Asymmetric Adaptability 1 ASYMMETRICAL ADAPTABILITY: DYNAMIC TEAM STRUCTURES AS ONE-WAY STREETS

We introduce the concept of “asymmetrical adaptability.” Using 66 teams, we replicated the common cross-sectional structural contingency finding that functional structures perform better within predictable environments, whereas divisional structures performed better within unpredictable environments. Unlike most tests of contingency theories, however, we directly tested whether teams could actually adapt in the manner directly implied by the theory and found evidence for “asymmetric adaptability.” Teams responded well to Functional-then-Divisional shifts, but they reacted poorly to Divisional-then-Functional shifts. We discuss the need to complement the static logic behind many contingency theories with a dynamic logic that explicitly challenges an assumption of symmetrical adaptation. Asymmetric Adaptability 3 ASYMMETRICAL ADAPTABILITY: DYNAMIC TEAM STRUCTURES AS ONE-WAY STREETS The applied behavioral and social sciences are replete with contingency theories. The core proposition underlying all contingency theories is that there is no “one best way” to solve all organizational problems. Instead, contingency theories argue that an approach that might be suitable under one specific set of circumstances may be unsuitable under a different set of conditions. This new set of conditions may demand an approach that is the exact opposite of what was formerly appropriate. Contingency theories of this type can be seen in many different areas (Miner, 1984). For example, with respect to allocation of rewards social interdependence theory suggests that competitive rewards should be employed when task interdependence among job incumbents is low but that cooperative rewards should be used when task interdependence is high (Deustsch, 1949). In Hackman and Oldham’s (1976) theory of job design, tasks should be designed one way if the job incumbents are high on growth need strength, but a different way if they are low in growth need strength. In the area of decision making, Vroom and Yetton’s (1973) model calls for autocratic decision making styles under one set of conditions, consultative styles under a different set of conditions, and group-based consensus procedures under yet a third set of conditions. In the area of leadership, the path goal theory groups leader behaviors into four different categories (directive, supportive, participate, and achievement-oriented) and lays out a whole host of factors that determine which set of behaviors to avoid or engage in, depending upon the circumstances (House 1971). This same type of contingency rhetoric can be found in countless other topic areas including socialization (Van Maanen & Schein, 1979), conflict Asymmetric Adaptability 4 management (Ruble & Thomas, 1976), communication structure and group performance (Shaw 1976), executive compensation (Balkin & Gomez-Mejia, 1987), organization design (Lawrence & Lorsch, 1967), and strategic planning (Grinyer, Al-Abazzaz, & Yasai-Ardekani, 1986). In addition to their core proposition denying the “one best way,” another common theme within contingency theories is their emphasis on adaptability. Naturally, if different times are going to demand different approaches, this implies that in a changing environment, the only way for an individual or collective to sustain success is to adaptively change with the times. Charles Darwin once noted that, “it is not the strongest of the species that survives, nor the most intelligent; it is the one that is most adaptable to change” (Darwin & Barlow, 1993, p. 122). This is a proposition that seems to be well accepted in current theorizing in the applied behavioral and social sciences. For example, Allred, Snow and Miles (1996) identify a new organizational form – the cellular organization – that epitomizes the central role of adaptability in current organizational approaches. According to Allred et al., cellular organizations are characterized by flexible boundaries, fluid role descriptions and flexible work processes, and “through experience and a common heritage, the cellular organization will learn, grow, and adapt to the uncertain environment and available opportunities” (p. 22). Similarly, Levitt et al. (1999) extol the virtues of virtual team design, where team structure is adaptively engineered to be aligned with project goals. Although the concept of such adaptive people and organizations is certainly alluring, much of the existing research on individual and group behavior suggests that the type of behavioral plasticity Asymmetric Adaptability 5 required by such systems may be difficult to achieve (Hendry, 1979, 1980; Smith & Nichol, 1981; DiMaggio & Powell, 1983; Pulakos, Arad, Donovan, & Plamondon, 2000). The preponderance of evidence in support of contingency approaches tends to be based on generalizations from static, between subjects research design, not within subject designs where a person or organization actually changes their approach from Time 1 to Time 2 (Hambrick, 1983; Miller & Friesen, 1983; Drazin & Van de Ven, 1985; Miller, 1988). Thus, little of the empirical research in support of contingency theories actually speaks directly to the issue of adaptability. Moreover, current approaches to adaptability have approached the concept contrasting adaptive systems with rigid systems. However, within the category of “adaptive systems,” it could very well be the case that adaptations that require a change in one direction may be more or less difficult relative to adaptations that require a change in the opposite direction. In addition, having shifted from one direction to another, it may not always be easy for system’s to cycle back to an earlier state. The purpose of this paper is to introduce the concept of “asymmetrical adaptability” and test the general proposition that adaptability in social systems can only be understood by directly examining both the point of origin and the destination of the adaptation. In general, we will argue that certain types of adaptation are going to be more natural than others, and that the prior experience of working under an earlier system will have marked impact on how people react to the adapted system. That is, although in a Euclidean sense the distance from Point A to Point B is the same as the distance between Point B to Point A, in a psychological sense, moving a social system from Point A to Point B may be more difficult than moving it from Point B to Point A. Whereas each type of movement may be characterized as “adaptive,” social systems may be Asymmetric Adaptability 6 more adaptable in some directions than others – hence the concept of “asymmetrical adaptability.” Although this general notion could be applied to any contingency theory, it is clearly beyond our powers to test this idea in all domains. Instead, the focus of this paper will be on structural changes in work groups. We focus on this specific area for several reasons. First, many organizations are adopting team-based structures, and thus there is a renewed interest in work groups (Devine et al., 1999). Second, few would dispute the relevance of the structure of the group as a central determinant of processes and outcomes (Hackman, 1992). Third, there is a well-accepted theory that deals with structure (Structural Contingency Theory; Burns & Stalker, 1961) that serves as a natural test bed for this general idea. In the following sections we will (a) review the general propositions of Structural Contingency Theory, (b) describe two possible “original structures” that a team might have to adapt from, (c) describe two “alternative structures” that a team may need to adapt to in order to stay aligned with their changing environment, (d) derive hypotheses regarding why it is more difficult to adapt in one direction compared to another, and (e) describe an empirical study designed to test this hypotheses. Structural Contingency Theory Characterizing Structures and Environments. Organizational structure describes how large numbers of persons are differentiated into smaller groups, as well as how the roles of members within these groups are differentiated and coordinated. One of the most critical dimensions of Asymmetric Adaptability 7 structure is departmentation (Wagner, 2000). Departmentation deals with the division of labor and refers to the degree to which work units are grouped based on functional similarity or on geographic/product market differentiation. In functional departmentation schemes, people are grouped based on the similarity of the work they perform, whereas divisional departmentation groups people by either the type of product produced or the geographic region served. At the team level, functional departmentation tends to create narrow, specialized roles, where the incumbent has less personal discretion and enhanced needs to coordinate with others. By contrast, divisional departmentation creates broader, more general roles, where the incumbent has wider personal discretion and reduced needs to coordinate with others (Burns & Stalker, 1961). Using an academic example, imagine a group of four students who have to do four case analyses for a specific class. One way they could divide up the case assignments is to make one person responsible for all the research that goes into each paper (research specialist), one person responsible for all the data analysis (data specialist), one person responsible for writing the report (writing specialist), one for the physical in-class presentation (presenter). The four could repeat this system of specialization for each of the four case analyses. This approach to task decomposition reflects a pure functional structure. Alternatively, the four could decide to let each student do one of the four projects independently, carrying out each of the four operations -research, data analysis, writing and presenting -on their own. This approach to task decomposition reflects a pure divisional structure. In the end, the exact same mission (preparing and presenting four case studies) is accomplished, but clearly in two very different ways. Asymmetric Adaptability 8 According to Structural Contingency Theory (SCT), there is no "one best way" to structure groups. Instead, this theory proposes that the group’s structure interacts with the nature of the task environment to influence performance (Burns & Stalker, 1961). In relatively predictable and stable environments, structures that employ functional departmentation tend to perform better than divisionally structured organizations. Functional structures are effective in this type of environment because they promote efficiency. Efficiency is created because redundancy across sub-units is minimized, and high levels of functional expertise can be developed (Pennings, 1992). Returning to our academic example, structuring the work functionally would be an efficient choice because the person who was doing the research might be able to do it all with a single trip to the library. Moreover, due to repetition, he or she might develop high levels of expertise in terms of how to do this one aspect of the task quickly and effectively. Although functional structures are efficient in relatively stable and predictive environments, these same structures tend to perform poorly in unstable and unpredictable environments. Unstable and unpredictable environments create changing and complex contingencies that overwhelm the specialized skills of individual team members. In unstable and unpredictable environments, divisional structures tend to perform better because they promote flexibility. Divisional structures are flexible because group members have broader capacities (i.e., they are less specialized) and their product or regional focus helps them react more quickly to local, idiosyncratic threats and opportunities. Returning to our running example, if the students choose a divisional structure, they will clearly be less efficient because each of the four members has to go to the library and do the research Asymmetric Adaptability 9 part of the task themselves. Since no one does this more than once, each is a "rookie" with each aspect of the task, and therefore they probably will not be as efficient in their activities as the "research specialist" in the functional teams. Moreover, divisional structures sometimes lack coordination, in the sense that “the right hand may not know what the left hand is doing,” and two different students may use the same research article to make incongruent points in their presentations. Although inefficient, if there was a sudden change in the task environment, the divisional structure would provide greater flexibility. For example, if one of the members became ill and the remaining team members had to do the fourth paper on their own. Because each of the divisional team members did each aspect of the task, any one of them could do it on their own. Alternatively, they could share the work in a number of different flexible manners because each has at least some experience with all aspects of the work. This change in the task environment would create more fundamental problems for the functional team, however, because none of them could do the paper on their own, and none of them have any experience in the one sub-task that was performed by the absent team member. This shows the clear efficiency--flexibility trade-off between the two structural choices, and illustrates why there is no one best way to structure teams. Instead of one best way, SCT argues that groups should be structured functionally in stable and predictable environments, but divisionally in unpredictable and unstable environments. Empirical research has supported this general proposition in both highly controlled laboratory experiments (Hollenbeck et al., in press) and more generalizable field studies (Drazen & Van de Ven, 1985). Given this established Asymmetric Adaptability 10 contingency, many have advocated that in the face of environmental change, groups need to be able to change their structure so that they are always in alignment. While this inference may logically follow from the existing data, it needs to be noted that this contingency has only been established via between subjects, cross-sectional studies. No one has ever directly documented that teams can actually switch from one structure to another without encountering unforeseen difficulties. Structural Starting Points and Adaptations. Within the SCT framework, a team that tries to change its structure could start from a number of different points, but for ease of exposition, we will focus on the endpoints of pure functional versus pure divisional schemes. On the one hand, an organization may start in a functional structure, and then need to adapt to a divisional structure. For example, the team may have started out in a stable and predictable environment, but because of some change in the competitive landscape (introduction of a new technology or new set of competitors), find that they do not have the required flexibility to compete effectively. According to SCT, this team should then adapt and change from their functional structure to a divisional structure (i.e., F-D adaptation). Alternatively, the team may have started out in a divisional structure, and then need to adapt to a functional structure. For example, the team may have started out in a new product development task where the user requirements and demands were at first uncertain, but then, after the eventual adoption of industry-wide standards, the task environment became more predictable and stable. This team may find that it lacks the efficiency to compete in the new environment, and SCT would suggest that this team should adapt and change from the divisional structure to a Asymmetric Adaptability 11 functional structure (i.e., D-F adaptation). Although on paper, it is no more difficult to redraw the organization chart from F-D to D-F, in operational reality it may be much more difficult for actual teams to shift in one direction versus the other. Indeed, there are three specific reasons why it might be more difficult for teams to adapt in the D-F direction relative to the F-D direction. First, in terms of task scope, teams that adapt in the F-D direction experience an increase in task scope. That is, they start out performing relatively narrow roles, and then switch to a system where they perform a more holistic job. Alternatively, those engaged in D-F adaptation experience a reduction in task scope. The job characteristics theory (Hackman & Oldham, 1976), as well as the empirical literature surrounding this theory makes it clear high task scope is associated with increased intrinsic motivation, which in turn is related to performance (Fried & Ferris, 1987). Thus, one might expect that the increase in intrinsic motivation associated with the F-D adaptation may make this type of shift more viable relative to a D-F adaptation where one experiences a reduction in task scope. Second, in terms of worker empowerment, teams that adapt in the F-D direction experience an increase in personal discretion and choice because of the shift. That is, the individual team member has more control over an entire product (or geographic region if the divisional structure is geographically based), not just one piece of a group project. On the other hand, team members making a D-F adaptation wind up sacrificing some level of personal discretion and choice as they take on more specialized roles. Thomas and Velthouse’s (1990) Theory of Empowerment, as well as the empirical research that supports this theory (Sprietzer, 1996) suggests that Asymmetric Adaptability 12 motivation is higher when people feel they have some degree of personal discretion over the work. Lack of discretion inhibits motivation, and hence, this again suggests that it may be more difficult to motivate people to engage in D-F adaptation relative to F-D direction, where the amount of personal discretion is enhanced. Finally, in terms of group norms, each of the two different structures place different demands on team members that could affect the group’s habits with respect to group processes such as communication. For example, higher levels of interdependence created by structuring a group functionally will result in relatively high levels of communication among group members. In contrast, the broad roles experienced by team members in divisionally structured groups force incumbents to concentrate of their relatively complex, high scope job. This need for concentration, when combined with the relative independence of divisionally structured team members makes communication less critical. According to Entrainment Theory (Ancona & Chong, 1996), once a set of norms and habitual activities become routine in a social system, these norms become self-reinforcing and entrained so they often persist over time – even after whatever original operational value that might have been attributed to the norms no longer persists. Indeed, there is direct empirical support for the notion that norms established early in a group’s existence often continue unabated even after the value of the norms is no longer evident (Bettenhausen & Murninghan, 1985). Thus, a group that starts out in a functional structure will develop norms for high levels of communication, and when this type of group shifts to a divisional structure, this high level of communication may persist. Although high levels of communication may not necessarily be required by the new Asymmetric Adaptability 13 structure, this may not necessarily harm the team’s effectiveness. In fact, it may even be beneficial in the sense that members can share the expertise they developed as functional specialists with each other as they enact their new expanded roles. On the other hand, a group that starts out divisionally will not develop norms for high levels of communication, but instead, members will be focused on concentrating on their own tasks. When this group shifts to a functional structure, if their initial norms and habits persist, the carryover will be dysfunctional because the functional structure demands high levels of communication. This will result in performance deficiencies that probably would not be experienced by teams that had simply started out in the functional structure in the first place. Hence, this again suggests that it may be more difficult for teams to engage in D-F adaptation relative to F-D adaptation. Thus, for at least three specific reasons, we feel that SCT is the type of contingency theory that might be fruitful for documenting “asymmetrical adaptability.” The theory it implies that there is no single best way to act, and that the appropriate solution (structure) to the organization’s problem (promoting performance) may change depending upon external circumstances (stability and predictability of the task environment). Thus, it implies systems need to be adaptive, but as currently stated, makes no allowance for the fact that certain adaptations may be much more difficult than others. That is, it does not recognize the possibility for asymmetrical adaptation and the notion that dynamic team structures may be one-way streets, where groups can move in one direction, but not another. Based on the theoretical analysis and empirical literature reviewed above, our general hypothesis, is that teams that experience a D-F structural adaptation will perform worse upon realignment than teams that experience an F-D structural adaptation. Asymmetric Adaptability 14 METHOD Research Participants and Task Research participants were 264 upper-level students who were arrayed into 66 fourperson teams. In return for their participation, each earned class credit, and all were eligible for cash prizes based upon performance. Participants engaged in a dynamic and networked computer simulation. The task was a modified version of the more generic Distributed Dynamic Decisionmaking (DDD) Simulation developed for the Department of Defense for research and training purposes in the area of command and control (see Miller, Young, Kleinman & Serfaty, 1998 for a complete description). In this task, teams monitor and enforce a “no-fly zone.” The specific variant of this task used in this research, MSU-DDD, was developed to be used in contexts where teams are comprised of anywhere from 2 to 5 members with little or no military experience. The grid representing the airspace to be defended is shown in Figure 1. This grid was partitioned in several ways. First, in terms of the person's physical location in the simulated geography, the grid was partitioned into four geographic quadrants of equal area (NW, NE, SW, SE), and each area was assigned to one of the team members (i.e., decision makers or DMs). The geographic region was also divided into three regions that varied in terms of the extent to which the areas needed to be protected from penetration by unfriendly forces. The regions were labeled neutral, restricted (a 12 by 12 grid in the center of the screen), and highly restricted (a 4 by 4 grid in the middle of the screen). The team's mission was to keep unfriendly forces from moving into the restricted and highly restricted areas, while at the same time, allowing friendly forces to move in and out of the same areas freely. Asymmetric Adaptability 15 When it came to monitoring the geographic space, each DM's base had a detection ring (Base DR in Figure 1) radius of roughly six grid units (demarcated by a circle like the one shown in Figure 1). The DM could detect the presence or absence of any track within this radius track. Each base also had an identification ring (Base IR in Figure 1) radius of roughly 4 grid units. He or she could discern the nature of the track (in terms of friendly versus unfriendly) once it was within this range. Any track outside the DR was invisible to the DMs, and therefore they had to rely on their teammates to monitor regions of the space that were outside their own quadrant. Each DM also had control various types of vehicles that could be launched, and then moved to different areas of the screen. These vehicles were semi-intelligent agents that could automatically perform certain functions (follow designated tracks, return to base to refuel, etc.), and hence the DM was a manager of these semi-intelligent agents. There were four different types of vehicles; (a) AWACS planes, (b) tanks, (c) helicopters, and (d) jets. Each of these vehicles varied in its capacities on four different dimensions; (a) range of vision, (b) speed of movement, (c) duration of operability, and (d) weapons capacity. The symbol representing each of the four vehicles is shown in Figure 1, along with the range of vision that characterized each vehicle (see surrounding circles). The various vehicles constituted a complex set of assets that ranged widely in their capacities. Each DM controlled four such vehicles that could all be operated concurrently. The characteristics and qualities of each vehicle are summarized in the first four rows of Table 1. Tracks were radar representations of forces moving through the geographic space monitored by the team. There were 12 unique types of tracks that varied in terms of being (a) friendly vs. Asymmetric Adaptability 16 unfriendly, (b) air-based vs. ground-based, (c) easy or difficult to disable, and (d) known or unknown upon identification. All tracks originated from the edge of the screen and proceeded inward. The last twelve rows of Table 1, summarizes the nature of the 12 different tracks. Prior to identification (e.g., when the track was close enough to be detected but not close enough to be identified) each track was represented by a question mark followed by a number that was set above a diamond (e.g., see the bottom right portion of Figure 1). The number reflected each track's unique identification number. Once the track came within the IR of either the base or a vehicle, the DM had the opportunity to identify the track. Once identified, the symbol representing the track changed from a diamond, to a rectangle with a letter-number combination such as those shown in the first column of Table 1 (see the middle portion of Figure 1). The number referred to the level of power needed to disable the track (low = 1, medium = 3 and high = 5) and had implications for what vehicle could perform certain tasks. Tanks could disable all tracks, helicopters could disable those numbered 1 and 3, and jets could only disable tracks numbered 1. The number 0 next to a letter indicated that the track was friendly, and that it should not be disabled. The letter indicated whether the track was air-based (A) or ground based (G). Air based tracks moved quickly, whereas ground-based tracks moved slowly. Thus, the track shown in the highly restricted zone at the center of Figure 1 was a friendly air track that did not need to be engaged. If it were a G5, that would mean that it was an unfriendly ground track that needed to be engaged – and could only be disabled by a tank. Asymmetric Adaptability 17 Manipulations and Measures Nature of Structural Adaptation: F-D versus D-F. Structural adaptation was manipulated between teams via the task. In the functional structure, vehicles were grouped by task specialty and assigned to DMs in order to create narrow, distinctive functional competencies (see Figure 2). In the functional structure, each DM managed four vehicles, all of the same type, taking on the role of tank command, helicopter command, jet command, or AWACS command. In this narrow role the person could manage one single type of task (e.g., the AWACS commander could only identify tracks and not engage them, the jet command could only engage A1 or G1 tracks but not any others), and hence had relatively low task scope. People working in these structures also were relatively low in discretion, in the sense that once a track became an issue, there was only one possible or logical person to execute the task (e.g., if an A5 or G5 track appeared, the only person who could successfully engage this task was the tank commander). Finally, the fact that each of the vehicles had their own unique strengths and weaknesses (e.g., the tank commander could engage everything, but could not see much, whereas the AWACS commander could see almost everything but not engage anything) meant that there was a high degree of interdependence in the functional structure. In the divisional structure, shown in Figure 3, vehicles were grouped geographically and assigned to DMs in order to create broad, general functional capacities. In the divisional structure, each DM was the manager of four vehicles, all of a different type. The DM then took basic responsibility for a specific geographic region (e.g., the northwest or southeast quadrant). Because of the complex array of strengths and weaknesses for each of the four vehicles, operating the four different platforms created a job with relatively high task scope. People Asymmetric Adaptability 18 working in the divisional structures also were relatively high in discretion, in the sense that once a track became an issue, any one of the four team members could execute the task, and there was no set demand where any one person had to do any one task. Finally, the fact that each of the team members in a divisional structure could manage any task themselves meant that there was a low degree of interdependence in the divisional structure. Each team operated under both structures, where the order of execution was manipulated. Half the teams started out in a functional structure in Stage 1, and then shifted into a divisional structure at Stage 2 (F-D Condition), whereas the other half of the teams started out in a divisional structure in Stage 1, and then shifted into a functional structure at Stage 2 (D-F Condition). Teams were randomly assigned to conditions, and hence were equivalent in all regards, except the nature of structural adaptation they were required to make. Task Environment. Teams were randomly assigned to the sequence of environmental conditions, and the sequence (Stage 1 versus Stage 2) was counter-balanced to control any order effects or effects for experience. In the unpredictable task environment, a random number generator determined the entry and exit point of each track. In addition, each track changed trajectory once over the course of its life, and thus the entry vector of the track could not be used to predict its exit point. In the predictable task environment, each track originated in the Northwest Quadrant and proceeded in a straight line diagonally, exiting the Southeast Quadrant. Thus, the origin and exit point of tracks in the predictable environment were easy to anticipate. Both types of task environments created their own unique challenges. The unpredictable condition was challenging because of the uncertainty with respect to where tracks were Asymmetric Adaptability 19 originating, but the predictable condition was also challenging because the heavy concentration of tracks flooding into a single area. Each condition contained 100 separate tracks and each stage lasted roughly 30 minutes. The nature of the structural adaptation and the nature of the task environment were completely crossed, creating a pure 2 X 2 factorial design where all the independent variables are orthogonal. Team Performance. At each of the two stages, the teams started off the simulation with 50,000 points and lost one point for each second that any unfriendly track was in the restricted zone and two points per second for each unfriendly track in the highly restricted zone. The teams also lost 300 points for disabling any friendly track. Performance was measured at both Stage 1 (point of origin structure) and Stage 2 (destination structure), and we also examined the difference in performance between the two stages. Data Analysis. Because the ultimate dependent variable of interest is a difference score reflecting the ease or difficulty in changing from one structure to another, we used the regression decomposition techniques describe by Edwards (1995) to analyze the data. Edwards’ (1995) procedure was developed specifically to deal with dependent variables that take on this form. This procedure allows the overall effect of each independent variable on the difference score to be decomposed into two complementary effects on the components, as well as the difference itself. In fact, the difference between the unstandardized B weights for each of the components equals the effect for the difference score. This procedure solves the basic interpretability problem traditionally associated with dependent variables that take the form of change scores (see Edwards, 1995 for more information on this technique). Asymmetric Adaptability 20 RESULTS Table 2 presents the results of regressing the difference in performance across stages, as well as performance at each stage on structure, the nature of the task environment and the interaction between the structure and task environment. The first column of this table shows the results for Stage 1 performance. In this regression, the nature of the structure and environment are entered in the first hierarchical step, followed by their interaction in a second step. This regression shows that at Stage 1, although there was slight tendency for divisional teams to outperform functional teams, this was not a statistically significant difference, and hence there is no main effect for either structure (F or D). Thus, in line with SCT, there is no one best structure across environments. There was also no significant main effect for the environment, although there was slight tendency for teams to perform better in the predictable environment. More critically, the results in the second step of this regression show that the interaction between structure and the environment was statistically significant, and accounted for an appreciable amount of variance (8%). The nature of this interaction is plotted in Figure 4, where, in order to aid interpretation, raw scores on the simulation were converted into z-scores. As is evident from this figure, in line with SCT, teams that were working in predictable environments performed better when structured functionally, whereas teams confronting unpredictable environments fared better when structured divisionally. This is a straightforward replication of a relatively ubiquitous SCT finding (Hollenbeck et al., in press; Drazen and Van De Ven, 1985), and it is just this type of result that has led many to suggest that organizations should change their structures in order to stay aligned with their environments. However, again it needs to be noted that this is a Asymmetric Adaptability 21 cross-sectional finding, and hence while documenting the contingency, it does not directly speak to adaptability. The second column of Table 1 shows the results for the Stage 2 performance, where each team changed their structure from F to D or D to F. Working from the bottom-up in this instance; we again see that there is a statistically significant interaction between structure and environment. This interaction explained 7% of the variance in performance, and when plotted, looked virtually identical to what is plotted in Figure 4. Again, all else equal, it was better for teams when their structure matched their environments in the manner prescribed by SCT. However, unlike in Stage 1, where there was no main effect for structure, in Stage 2 we see a statistically significant main effect for structure that explained 6% of the variance in performance. The nature of this main effect was such that, regardless of the nature of the environment, there was a tendency for F-D teams to outperform D-F teams. The relative magnitude of this difference is explicitly tested for significance in the third column of Table 3. When it comes to analyzing the differences between Stage 1 and Stage 2 performance, one sees only a main effect for structure, indicating that there were significantly different performance trajectories for the teams in the two conditions. In line with the crosssectional generalization of SCT, teams that adapted from functional to divisional structures improved when they changed. However, in contrast to the cross-sectional generalization of SCT, teams that tried to switch from a divisional to functional structure performed worse. Thus, whereas the functional versus divisional difference had a negligible effect on performance when this was all the teams knew (Stage 1), after teams had experience with one of the structures, it Asymmetric Adaptability 22 was more difficult for the ex-divisional teams to manage their new functional structure than it was for the ex-functional teams to manage their new divisional structure. Thus, there was asymmetrical adaptation when one considers the D-F versus F-D transition. DISCUSSION The field of organizational behavior is dominated by many different contingency theories, and the general idea behind all of theories of this type is that there is no “one best way” to solve all organizational problems. Instead, contingency theories imply that people or organizations need to engage in one set of behaviors when confronted with one set of conditions, but engage in a different set of behaviors under an alternative set of conditions. This implies that if people want to sustain excellence over time and changing environmental conditions, they need to be adaptive and able to switch back and forth from one set of routines to another. Although at a conceptual level, it is hard to argue with the virtues of adaptability, to date there has been little recognition of dynamic influences on behavior that may make some types of adaptation more or less natural relative to others. Instead, inferences regarding contingent relationships drawn from between subject, cross-sectional research have been generalized to make prescriptions about the desirability of adaptability. However, inferences about the desirability of adaptability really need to be based upon evidence from within subject, longitudinal studies where groups actually change from one state or set of routines to another. Asymmetric Adaptability and Team Structure. We introduced the concept of “asymmetrical adaptability” to capture the idea that the adaptation process can only be fully understood by Asymmetric Adaptability 23 directly analyzing the point of origin and the destination point associated with specific types of changes. This implies that movement from Point A to Point B may not be psychologically equivalent as moving from Point B to Point A. Using Structural Contingency Theory as an exemplar, we replicated the common, cross-sectional contingency finding at Stage 1 that indicated that functional structures perform better within predictable environments, whereas divisional structures performed better within unpredictable environments. The dynamic implication of this finding is that if a team’s task environment changes from Time 1 to Time 2, then, in order to stay in a fit with its environment, the team should change structures. Unlike most tests of contingency theories, in this study, we tested directly to see if teams could actually adapt in the manner implied by the theory and found evidence for “asymmetric adaptability.” Whereas both Divisional to Functional (F-D) and Functional to Divisional (F-D) shifts can both be considered “adaptive” under the right circumstances, in fact, one of these adaptations was much easier to negotiate relative to the other. Specifically, structural adaptation seemed to be a one-way street in the sense that it was much more natural for team to shift from Functional to Divisional (F-D adaptability) structure, than to switch in the other direction. We speculated that the F-D adaptation was more natural to make because it reflected an increase in task scope and discretion for the team members over time, and because the norms developed in divisional structures regarding communication were counter-productive in the new functional structure. Asymmetric Adaptability in Alternative Domains. Although it was beyond the scope of this one study to test all existing contingency theories in this same manner, we think is interesting to Asymmetric Adaptability 24 speculate on how various contingency theories might stand up to the same kind of analysis. For example, the Vroom-Yetton model of leadership is a contingency theory that states that the decision making process that the leader engages in should depend upon characteristics of the followers and the situation (Vroom and Yetton, 1973). Within one set of circumstances, the model might recommend that the leader use a process labeled GII. In this instance, the leader is supposed to share the problem with subordinates, and together they should generate and evaluate alternatives. The goal would be to work slowly and attempt to reach consensus on a solution. The leader serves in the role of the chairperson, coordinating the discussion, and keeping it focused on the problem. The leader makes sure the critical issues are discussed but does not try to influence the group to adopt his or her solution. In the end, the leader needs to be willing to accept and implement any solution that has support from the group. Alternatively, within a different set of circumstances, the model might recommend that the leader use a process labeled AI, where the leader solves the problem himself or herself, using personal information, and not involving the subordinates in any way. While accepting the static logic that might lead this theory to recommend such different styles under different circumstance, the concept of asymmetrical adaptability makes one question the dynamic logic involved when, in a real operational setting, the social system tries to move from one state to the other. Specifically, if the original circumstances dictate a series of initial decisions where the AI style is appropriate and executed, but then circumstances change, it may be quite natural for the group to adapt from a an AI to GII style because the group members’ roles and influence are expanded in Asymmetric Adaptability 25 the new adapted situation. However, if the original circumstances allow for a long series of GII style decisions, but then circumstances change – demanding the leader adopt a new AI style – will this shift be as easy to execute as the other? In this instance, group members are being asked to sacrifice influence and discretion. Will their reaction to the AI style be the same as those who only experience the AI style at Stage 1, and had never experienced GII? There may very well be asymmetrical adaptability in this situation, such that it easier for a team to evolve from a series of AI to GII decision rules, than it would be for the same team to adapt from a series of GII to AI decision rules. Another contingency theory that might be instructive to analyze in this same manner is social interdependence theory (Deutsch , 1949). Social interdependence theory deals with competitive versus cooperative reward allocations, and like all contingency theories, suggests that there is no one best way to allocate awards. According to this theory, if task interdependence among team members is low, then the organization should employ competitive rewards that pit one member against the other. However, when the task demands high levels of interdependence among team members, the organization should employ cooperative rewards, where all team members experience the same outcome regardless of their individual contribution. The general static logic underlying this theory is that cooperative rewards promote collaboration and teamwork under conditions of high interdependence, whereas competitive rewards prevent social loafing among group members when interdependence requirements are low. Again, one can accept the static logic underlying this theory, and at the same time question the dynamic logic if the implication is that group should change from one reward allocation structure Asymmetric Adaptability 26 to another if the level of task interdependence changes. For example, a group might originate in a situation where there is low task interdependence, and work under a competitive reward structure where every person is looking out for himself or herself. Over the course of time, this will affect the behaviors and interpersonal relationships of these people in a particular direction. What if the level of task interdependence changes, however, and this group has to adapt to a new, cooperative reward structure. It is not clear that the level and type of cooperative behavior that would be exhibited by these people, would in any way resemble what would be seen in a group that was working under a cooperative structure without ever having experienced the competitive structure. Indeed, one may see a type of “cutthroat cooperation” in the former group that would differ dramatically from the cooperation seen in the group that never experienced competition in this context. Alternatively, a group that might have started off in a cooperative reward allocation structure may experience less disruptive processes if asked to switch to a new competitive structure. Having established a set of supportive relationships at Stage 1, the type of competition seen as Stage 2 may be a type of “friendly competition” that does not create the kind of dynamic difficulties that might be experienced in the “cutthroat cooperation” group. This would again imply that there is asymmetrical adaptability in the sense that is more natural for a group to evolve from cooperation to competition, than from competition to cooperation. Although any one of a number of other contingency theories could be analyzed in a similar manner, the overall point that should be drawn from this discussion should be clear. The static logic that provides the basis for many contingency theories needs to be complemented by a Asymmetric Adaptability 27 dynamic logic. This dynamic logic needs to consider whether the changes that are dictated by sequential applications of different behavioral routines lead to the type of asymmetrical adaptability that was documented in this study. Limitations and Directions for Future Research. First, although this study called into question the desirability of D-F shifts in team structure, it did not explore all the possible alternatives short of a total D-F shift than might have been available for these teams. For example, when the environment became more predictable, it might have been better for the D-F teams to switch from D to something a little less extreme relative to a strict F structure. That is, operationally speaking, instead of shifting from a situation where they controlled all four types of vehicles to a situation where they controlled only a single vehicle, the D-F teams may have been better off shifting to a “compromise structure” that allowed them control over two vehicles rather than just one. This would be a shift in the functional direction, but a not a total shift to a functional structure. This type of compromise structure may make for a better destination for these teams given their point of origin. Alternatively this structure could serve as a “transition structure” as the team moves from D to F in two or three steps rather than one. Second, although this study suggested three different reasons why D-F adaptability is more difficult relative to F-D adaptability, we did not separately manipulate each of these three factors (i.e., create three extra experimental groups) to see which was the most critical factor in the process. If one or two factors could be isolated as the primary mediating mechanism, this may point to training programs that might be used to mitigate some of the problems inherent in making D-F transitions. That is, if the primary problem with the D-F transition teams is their lack Asymmetric Adaptability 28 of communication in the new structure, this might be anticipated and training programs designed to enhance communication frequency might be put into place. The purpose of this study was to demonstrate the concept of asymmetric adaptability, however, and we were only secondarily interested in this specific domain. In our opinion, future research directed at extending these findings to other domains may be more valuable than research pinning down exactly why there was asymmetry in this one particular domain. Third, because this study was conducted in a laboratory context, there are the traditional concerns one might have regarding the external validity of these findings. On the one hand, there are certain features of this task and our research participants that do achieve what Berkowitz and Donnerstein (1982) refer to as mundane realism. Most weapons directors or those operating consoles in command and control situations are lieutenants, and thus our research participants are about the right age and education level. In addition, the command and control task is one where people sit at computer monitors and collect information exactly as is done in this simulation (which was developed for the Department of Defense). Also, these people tend to be assembled into crews based upon rotations that preclude their working together for long periods of time (i.e., many of these crew have limited histories and futures). On the other hand, it is also clear that we could not ever simulate the psychological processes involved in real warfare in this laboratory context. While the consequences of decisions for our research participants were not as dramatic as they would be in a real situation, this was a psychologically engaging task and research participants were visibly upset when they performed poorly or made errors. The research participants were also aware of the financial bonuses that Asymmetric Adaptability 29 were available to the top performing teams, and the valence and probability of such rewards motivated them to perform well. Thus, there were consequences associated with performance that mattered to these people, so we believe that "psychological realism" (Berkowitz and Donnerstein, 1982) was quite high. Beyond the issues of mundane and psychological realism, however, one needs to keep the nature of the research question in mind when assessing the relevance of external validity. We are less concerned with actual command and control situations than we are in testing the dynamic application of Structural Contingency Theory in a team context. Since there is no formal aspect of this theory that would imply it would not work in this specific context, this context provides a legitimate venue within which to test the theory. As Ilgen (1986) noted, this is precisely the type of question that is well suited to laboratory contexts. In addition, it should be noted that this study simply could never have been conducted in the field. That is, one of the major problems with trying to scientifically study real command and control situations is that the number of teams is small, their tasks geographically and politically idiosyncratic, and their availability limited. There are no cases where one has multiple teams that experience the exact same tasks, in the exact same order, in the exact same context with everything but structure controlled. Finally, the whole issue of external validity needs to be considered in the light of the fact that to technically achieve external validity within one study, one has to randomly select research participants, tasks and times from some meaningful population. Clearly this was impossible in Asymmetric Adaptability 30 this context, as well as most others. Indeed, one never sees a study where the tasks chosen for the research were randomly selected from some meaningful population of tasks, and therefore, it is virtually impossible to meet the technical requirements for generalizing across tasks. Certainly, one cannot conclude that simply because a study was conducted in one specific field context, that its findings would be generalizable to all other field contexts (Flanagan & Dipboye, 1981). Fortunately, as Cook and Campbell (1979) have noted, “a strong case can be made that external validity is enhanced more by many heterogeneous small experiments than by one large experiment employing random selection of subjects, tasks and times” (p. 80). Moreover, as directly shown by Anderson, Lindsay and Bushman (1999), the correlation between effect sizes obtained in laboratory settings and field settings generally exceed .70. Thus, there is the hope that the generalizability of the findings reported here will become evident as other researchers, perhaps interested in these findings, replicate this study with other small experiments with different samples and tasks conducted at different times. Asymmetric Adaptability 31 REFERENCESAllred, B. B., C. C. Snow, and R. E. Miles.1996 “Characteristics of managerial careers in the 21 century.” Academy of ManagementExecutive 10: 17-27. Ancona, D. and C. Chong.1996 “Entrainment: Pace, cycle, and rhythm in organizational behavior.” In L. L. Cummings andB. M. Staws (Eds.) Research in Organizational Behavior, vol. 18 (pp. 251-284) JAI Press Inc. Anderson, C.A., J. L. Lindsay, and B. J. Bushman.1999 “Research in the psychological laboratory: Truth or triviality.” Current Directions inPsychological Science, 8: 3-9. Balkin, D. B., and L. R. Gomez-Mejia.1987 “Toward a contingency theory of compensation strategy.” Strategic Management Journal,8: 169-182. Berkowitz, L. and E. Donnerstein.1982 “External validity is more than skin deep.” American Psychologist, 37: 245-257. Bettenhausen, K. and J. K. Murnighan.1985 “The emergence of norms in competitive decision-making groups.” Administrative ScienceQuarterly, 30: 350-372. Asymmetric Adaptability 32 Burns, T., and G. M. Stalker.1961 “The management of innovation.” London: Tavistock. Cook, T.D., and D. T. Campbell.1979 “Quasi-experimentation: Design and analysis issues for field settings.” Chicago: RandMcNally. Darwin, C and N. Barlow.1993 “The autobiography of Charles Darwin: 1809-1882.” New York:W.W. Norton. Devine. D.J., L. D. Clayton, J. L. Philips, B. B. Dunford, and S. B. Melner.1999 “Teams in organizations: Prevalence, characteristics, and effectiveness.” Small GroupResearch, 30: 678-711. Deutsch, M.1949 “A theory of cooperation and competition.” Human Relations, 2: 129-152. Dimaggio, P. J., and W. W. Powell.1983 “The iron cage revisited: Institutional isomorphism and collective rationality inorganizational fields.” American Sociological Review, 48: 147-160.