Marks, Sabella, Burke, & Zaccaro, 2002; Mathieu, Heffner, Goodwin, Cannon-Bowers, & Salas, 2005; Mathieu, Heffner, Goodwin, Salas, & CannonBowers, 2000; Stout, Cannon-Bowers, Salas, & Milanovich, 1999). We tested the influence of the reflexivity intervention not only on performance, but also on aspects of the group process: on strategy development, on strategy implementation and on the similarity of the mental models. We expect that the link between reflexivity and team performance is mediated by these process factors. Each of these process factors is shortly discussed below. Strategy development Strategy development in groups has been described as the selection, organization, and adaptation of actions as well as the development of alternative courses of action adapted to changing task requirements (Marks et al., 2001; Salas et al., 2005; Tschan et al., 2000). Several studies found strategy development a critical factor for group performance, for example, for a business simulation task (Fussell et al., 1998); but also in cockpit crews (Orasanu, 1994), or in hospital teams (Edmondson, Bohmer, & Pisano, 2001). Although strategy discussion may not be fruitful for groups with easy or routine tasks (Espinosa et al., 2004; Hackman et al., 1976), it is crucial for teams with novel and complex tasks and teams working under difficult conditions (Cannon-Bowers, Salas, Blickensderfer, & Bowers, 1998). Team leaders may play an especially important role in strategy development and communication (e.g., Sniezek, 1992), because initiating and maintaining strategy discussions can be seen as part of their role (Edmondson, Roberto, & Watkins, 2003; Kozlowski et al., 1996; Marks et al., 2000). We therefore concentrate on the commanders’ strategy communication. Strategy implementation Strategy development is a necessary, but not sufficient condition for high performance as even the best plans and most elaborate strategies only enhance performance if they are actually implemented. This is not automatic, as strategies may be too abstract to be implemented directly, or problems of synchronization and coordination may hinder or delay the implementation of a good strategy. Several authors—for example, Frese and Zapf (1994) for individual actors and Marks et al., 2001 for teams—distinguish between a mode of planning or strategy development and an action phase, where strategies are ‘put to work’. Although strategy implementation is not automatic, it is enhanced by a planning phase (Stout et al., 1999). We therefore expect that communication of strategies

130

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

fosters strategy implementation, but also that strategy implementation mediates between strategy communication and performance. In contrast to communicating strategies, which in our setting is a central role of the commander, it is the specialists’ task to implement the strategies, and we therefore focus on the specialists’ behavior with regard to implementation. Shared mental models Recent theoretical and empirical work has emphasized the importance of shared cognition, or shared mental models, for good coordination and high performance in teams (Cannon-Bowers, Salas, & Converse, 1993; Klimoski & Mohammed, 1994; Marks et al., 2002; Mathieu et al., 2000; Salas et al., 2005). Shared mental models have been described as ‘‘knowledge structures held by members of a team that enable them to form accurate explanations and expectations of the task’’ (Cannon-Bowers et al., 1993, p. 288). Several studies have found empirical support for a positive relationship between shared mental models and group performance (Marks et al., 2000, 2002; Mathieu et al., 2000; Stout et al., 1999). Researchers have argued that different types of mental models exist; the most commonly made distinction is between shared knowledge of the task (task-related mental model) and shared knowledge concerning roles and responsibilities of team members (team-related mental model; Cannon-Bowers et al., 1993; Mathieu et al., 2005). Recent studies have found that the team-related mental model seems to play an especially important role for group performance (Mathieu et al., 2000, 2005; Smith-Jentsch, Mathieu, & Kraiger, 2005). In accordance with these models, we expect that strategy discussion leads to a similar understanding of team members’ roles and responsibilities, interaction patterns and information flow, that is to a more similar team mental model (Cannon-Bowers et al., 1993; Mathieu et al., 2000), and we expect that a more similar mental model in groups is related to strategy implementation, and to group performance. We realize that the similarity of mental models is not equivalent to their quality. Provided, however, those models are not completely inadequate, agreeing on a common strategy may provide some advantage even if the content of the mental models is not optimal.

members. The task of the group was to assess the potential threat of each airplane, which could change during the session. This required continuous observation of all planes in the airspace, exchanging plane information among group members, and integrating the information concerning a given plane to assess its threat. The role of the commander included deciding on the threat level of the plane, but also instructing specialists on optimal performance strategies. Exchange of plane information as well as discussion of procedures and strategies was possible by an e-mail system. Three conditions were run—a control condition, a group reflexivity condition, where the group was instructed to commonly reflect on the group’s strategies, and an individual reflexivity condition, where each group member individually reflected on the group’s strategies. We expect reflexivity to be most useful after the group has gained some experience with the task (Hackman & Wageman, 2005). Therefore, the intervention occurred after the third shift (see Methods section for a more detailed description of the task and the experimental conditions). Based on the considerations above, we formulated the following hypotheses: Hypothesis 1. Reflexivity performance.

intervention

enhances

(1.1) Team performance in the reflexivity conditions is better than in the control condition. (1.2) Team performance in the group reflexivity condition is better than in the individual reflexivity condition. Hypothesis 2. Reflexivity intervention influences group processes. (2.1) Team leaders (commanders) more often communicate strategies and request the implementation of strategies (called commander strategy communication from now on) in the reflexivity conditions than in the control condition. (2.2) Team members (specialists) implement more strategies (called specialist strategy implementation from now on) in the reflexivity conditions than in the control condition. (2.3) Team members develop more similar mental models in the reflexivity conditions than in the control condition.

The present study and hypotheses In the present study, hierarchically structured groups of three worked on a computer-simulated, team-based military air-surveillance task (TAST): for six practice shifts and one test shift of 15 min each, a group of one commander and two specialists had to observe an airspace with planes entering and flying around. Information about each plane was distributed among the group

Hypothesis 3. We propose a process model describing the influence of the reflexivity intervention on the team process and on team performance. (3.1) We expect the link between reflexivity and team performance to be mediated by the commander’s strategy communication, the similarity of mental models and specialist strategy implementation.

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

(3.2) We also expect the link between commander strategy communication and performance to be mediated by the similarity of mental models and the strategy implementation of the specialists. Methods Participants Participants were recruited through the subject pool of the psychology department of a Swiss university, and through flyers at the university and local colleges (Gymnasiums). Each participant was paid CHF 75 (ca. US $60) for full participation; psychology students could obtain credits instead of payment. Ten teams had to be excluded due to network problems that led to loss of data, three teams were excluded because they were identified as multivariate outliers. Inspection of the communication in one team revealed that they had discussed the task between meetings, which was prohibited. In two other teams, team members did not work on the task seriously in the last shift 7, but were chatting about non task-related matters, and therefore, their performance in the last shift 7 did not correspond with previous task accomplishment.1 The data of 49 teams were included in the analyses. Teams were randomly assigned to conditions; there were 15 teams in the individual reflexivity condition, 17 teams in the group reflexivity condition, and 17 teams in the control condition. The 147 participants (97 women and 50 men) were assigned to teams of three, based on their scheduling preferences. Mean age was 24.5 years (SD = 7.5). Most participants were students at a university (76.2%) or college level (16.3%), the rest (7.5%) were employed. Only participants who did not know each other were assigned to the same team. Experimental task The task used for this study was a team-based military air-surveillance task (TAST) (Choi & Levine, 2004; Gabrys, 1994). Three team members worked in different rooms on computers connected via a network. They communicated exclusively through an e-mail system that allowed sending and receiving typed text messages. The groups had a hierarchical structure, with a commander and two specialists. Participants were randomly assigned to one of the roles. The groups’ task was to observe planes moving in an air space and to determine the threat level of each plane at each moment during seven 15 min work shifts.

1 Results including these three teams showed substantially the same pattern.

131

The both complex and dynamic task is similar to TIDE2 used by Hollenbeck, Ilgen and colleagues (see, e.g., Hollenbeck et al., 1995). However, in contrast to TIDE2, planes in this simulation move in the airspace, and therefore, have to be assessed repeatedly over the observation period. Up to four planes could be present during any one shift. Plane information and formula Nine items of plane information were available to the team, but distributed among team members. Only specialist A had access to information about altitude (feet), distance (distance from base in miles), corridor (being in, at the edge, or out of a prescribed corridor), and size of plane (small, medium, large). Only specialist B had access to information about speed (miles per hour), angle (ascent or descent; in degrees), direction (in degrees) and radar (weather, none, and weapon). Depending on the raw values of the characteristics, their danger potential was either low (1), medium (2), or high (3) for altitude, distance, size, speed, angle, and radar, and none (0), medium (1) or high (2) for corridor and direction. The specialists observed the raw values of plane and flight characteristics, but had information sheets allowing them to transform raw values into danger potential values. They could either send raw values or danger potential values to the commander. Only the commander had access to information about IFF (identify friend or foe: friend, civilian, and enemy plane), with a danger potential of 0, 1, and 2. Threat assessment It was the commanders’ task to calculate the threat for each plane. Only they knew about and were trained in the use of a formula that integrated the various items of plane information. In the formula, total threat was the result of three subtypes of threat. Position threat was calculated by adding the danger potential of altitude (1–3) and distance (1–3), and multiplying that sum by the value for corridor (0–2). Because of the multiplicative formula, a value of 0 for corridor yielded a result of 0 for position threat. Corridor, therefore, was the critical parameter for position threat. Maneuverability threat was calculated by adding the danger potential of speed (1–3) and angle (1–3), and multiplying that sum by direction (0–2). The maneuverability threat was 0 if the plane was flying away from the base, so here direction was the critical parameter. Plane type threat was calculated by adding the danger potential of size (1–3) and radar type (1–3). The values of the three threat subtypes had to bee added and finally multiplied by the danger potential of IFF (0–2). Due to this multiplication, the threat value of friends was always 0, regardless of the other plane characteristics, so IFF was the critical parameter for overall threat. The specialists sent information on plane characteristics to the commander via the e-mail system. The

132

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

commander calculated the threat level for each plane and then assigned it to the plane. The result of the threat assignment appeared on the screens of the specialists, represented by one (low threat) to seven (high threat) stars. As the airplanes moved in the airspace, plane parameter values changed over time, so the team had to monitor plane information continually and the commander had to update the threat assignment in a timely manner. Feedback on performance for each plane and an overall performance score were provided to everyone automatically after each shift (see below). Training Before the first shift, the team participated in a training session. A slide show on the task, plane characteristics and their danger potential was presented, followed by hands-on training on the computer. During training, team members were in the same room, working on different computers in their specific roles. Special emphasis was put on learning to use the message facilities and how to look up information. The feedback screen was explained. After team training the commander individually received special instructions on how to use the formula to calculate threat values and how to assign threat levels on the computer. Commanders were reminded that a multiplication by zero equals zero. Procedure Seven shifts of 15 min were completed by each team in two sessions of about two and a half hours each. The sessions were one week apart (three shifts on day one and four shifts on day two), in order to minimize fatigue. On arrival participants filled in an informed consent form, and the general procedure was explained. Then they were seated in front of a personal computer for the standardized training procedure. After training team members were led to three different rooms, each equipped with a computer connected by a network. The teams now worked for three training shifts of 15 min each. Team members were asked not to discuss the experiment between the first and second sessions. Manipulation checks showed that this request was respected by all but one group, which was excluded from data analysis. The second session began with reflexivity (or control) intervention (see below). After the intervention, the teams worked for three more shifts of 15 min, followed by the test shift (shift 7). Immediately after completing test shift 7, participants filled in the shared mental model questionnaire. Neither performance goals nor performance rewards were given to the teams. After task completion, the teams were thanked, debriefed, and paid. Data on the behavior of team members were collected through a computerized protocol that recorded all actions and communication exchanges (e.g., ‘‘specialist B looks up corridor of airplane ax155’’ or ‘‘specialist A reads message 41 from the commander’’), as well as

the time of occurrence. Message content was also recorded and later coded. Experimental manipulation Three different conditions were created: Individual reflexivity, group reflexivity, and a control condition. At the beginning of the second session, participants in the reflexivity conditions were instructed to reflect on the air traffic control task. Each participant received a sheet describing in three steps how to engage in reflection on teamwork and the task: (1) Step one suggested reviewing task performance on the first day: ‘‘How did you ask for information? How did you pass on information? How was the team organized?’’ (2) Step two instructed participants to consider potential improvements in performing the task: ‘‘Are there alternatives to your chosen task performance procedures, and if so, what are they?’’ (3) Step three asked participants to develop suggestions for task improvement for future work. The instructions did not suggest specific strategies for the task. In the group reflexivity condition, team members discussed the questions as a group using the e-mail system. In the individual reflexivity condition, participants were asked to reflect individually, to write down their reflections and send them to the ‘‘archive’’ using the e-mail system. In the control condition, the groups were asked to discuss, via e-mail ‘‘the conditions of professional success’’, a topic unrelated to the air traffic control task. Reflexivity and control interventions were implemented at the beginning of the second day, which is about the midpoint of the experimental group’s life. Based on Gersick’s group developmental model (1988; Hackman & Wageman, 2005), this is as an optimal point in time to reflect on strategies. Evidence from the communication protocols of the reflexivity sessions revealed that team members remembered their earlier performance in detail. Manipulation check To determine if the participants followed the instructions, two coders independently rated the e-mail communication or the e-mails sent to the archive. All groups and individuals in the reflexivity conditions were indeed following the instructions; they reviewed their task performance on the first day and developed alternative performance procedures for the second day. None of the groups in the control condition exchanged e-mails about the air-surveillance task. The coding was straightforward and both coders agreed completely. Measures Team performance The TAST-program automatically calculates an overall performance score for each plane based on a second by second comparison of the real threat level (as programmed) and the threat assignment of the team: the

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

smaller the difference, the higher the performance. These second by second differences are averaged over the time a plane is in the airspace. An adjustment in the formula takes into account that no team can assign the threat level of a plane correctly before gathering the essential information on a plane (tolerated time lapse of 20 s). The performance value is standardized between 0 and 100, a value of 100 indicating perfect agreement between the threat level of a plane and the threat assignment of a team. Finally, the overall performance of a shift is calculated as the average performance per plane in a shift. Performance of the test shift (shift 7) was used as a dependent variable. Since initial differences in performance might be important, performance in shift 3 (the last shift before experimental intervention) was used as a control variable. For all analyses, the team performance measure was arcsine-transformed, as recommended by Fleiss (1981) for proportional data. However, we report percentages in descriptive statistics. Commander strategy communication Based on hierarchical task analysis we identified four main task adaptive strategies the commanders could propose to the specialists in order to enhance coordination among team members (Tschan et al., 2000). If those strategies were implemented by the specialists, this would lead to a considerable reduction in the workload of the experts and the commander. Only the commander could suggest the four strategies, as they all are related to the features of the formula as follows: (1) The commander could ask the specialists to preprocess information. Receiving threat values instead of raw values (e.g., 2, signifying medium threat for distance rather than the distance in feet, which then the commander would have to translate into the threat level), would reduce the commander’s workload and the risk of information overflow (Fussell et al., 1998). (2) Each of the specialists had access to plane information necessary to calculate a partial threat value (either position threat or manoeuvrability threat). The optimal strategy for distributing the workload would be to teach each specialist how to calculate their part of the formula. This would allow the specialists to send precalculated partial threats instead of information on each parameter, and this would greatly reduce the workload of the commander. (3) Friendly planes represent no threat. Therefore, the commander (who was the only person to see the IFF status of the plane) could tell the specialists which planes are friendly and ask them to ignore such planes. This reduces the specialists’ and subsequently also the commander’s workload. (4) Corridor (specialist A) and direction (specialist B) are especially critical parameters to observe, according to the formula (see above). As only the commander knows which of the plane characteristics are critical, he or she should advise the specialists to observe those critical characteristics more closely.

133

To assess strategy communication, all messages were content coded. Strategy information in all shifts was coded as either general (such as ‘‘lets cooperate well’’ or ‘‘send all information’’), or as referring to one of the four task adaptive strategies described above. Ten percent of the material was coded by two independent raters who had excellent interrater agreement (Cohen’s j = .92). As the commander is the one who has all the strategic information, it is strategic communication by the commander that is of main interest. We therefore concentrated on the communication of task adaptive strategies (as opposed to general strategies) of the commander. We counted in how many shifts a specific task adaptive strategy was communicated by the commander to the specialists during shifts 4–6 (i.e., the first shift after experimental intervention to the last shift before the test shift, considered as the strategy development phase). In the group reflexivity condition, strategy communication during the reflection phase was also included. Strategy communication in test shift 7 was not included in this measure, since performance in that shift was used as the criterion (see below). Specialist strategy implementation Specialist strategy implementation for each of the four critical strategies was calculated as follows: (1) Preprocessing of information was operationalized as the proportion of all plane information sent to the commander as threat values (as compared to raw values). (2) To measure partial-threat calculation by the specialists, we calculated the proportion of plane information sent as partial-threat values (position or manoeuvrability). (3) To measure if specialists refrained from observing friendly planes, we calculated the proportion of plane information look-ups by specialists that referred to friendly planes. (4) To assess whether the critical parameters corridor or direction were observed more often, we calculated the proportion of plane information look-ups of these parameters on all parameter look-ups. The computerized protocol of specialist behavior allowed directly extracting the information for all four measures. All four proportions were calculated relative to the total parameter information that was looked up or sent. After reversing the score for friendly planes and arcsine-transforming all proportions (Fleiss, 1981), the four scores were z-standardized and combined into one score representing strategy implementation. Strategy implementation in test shift 7 is considered one of the criterion variables. Shared mental model (SMM) As we are concerned with communication on task adaptive behaviors and coordination, the team-interaction model is the subcategory of mental models of interest to us. It was assessed by asking each participant to rate 25 concepts on a 9-point Likert scale in terms of

134

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

their importance for attaining high performance in the task, ranging from 1 (not important) to 9 (very important). In accordance with the hierarchical structure of the group and the pivotal role of the commander, the concepts covered primarily the role of the commander in coordinating the team. Example items are: ‘‘The commander explains to the specialists which parameters are important and which are not’’, ‘‘The commander is clear about what information he or she expects’’. The similarity of the mental models in the team was calculated in two steps. First, the individual mental model was calculated using the Pathfinder algorithm (Cooke et al., 2003). The Pathfinder algorithm is based on mathematical graph theory and was developed to graphically represent structures of mental models, using proximity (or similarity) between concepts, expressed as the difference in the rating of each concept. We calculated a proximity matrix based on the ratings of the concepts for each participant and submitted this matrix to the Pathfinder algorithm to derive the individual mental models, using KNOT software (Knowledge Network Organizing Tool; Schvaneveldt, 1990). In the second step, the similarity of the individual mental models was calculated. The Pathfinder C (closeness) coefficient reflects the overlap between two mental models on a scale from 0 (no similarity) to 1 (perfect similarity). Similarity measures of each two-by-two combination of networks were averaged to obtain the similarity for the three group members. This procedure is similar to that used by Stout et al. (1999).

Results Table 1 presents descriptive statistics and correlations between all study variables. Effects of reflexivity on team performance. Hypothesis 1 stated that the reflexivity intervention would positively influence team performance (1.1) and that this effect

would be most pronounced for the group reflexivity condition (1.2). Hypothesis 1 was tested with ANCOVA. Performance in shift 3 (the last shift before reflexivity intervention) was entered as a covariate, to control for effects of earlier performance. The dependent variable was performance in test shift 7. In the group reflexivity condition, average team performance was 77.8% (SD = 6.8), n = 17; in the individual reflexivity condition it was 81.0% (SD = 4.8), n = 15, and in the control condition teams performed at 76.6% (SD = 5.5), n = 17, F (2, 45) = 3.43, p < .05, g2 = .13. Hypothesis 1.1 tested the two reflexivity conditions against the control condition. Hypothesis 1.2 tested group reflexivity against individual reflexivity. Planned contrasts revealed a significantly higher performance in the individual reflexivity condition than in the control condition (p = .013). However, performance in the group reflexivity condition was significantly different neither from performance in the control condition (p = .309) nor in the individual reflexivity condition (p = .114). If the two reflexivity conditions were combined, yielding a value of 79.3% (SD = 6.10), the reflexivity conditions were significantly different from the control condition, F (1, 46) = 4.13, p < .05, g2 = .08. Hypothesis 1.1 was therefore supported but Hypothesis 1.2 was not. Effects of reflexivity on group processes. Hypothesis 2 tested the effects of the two reflexivity interventions on group process variables with ANCOVAs, again entering performance of shift 3 as a covariate. We expected more commander strategy communication in the two reflexivity conditions than in the control condition (2.1). Commanders communicated a mean of 9.00 (SD = 5.72) strategies in the individual reflexivity condition, a mean of 10.71 (SD = 7.60) strategies in the group reflexivity condition and a mean of 3.88 (SD = 3.31) in the control condition, F (2, 45) = 6.67, p < .01, g2 = .23. Planned contrasts revealed significantly more commander strategy communication in the individual reflexivity condition (p = .014) and in the group reflexivity condition

Table 1 Means, standard deviations and intercorrelations among all study variables Variable

M

SD

1

2

3

4

5

1. Reflexivity intervention

1 = control 2 = reflexivity 67.76 78.38 7.82 .00 .33

10.54 5.98 6.43 .49 .07

.069 .222 .451** .386** .476**

— .464** .132 .340* .003

— .533** .603** .302*

— .596** .561**

— .526**

2. 3. 4. 5. 6.

Performance shift 3a Performance test shift 7a Strategy communication commander shifts 4–6ad Strategy implementation specialists test shift 7b Shared mental model shift 7c

Note: N = 49 groups. a Correlations are based on the arcsine transformed variables. b Arcsine transformed and z-standardized. c Based on pathfinder c coefficient. d In group reflexivity also communication during the reflexivity shift is included. * p < .05. ** p < .01.

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

(p = .001) than in the control condition, but no differences between the two reflexivity conditions. We also expected more specialist strategy implementation by the specialists in the two reflexivity conditions than in the control condition (2.2). Specialists implemented more strategies in the individual (M = .14, SD = .42) and in the group reflexivity condition (M = .14, SD = .46) than in the control condition (M = .26, SD = .49); note that strategy implementation is a z-standardized variable. The main effect was significant, with F (2, 45) = 5.28, p < .01, g2 = .19. Planned contrasts revealed significantly more specialist strategy implementation in the individual (p = .009) and the group reflexivity condition (p = .006) than in the control condition, but no differences between the two reflexivity conditions. Finally, as postulated in Hypothesis 2.3, the team interaction mental models in the individual (M = .35, SD = .07) and in the group reflexivity condition (M = .36, SD = .08) were more similar than mental models in the control condition (M = .28, SD = .04). The main effect was significant with F (2, 45) = 6.72, p < .01, g2 = .23, and planned contrasts revealed significantly more similar mental models in the individual reflexivity condition (p = .006) and in the group reflexivity condition (p = .002) than in the control condition, but no differences between the two reflexivity conditions, thus supporting Hypothesis 2.3. We proposed a model describing the influence of reflexivity and group process variables on team performance (Hypothesis 3). Given that we found no differences in the group process variables between the two reflexivity conditions, we pooled them for the following analyses. We expected the link between reflexivity and team performance to be mediated by three process variables: strategy communication by the commander, strategy implementation by the specialists and similarity of team members’ mental models. This was tested as a path analysis in a series of multiple regressions, following the procedure for testing mediations proposed by Baron and Kenny (1986). Table 2 displays the results of the regression equations, the results are also summarized in Fig. 1. In essence, performance was regressed on all predictor variables, and in the following steps the more proximal predictors (those to the right of Fig. 1) were regressed on the more distant ones (those to the left of Fig. 1): Performance in test shift 7 was regressed on strategy implementation by the specialists in the same shift, shared mental models measured at the end of test shift 7, strategy communication by the commander in the shifts 4–6 (the shifts between reflexivity intervention and test shift 7), and a dummy variable representing reflexivity intervention. Performance in shift 3 was entered as a control variable. In a second regression, strategy implementation was regressed onto shared mental models, strategy communication, reflexivity interven-

135

tion and performance in shift 3. In the third regression, shared mental models were regressed onto strategy communication, reflexivity intervention and performance in shift 3. Finally, in the fourth regression, strategy communication was regressed onto reflexivity intervention and performance in shift 3. The path analysis showed that the relationship between reflexivity and performance in test shift 7 was indeed mediated by commander strategy communication, shared mental models and specialist strategy implementation (3.1). The link between commander strategy communication and performance in the test shift 7 was partially mediated by shared mental models and specialist strategy implementation (3.2), Sobel-test, Z = 1.72, p < .05, one-tailed (Sobel, 1982). The link between reflexivity and specialist strategy implementation was mediated by commander strategy communication and shared mental models, Sobel-test, Z = 2.15, p < .05, one-tailed. Finally, the link between reflexivity and shared mental models was partially mediated by commander strategy communication, Sobel-test, Z = 2.05, p < .05, one-tailed. There were direct paths between reflexivity intervention and commander strategy communication and specialist strategy implementation and performance in test shift 7. These results support the suggested model to a considerable degree. Additional analyses. Contrary to our expectations, the results for the group reflexivity condition were not superior to the individual reflexivity condition. Quite the contrary, they were inferior. We therefore conducted some additional analyses to clarify what might be responsible for this aspect of our results.2 One analysis concerned the role of the commander in the two reflexivity conditions. Specifically, we reasoned that more active commanders (relative to their specialists) might improve performance and that the individual condition might enhance active behaviors by the commanders. We therefore coded all messages with regard to their quality as initiation or response. Initiation consisted of command, observation, positive feedback, negative feedback, and questions. Response consisted of acceptance, rejections and expression of uncertainty. Active leadership of the commander was then calculated as the (arcsine-transformed) proportion of initiating communication acts of the commander relative to all communication acts of all three team members. As before, this refers to shifts 4–6, and as before, we also included active leadership during the reflection shift in the group reflexivity condition. Twenty-five percent of the material was double coded, Cohen’s j was .92 for initiation and .94 for response, indicating excellent reliability.

2 We thank two anonymous reviewers and the editor for suggesting such analyses.

136

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

Table 2 Hierarchical multiple regressions for the path model Variable Regression 1: dependent variable is performance in test shift 7 Step 1 Performance shift 3 Reflexivity (dummy variable) Step 2 Performance shift 3 Reflexivity (dummy variable) Strategy communication commander shifts 4–6 Step 3 Performance shift 3 Reflexivity (dummy variable) Strategy communication commander shift 4–6 Strategy implementation specialists test shift 7 Shared mental model test shift 7

B

SEB

.344 .052 .290 .009 .007 .219 .002 .005 .067 .068

.090 .025 .081 .025 .002 .085 .026 .002 .030 .187

1.044 .197 .036

**

.481 .255*

.163**

.442**

.057

.406 .042 .461** *

.306 .009 .326* .336* .052

.252** **

.368 .411** .418**

.171**

.456**

.047*

*

.291 .194 .471**

.391 .128 .010 .889

**

.302 .118 .349* .275*

Regression 3: dependent variable is shared mental models in test shift 7 Step 1 Performance shift 3 .016 Reflexivity (dummy variable) .074

.071 .020

.030 .478**

Step 2 Performance shift 3 Reflexivity (dummy variable) Strategy communication commander shift 4–6

.065 .021 .002

.043 .274* .443**

1.087 .120 .027 1.814

.406** **

.404 .127 .010

Step 3 Performance shift 3 Reflexivity (dummy variable) Strategy communication commander shift 4–6 Shared mental model test shift 7

DR2

.249**

Regression 2: dependent variable is strategy implementation of the specialists in test shift 7 Step 1 Performance shift 3 1.322 .450 Reflexivity (dummy variable) .419 .127 Step 2 Performance shift 3 Reflexivity (dummy variable) Strategy communication commander shift 4–6

R2adj

b

.194**

.337** .023 .042 .005

Regression 4: dependent variable is strategy communication of the commander in shifts 4–6 Performance shift 3 7.737 6.117 Reflexivity (dummy variable) 6.174 1.733

.164 .462**

.151**

.196**

N = 49 groups. * p < .05. ** p < .01.

The analyses revealed that commanders, indeed, showed more active leadership in the individual (M = .82, SD = .11) than in the group reflexivity condition (M = .50, SD = .13), t (25.65) = 6.83, p < .001. To test the effect of team leader activity on performance effects in the two reflexivity conditions, we conducted a regression analyses with team performance as the dependent variable, and commander activity as predictor. Performance in shift 3 was, again, entered as a control variable. This analysis yielded a significant main effect with R2adj ¼ :28, F (2, 29) = 6.93, p < .01, and bcommander activity = .318, p < .05. Thus, group reflex-

ivity seemed to suppress active leadership behavior as defined above. A second analysis turns away from the focus on the commander and includes the experts as well. As explained above, we had concentrated on the commander because only the commanders had the specific knowledge that suggested to them to develop the four adaptive strategies of this task. The experts might actually have contributed to the discussion in a different way, that is, by talking about general strategies, which may, not have been particularly useful for the task. However, this information asymmetry (Stasser & Titus, 1985) between

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

PERFORMANCE SHIFT3 (CONTROL)

.481**/.306* .368**/.302**

.411**/.118(ns) .471**/.349*

REFLEXIVITY INTERVENTION VS CONTROL

137

.462**

COMMANDER STRATEGY COMMUNICATION (SHIFTS 4-6)

.443* .478**/.274*

SPECIALIST STRATEGY IMPLEMENTATION TEST SHIFT (7)

.336*

GROUP PERFORMANCE TEST SHIFT (7)

.275*

SHARED MENTAL MODELS TEST SHIFT (7)

.461**/.326* .255*/-.009(ns)

Fig. 1. Mediators of the relationship between reflexivity interventions and performance: A path model. Note. Numbers represent standardized beta weights. Only relationships where p < .05 were retained. Numbers before the slash represent coefficients for step 1 (with performance in shift 3 and reflexivity as predictors), numbers after the slash represent coefficients of the final model (see Table 2). N = 49. R2adj ¼ :442 . *p < .05. **p < .01.

the team members may have misled teams in the group reflexivity condition to discuss more shared (general) strategies. In extending the analysis to the experts, and thus to the group as a whole, we therefore included task adaptive as well as general strategies, counting both the number of task-adaptive strategy communications and the number of general strategy communications appearing in all messages of either commander or specialists. We also calculated the proportion of task-adaptive strategy communications on all communications in the team. Indeed, the amount of team communication related to general strategies was much lower in the individual reflexivity condition (M = 8.87; SD = 6.23) than in the group reflexivity condition (M = 44.35; SD = 17.09), t (20.65) = 7.98, p < .001. Moreover, the proportion of team strategy communication related to task adaptive strategies on all strategy communication (task adaptive strategies and general strategies) was higher in the individual (M = .69, SD = .24) than in the group reflexivity condition (M = .45, SD = .16), t (30) = 3.567, p < .01, based on arcsine transformed proportions. To test the effect of the proportion of task adaptive strategy discussion on performance in the two reflexivity conditions, we conducted a regression analyses with team performance as the dependent variable, the arcsine transformed proportion of task adaptive strategy communication of the group as predictor, and performance in shift 3 as control variable. This analysis yielded a trend for the main effect with R2adj ¼ :27, F (2, 29) = 6.75, p < .01, and bproportion of task adaptive strategy discussion = .326, p = .055. Thus, it seems that the group reflexivity condition induced quite some discussion about general, rather than task-specific strategies, and these general strategy discussion added ‘noise’ rather than helpful communica-

tion, and therefore, lowered performance. This effect was mainly due to the behavior of the experts.

Discussion The purpose of this study was to test whether guided reflexivity interventions would lead to higher team performance. We encouraged hierarchically organized groups performing a simulated team-based military air-surveillance task to reflect on their task strategies either individually or in a group discussion. We expected the effect of reflexivity on performance to be mediated by more strategy communication, similarity of team interaction mental models, and more implementation of strategies. With the exception of the effects of individual vs. group reflexivity, our results largely support these assumptions. We will discuss each of the hypotheses below. Reflexivity and performance Previous research on reflexivity has related group performance to the habitual amount of reflexivity in groups, but not used reflexivity as a training tool (e.g., Carter & West, 1998; Schippers et al., 2003). We have shown that this concept can rather easily be transformed into an intervention tool. The postulated effect of reflexivity on performance was supported: Groups in the reflexivity conditions performed better after controlling for performance in earlier shifts. The latter is important as performance on the first day was in itself a good predictor of performance on the second day. Earlier research (Gersick & Hackman, 1990) has shown that groups often adopt their strategies early and do not

138

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

change their habitual routines. Our results suggest that retaining early suboptimal habits could be prevented, or at least attenuated, by the reflexivity intervention. Thus, our findings underscore the importance of continued strategy discussion even for experienced groups (Hackman & Wageman, 2005; Orasanu, 1994). Individual and group reflexivity We expected the effects of reflexivity on performance to be stronger in the group than in the individual reflexivity condition. We thought that in the individual reflexivity condition the extra time needed during task performance for communicating gained insights would hamper the exchange of strategic information. This was not the case. Performance of the two conditions was not significantly different and the pattern indicated better, not worse performance in the individual reflexivity condition. In fact, teams in the group reflexivity condition did not perform significantly better than groups in the control condition. In addition, group and individual reflexivity showed a very similar pattern for the measured process variables. In both reflexivity conditions, the commander communicated significantly more strategies than in the control condition, the experts implemented more strategies than in the control condition, and the similarity of the shared mental model was higher than in the control condition. All three process variables included into the model showed no differences between individual and group reflexivity. The explanation for the better performance in the individual vs. group reflexivity condition evidently does not lie in these process variables. Although pertinent analyses could only be carried out post hoc, trying to understand this unexpected result is important both for theory and application purposes. Following suggestions by anonymous reviewers, we therefore analyzed the behaviors by commanders and experts in a more fine-grained way. One possibility was that commanders in the individual reflexivity condition would take a more active leadership role later on. Only the commander was given specific information on how to combine plane characteristics to threat assignment. This made strategy development highly dependent on his or her task related knowledge. It may, therefore, be especially crucial that the commander has the opportunity to reflect, even without immediately exchanging ideas with the specialists (Kozlowski et al., 1996). Additional analyses revealed, indeed, that commanders in the individual reflexivity condition more actively took the leadership role than commanders in the group reflexivity condition. Furthermore, active leadership behavior of the commanders did, indeed, predict team performance. As these results are based on a post hoc analysis, further investigation of this process is warranted.

Another possibility refers to extending the analyses from the focus on the commanders’ communication to the experts, thus involving the team as a whole. As the focus on the commanders was based on their specific task-related knowledge, a change in the variable measured suggested itself as well. We had focused on communication that referred to those strategies that we had identified as promising by task analysis (Tschan et al., 2000). The possibility for the experts to contribute to these strategies was limited, basically being confined to asking questions that would induce the commander to share his or her strategic knowledge. The experts might well, however, contribute in a less helpful way, that is, by discussing general strategies (such as ‘‘let’s cooperate well’’) rather than task-adaptive strategies. To the extent that this would happen, it would introduce ‘noise’ into the discussion, implying coordination losses (Steiner, 1972) by distracting the group from careful elaboration of task-adaptive strategies. Indeed, the additional analyses showed that in the group reflexivity condition, the proportion of general strategies discussed was much higher than in the individual reflexivity condition. Moreover, the higher the proportion of general, as opposed to task-adaptive, strategies, the lower the performance of the group. This indicates that the participation of the specialists in the strategy discussion actually distracted from the focus on really helpful strategies. By contrast, individual reflection seemed to induce the commanders to focus on the helpful strategies, and this by far outweighed the disadvantages of individual reflexivity. Of course, these are post-hoc analyses, and they should be treated with caution until they are replicated. They are, however, in line with research on managers showing that individual reflection resulted in a learning advantage over reflection in peer groups, mainly because the peer groups did not discuss the managers’ problems in a sufficiently specific way (Daudelin, 1996). If confirmed by future research, these two additional analyses shed some light on potential disadvantages of reflective group discussions. Rather than increasing diversity and enhancing the chance that important aspects will be discussed, group discussions may actually restrict the topics that are reflected upon, as suggested by the research tradition on hidden profiles (Stasser & Titus, 1985). Furthermore, in our case of groups with clear information asymmetry, restricting the discussion to shared knowledge seems to imply a restriction to less specific, less task-adaptive strategies, as it is the general strategies, such as ‘‘let’s cooperate well’’ that are common knowledge. (This implies, of course, that these results cannot be generalized to groups where everybody is an expert, even if they are replicated for groups like ours.) Such general discussions may even discourage the commander to take a more active role. To the extent that this is true, individual reflexivity should be preferred. As group discussions may have other advanta-

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

ges, however, such as increasing commitment to decisions taken, one might also consider a more specific instruction (e.g., by explaining the difference between general and task-adaptive strategies), or an instruction that emphasizes the commander’s status and encourages the experts to elicit his or her knowledge. To the extent that this aim is reached, more specific knowledge should be exchanged, and the active role of the commander should be encouraged. Again, however, these considerations are based on post-hoc analyses, which need to be treated with caution unless replicated. The path model We tested a model relating reflexivity to strategy communication, mental model, strategy implementation and performance. To the best of our knowledge, no empirical investigation of the effects of reflexivity on group process variables has been carried out before. The links proposed in the path model were largely supported. The commanders in the reflexivity conditions communicated task-adaptive strategies in more than twice as many shifts as the commanders in the control condition, and the specialists in the reflexivity conditions implemented more strategies than the specialists in the control condition. The effect of commander strategy communication on performance was only partially mediated by specialist strategy implementation, a direct effect remains. This could be due to the fact that commanders who suggest more strategies to their group also apply better individual strategies for using information received by the specialists. Again, further research is needed on this issue. Reflexivity enhanced the similarity of the team interaction mental model. In recent years there has been increasing evidence that supports the link between team process, shared mental models, and performance (Marks et al., 2002; Mathieu et al., 2000; Stout et al., 1999), and it has been shown that interventions in teams influenced shared mental models (Smith-Jentsch, Campbell, Milanovich, & Reynolds, 2001). Our results parallel those of these studies, as reflexivity intervention did enhance the similarity of the team members’ mental models. This effect was partially mediated by commander strategy communication, suggesting that the commander did contribute to an important extent to the shaping of the mental models of the team members (Marks et al., 2000). Moreover, the reflexivity intervention also had a direct effect on shared mental models. The link between mental model and group performance was fully mediated by strategy implementation. These are important findings, since evidence that strategy development and planning can influence the similarity of mental models among team members is still sparse. We argued above that the similarity of the mental models—although not equivalent quality—may still yield an advantage, as long as those models are not com-

139

pletely inadequate. Our results support this contention. Furthermore, we concentrated on the team interaction aspects of the mental model. Simple agreement on a given interaction strategy is likely to carry some advantages in terms of coordinated activity regardless of content, as long as this content is not completely wrong. Things might well be different for the task-related mental models. Note, however, that mental models were measured immediately after task completion, but strategy implementation was measured during task performance. This procedure was chosen in order to not influence strategy building with specific questions while the group was still performing the task. We do, however, assume that the shaping of the mental models took place during the whole seven shifts and that there is a mutual influence of the mental models and the group process. More refined studies are needed, however, to investigate the mutual influence of process variables and shared mental models over time. Regarding the model as a whole, our additional analyses suggest one important modification. The part of the model that refers to strategy communication should focus: (a) not only on the commander but on the group as a whole and (b) include communication about both, task adaptive and general, strategies, as a high proportion of general strategies actually seems to lower performance. Strengths and limitations That we measured mental models after shift 7 certainly is a weak point of our study, as mentioned above. The most important limitations of the present research, however, concern the generalizability of the results. As in all experimental studies, it has to be shown to what degree the effects and relationships discovered apply to other groups and teams. Generalizations from one type of group and across different types of tasks have to be made with caution (Arrow, McGrath, & Berdahl, 2000; McGrath, 1984). This also applies to the present study where teams were hierarchically structured, the task was dynamic and the communication was computer-supported. Further research exploring the effects of reflexivity interventions in different types of tasks and in other settings will be needed. Another limitation of the study is the relatively small number of groups included. This necessitates caution about drawing premature conclusions, and underscores the need for replication studies. Despite these limitations, we feel that this study contributes to the understanding of intervention procedures in groups concerning the development and implementation of strategies and the shaping of shared mental models in groups. Analyzing these processes helps to understand the effects of interventions such as guided

140

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

reflection, and they can help to improve training methods for optimal team performance. More specifically, the study shows that a very general training procedure—encouraging groups or individuals working in groups to reflect on what they have done so far, and how they think they could improve their strategies in the future—has clear effects on group processes and performance. This is important, as guided reflection is a method that can be applied rather easily for many different teams in many different situations and does not require extended task-analysis and specific training procedures. This is not at all an argument against more task-specific team training methods (e.g., the so called self-correction or team dimensional training; see Smith-Jentsch, Zeisig, Acton, & McPherson, 1998), which are likely to have greater effects on group performance. However, for many groups and tasks, developing task-specific training procedures is very costly, whereas reflexivity is inexpensive. Furthermore, we were able to show where and how reflexivity actually influences the group process as we assessed the effect of reflexivity on commander communication and group member behavior at a later time. This allows for important insights into the group process. In fact, Mathieu et al. (2005) have very recently stressed that more detailed analyses of the relationship between reflective periods of the group’s process, action periods and shared mental models are needed. Finally, this study contributes to the ongoing discussion about the importance of mental models and shared cognition in groups—specifically their relationship to aspects of the group process, but also to performance. The role of a discussion, or reflection, on strategies in a very specific and focused way, that is, evaluating one’s behavior and performance with regard to detecting, and improving, task-adaptive strategies, for improving mental models, implementing good strategies, and improving performance, gains considerable support from our results.

Acknowledgments This research was supported by the Swiss National Science Foundation, Contract 1114-056997.99. We wish to thank Joseph E. McGrath, Richard E. Moreland, Margaret A. Neale, and three anonymous reviewers for their very helpful advice on earlier versions of this paper.

References Argote, L. (1989). To centralize or not to centralize: the effects of uncertainty and threat on group structure and performance. Organizational Behavior and Human Decision Processes, 43, 58–74. Arrow, H., McGrath, J. E., & Berdahl, J. L. (2000). Small groups as complex systems. Formation, coordination, development, and adaptation. Thousand Oaks: Sage.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 1173–1182. Bettenhausen, K. L., & Murningham, J. K. (1985). The emergence of norms in competitive decisionmaking groups. Administrative Science Quarterly, 30, 350–372. Cannon-Bowers, J. A., Salas, E., Blickensderfer, E., & Bowers, C. (1998). The impact of cross-training and workload on teamfunctioning: a replication and extension of initial findings. Human Factors, 40(1), 92–101. Cannon-Bowers, J. A., Salas, E., & Converse, S. (1993). Shared mental models in expert team decision making. In J. N. Castellan, Jr. (Ed.), Individual and group decision making (pp. 221–246). Hillsdale, NJ: Lawrence Erlbaum. Cannon-Bowers, J. A., Tannenbaum, S. I., Salas, E., & Volpe, C. E. (1995). Defining competencies and establishing team training requirements. In R. A. Guzzo & E. Salas (Eds.), Team effectiveness and decision making in organizations (pp. 333–380). San Francisco: Jossey-Bass. Carroll, J. M., & Rosson, M. B. (1987). Paradox of the active user. In J. M. Carroll (Ed.), Interfacing thought: Cognitive aspects of humancomputer interaction (pp. 80–111). Cambridge, MA: MIT Press. Carter, S. M., & West, M. A. (1998). Reflexivity, effectiveness, and mental health in BBC TV-production teams. Small Group Research, 29, 583–601. Choi, H.-S., & Levine, J. M. (2004). Minority influence in work teams: the impact of newcomers. Journal of Experimental Social Psychology, 40, 273–280. Cooke, N. J., Kiekel, P. A., Salas, E., Stout, R., Bowers, C., & Cannon-Bowers, J. A. (2003). Measuring team knowledge: a window to the cognitive underpinnings of team performance. Group Dynamics: Theory, Research, and Practice, 7(3), 179–199. Daudelin, M. W. (1996). Learning from experience through reflexion. Organizational Dynamics, 24(3), 36–48. De Dreu, C. K. W. (2002). Team innovation and team effectiveness: the importance of minority dissent and reflexivity. European Journal of Work and Organizational Psychology, 11(3), 285–298. Edmondson, A. C., Bohmer, R. M., & Pisano, G. P. (2001). Disrupted routines: team learning and new technology implementation in hospitals. Administrative Science Quarterly, 46(4), 685–716. Edmondson, A. C., Roberto, M. A., & Watkins, M. D. (2003). A dynamic model of top management team effectiveness: managing unstructured task streams. The Leadership Quarterly, 14, 297–325. Ericsson, K. A., & Lehmann, A. C. (1996). Expert and exceptional performance: evidence of maximal adaptation to task constraints. Annual Review of Psychology, 47, 273–305. Espinosa, J. A., Lerch, F. J., & Kraut, R. E. (2004). Explicit versus implicit coordination mechanisms and task dependencies: one size does not fit all. In S. M. Fiore & E. Salas (Eds.), Team cognition: Understanding the factors that drive process and performance (pp. 107–129). Washington: APA. Fleiss, J. L. (1981). Statistical methods for rates and proportions (second ed.). New York: Wiley. Frese, M., & Zapf, D. (1994). Action as the core of work psychology: A German approach (second ed.. In H. C. Triandis, M. D. Dunette, & L. M. Hough (Eds.). Handbook of industrial & organizational psychology (Vol. 4, pp. 271–340). Palo Alto, CA: Consulting Psychologists Press. Fussell, S. R., Kraut, R. E., Lerch, F. J., Scherlis, W. L., McNally, M. M., & Cadiz, J. J. (1998). Coordination, overload and team performance: Effects of team communication strategies. Paper presented at the Conference on Computer Supported Cooperative Work, New York. Gabrys, G. (1994). Instructions for using the new NETATC program. Pittsburgh and Berne: Learning, Research, and Development

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142 Centre, University of Pittsburgh, PA; Institute of Psychology, University of Berne, Switzerland. Gersick, C. J. G. (1988). Time and transition in work teams: toward a new model of group development. Academy of Management Journal, 31, 9–41. Gersick, C. J. G., & Hackman, J. R. (1990). Habitual routines in taskperforming groups. Organizational Behavior and Human Decision Processes, 47, 65–97. Gevers, J. M. P., van Eerde, W., & Rutte, C. G. (2001). Time pressure, potency, and progress in project groups. European Journal of Work and Organizational Psychology, 10(2), 205–221. Gladstein, D. L. (1984). Groups in context. A model of task group effectiveness. Administrative science quarterly, 29, 499–517. Guzzo, R. A., Wagner, D. B., Maguire, E., Herr, B., & Lawley, C. (1986). Implicit theories and the evaluation of group process and performance. Organizational Behavior and Human Decision Processes, 37, 279–295. Hackman, J. R. (2002). Leading teams. Setting the stage for great performances. Boston: Harvard Business School Press. Hackman, J. R., Brousseau, K. R., & Weiss, J. A. (1976). The interaction of task design and group performance strategies in determining group effectiveness. Organizational Behavior and Human Performance, 16, 350–365. Hackman, J. R., & Wageman, R. (2005). A theory of team coaching. Academy of Management Review, 30(2), 269–287. Hollenbeck, J. R., Ilgen, D. R., Sego, D. J., Hedlund, J., Major, D. A., & Phillips, J. (1995). Multilevel theory of team decision making: decision performance in teams incorporating distributed expertise. Journal of Applied Psychology, 80(2), 292–316. Karau, S. J., & Kelly, J. R. (1992). The effects of time scarcity and time abundance on group performance quality and interaction process. Journal of Experimental Social Psychology, 28, 542–571. Klimoski, R., & Mohammed, S. (1994). Team mental model: construct or metaphor? Journal of Management, 20(2), 402–437. Kozlowski, S. W. J., Gully, S. M., Salas, E., & Cannon-Bowers, J. A. (1996). Team leadership and development: theory, principles, and guidelines for training leaders and teams. In M. M. Beyerlein, D. A. Johnson, & S. T. Beyerlein (Eds.). Advances in interdisciplinary studies of work teams: Team leadership (Vol. 3, pp. 253–291). Greenwich, CT: JAI Press. Marks, M. A., Mathieu, J. E., & Zaccaro, S. J. (2001). A temporally based framework and taxonomy of team processes. Academy of Management Review, 26(3), 356–376. Marks, M. A., Sabella, M. J., Burke, C. S., & Zaccaro, S. J. (2002). The impact of cross-training on team effectiveness. Journal of Applied Psychology, 87(1), 3–13. Marks, M. A., Zaccaro, S. J., & Mathieu, J. E. (2000). Performance implications of leader briefings and team-interaction training for team adaptation to novel environments. Journal of Applied Psychology, 85(6), 971–986. Mathieu, J. E., Heffner, T. S., Goodwin, G. F., Cannon-Bowers, J. A., & Salas, E. (2005). Scaling the quality of teammates’ mental models: equifinality and normative comparisons. Journal of Organizational Behavior, 26, 37–56. Mathieu, J. E., Heffner, T. S., Goodwin, G. F., Salas, E., & CannonBowers, J. A. (2000). The influence of shared mental models on team process and performance. Journal of Applied Psychology, 85(2), 273–283. McGrath, J. E. (1984). Groups, Interaction and Performance. Englewood Cliffs, NJ: Prentice-Hall. Moreland, R. L., & Levine, J. M. (1992). Problem identification by groups. In S. Worchel, W. Wood, & J. A. Simpson (Eds.), Group process and productivity (pp. 17–47). Newbury Park, CA: Sage. Orasanu, J. M. (1994). Shared problem models and flight crew performance. In N. Johnston, N. McDonald, & R. Fuller (Eds.), Aviation Psychology in Practice (pp. 255–285). Hants: Avebury Technical.

141

Peterson, S. E. (1992). College students’ attributions for performance on cooperative tasks. Contemporary Educational Psychology, 17, 114–124. Salas, E., Sims, D. E., & Burke, C. S. (2005). Is there a ‘‘big five’’ in teamwork? Small Group Research, 36(5), 555–599. Schippers, M. C., Den Hartog, D. N., Koopman, P. L., & Wienk, J. A. (2003). Diversity and team outcomes: the moderating effects of outcome interdependence and group longevity and the mediating effect of reflexivity. Journal of Organizational Behavior, 24, 779–802. Scho¨n, D. (1983). The reflective practitioner: How professionals think in action. New York: Basic Books. Schvaneveldt, R. W. (Ed.). (1990). Pathfinder associative networks: Studies in knowledge organization. Norwood, NJ: Ablex Publishing Cooperation. Smith-Jentsch, K. A., Campbell, G. E., Milanovich, D. M., & Reynolds, A. M. (2001). Measuring teamwork mental models to support training assessment, development, and evaluation: two empirical studies. Journal of Organizational Behavior, 22, 179–194. Smith-Jentsch, K. A., Mathieu, J. E., & Kraiger, K. (2005). Investigating linear and interactive effects of shared mental models an safety and efficiency in a field setting. Journal of Applied Psychology, 90(3), 532–535. Smith-Jentsch, K. A., Zeisig, R. L., Acton, B., & McPherson, J. A. (1998). Team dimensional training: A strategy for guided team selfcorrection. In J. A. Cannon-Bowers & E. Salas (Eds.), Making decision under stress (pp. 271–297). Washington DC: APA. Sniezek, J. A. (1992). Groups under uncertainty: an examination of confidence in group decision making. Organizational Behavior and Human Decision Processes, 52(1), 124–155. Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equations models. In S. Leinhart (Ed.), Sociological methodology (pp. 290–312). San Francisco: Jossey-Bass. Stasser, G., & Stewart, D. (1992). Discovery of hidden profiles by decision-making groups: solving a problem versus making a judgment. Journal of Personality and Social Psychology, 93(3), 426–434. Stasser, G., & Titus, W. (1985). Pooling of unshared information in group decision making: biased information sampling during group discussion. Journal of Personality and Social Psychology, 48, 1467–1478. Steiner, I. D. (1972). Group process and productivity. New York: Academic Press. Stout, R. J., Cannon-Bowers, J. A., Salas, E., & Milanovich, D. M. (1999). Planning, shared mental models, and coordinated performance: an empirical link is established. Human Factors, 41(1), 61–71. Thomas-Hunt, M. C., Ogden, T. Y., & Neale, M. A. (2003). Who’s really sharing? Effects of social and expert status on knowledge exchange within groups. Management Sciences, 49(4), 464–477. Tjosvold, D., Hui, C., & Yu, Z. (2003). Conflict management and task reflexivity for team in-role and extra-role performance in China. The International Journal of Conflict Management, 14(2), 141–163. Tjosvold, D., Tang, M. M. L., & West, M. A. (2004). Reflexivity for team innovation in China. The contribution of goal interdependence. Group & Organization Management, 29(5), 540–559. Tschan, F., Semmer, N. K., Na¨gele, C., & Gurtner, A. (2000). Task adaptive behavior and performance in groups. Group Processes and Intergroup Relations, 3(4), 367–386. Weingart, L. R. (1992). Impact of group goals, task component complexity, effort, and planning on group performance. Journal of Applied Psychology, 77(5), 682–693. West, M. A. (1996). Reflexivity and work group effectiveness: A conceptual integration. In M. A. West (Ed.), Handbook of Work Group Psychology (pp. 555–579). Chichester: John Wiley.

142

A. Gurtner et al. / Organizational Behavior and Human Decision Processes 102 (2007) 127–142

West, M. A. (2000). Reflexivity, revolution, and innovation in work teams. In D. A. Johnson & S. T. Beyerlein (Eds.). Advances in interdisciplinary studies of work teams: Product development teams (Vol. 5, pp. 1–29). Stanford, CT: JAI Press.

Wittenbaum, G. M., Hubbell, A. P., & Zuckerman, C. (1999). Mutual enhancement: Toward an understanding of the collective preference for shared information. Journal of Personality and Social Psychology, 77(5), 967–978.