ASYMMETRIC ADAPTABILITY: DYNAMIC TEAM STRUCTURES AS ONE-WAY STREETS

ASYMMETRIC ADAPTABILITY:DYNAMIC TEAM STRUCTURES AS ONE-WAY STREETS

HENRY MOONEmory University

JOHN R. HOLLENBECKSTEPHEN E. HUMPHREY

DANIEL R. ILGENBRADLEY WEST

Michigan State University

ALEKSANDER P.J. ELLISUniversity of Arizona

CHRISTOPHER O. L. H. PORTERTexas A&M University

This study tested whether teams working on a command and control simulationadapted to structural change in the manner implied by contingency theories. Teamsshifting from a functional to a divisional structure showed better performance thanteams making a divisional-to-functional shift. Team levels of coordination mediatedthis difference, and team levels of cognitive ability moderated it. We argue that thestatic logic behind many contingency theories should be complemented with a dy-namic logic challenging the assumption of symmetrical adaptation.

Within the organizational sciences, taking astructural contingency theory approach to optimalorganizational performance implies that there aredifferent task environments, different ways inwhich to structure organizations, and positive im-plications for performance when the structure of anorganization and the dictates of the task environ-ment have a proper fit (Burns & Stalker, 1961; Law-rence & Lorsch, 1967; Miller, 1988; Pennings,1992). Simply, structural contingency theory advo-cates an “if this, then that” structure by environ-ment contingency. Support for this proposition canbe found in cross-sectional research on both large-scale organizations (Drazin & Van de Ven, 1985)and smaller work teams (Hollenbeck et al., 2002).

Many have used the structural contingency the-ory proposition that no one structure is always bestto argue that in the current, fast-paced, technology-

driven business environment, organizations needto be designed around flexible, team-based struc-tures (Townsend, DeMarie, & Hendrickson, 1998).Townsend and colleagues viewed newly flexibleorganizations as demonstrating “a pronouncedstructural difference from traditional workgroupparticipation because of their ability to transformquickly according to changing task requirementsand responsibilities” (1998: 23). Similarly, Levitt,Thomsen, Christiansen, Kunz, Yan, and Nass(1999) extolled the virtues of virtual team design, inwhich team structure is adaptively engineered to bealigned with project goals. Allred, Snow, and Miles(1996) characterized this emerging model of orga-nizations as “cellular structures,” a term implyingboth individual units’ ability to function indepen-dently and the ability of multiple units to engage inmore complex functions through interdependentaction, depending upon environmental demandsand constraints.

It is hard to argue with the virtues of flexibility,and the concept of infinitely adaptive persons andorganizations is certainly alluring. However, thedifficulties in maintaining such high levels of ad-aptation in organizations should not be underesti-mated. The preponderance of evidence in supportof contingency approaches tends to be based ongeneralizations from static, between-groups re-

This research was supported in part, by Grant N00014-96-1-0983 from the Cognitive and Neural Sciences Divi-sion of the Office of Naval Research. Although supportfor this work is gratefully acknowledged, the ideas ex-pressed herein are those of the authors and not necessar-ily endorsed by the funding agency. The authors wouldlike to thank Christine Jackson, for assistance in datacollection, as well as Jeffrey Daniels and Phillip Young,for programming assistance.

� Academy of Management Journal2004, Vol. 47, No. 5, 681–695.

681

search designs, not within-group designs in whicha team or organization actually changes its struc-ture between two data collection times. In fact,little of the empirical research in support of con-tingency theories in general, or structural contin-gency theory specifically, has spoken directly tothe issue of change and adaptability over time oracross environments. Thus, despite the conceptualattractiveness of this type of reconfigurability,structural contingency theory has come under at-tack by those documenting the conditions thatmake change difficult (DiMaggio & Powell, 1983).

The debate between those who advocate thevalue of structural adaptation and those who ques-tion its feasibility, however, has been framed in allor none terms. To date, there has been no discus-sion of the direction of change that various types ofreconfiguration require. That is, both sides of thisdebate have presumed that structural adaptability(or structural inadaptability) is symmetric for alltypes of changes; no theory exists according towhich one type of change might be systematicallymore difficult to accomplish than another, no doany empirical data test this proposition.

The purpose of this article is to introduce theconcept “asymmetric adaptability” and to presentthe results of a test of the general proposition thatadaptability in social systems can only be under-stood by directly examining both the origin and thedestination of the adaptation. In general, we arguethat certain types of adaptation are going to be morenatural than others, and that the prior experience ofworking under an earlier system influences howpeople react to the adapted system. That is, al-though in a Euclidean sense, the distance frompoint A to point B is the same as the distance frompoint B to point A, in a psychological sense, mov-ing a social system from point A to point B may bemore difficult than moving it from point B to pointA. Moreover, once a social system has been movedfrom point A to point B, the challenge inherent inreconfiguring the system back in the reverse direc-tion may be greater than the challenge associatedwith the initial reconfiguration.

In the following sections of this article, we (1)review the general propositions of structural con-tingency theory, (2) describe, through illustrativeexamples, two possible points of origin that ateam’s structure might take on initially (that is,might adapt from) and two destinations that a teammay need to subsequently move its structures to inorder to stay aligned with its environment (that is,adapt to), (3) derive hypotheses regarding why it ismore difficult to adapt in one direction than in theother, (4) suggest why coordination may be themediation mechanism that underlies asymmetry,

and (5) propose cognitive ability as a moderator ofasymmetry. These ideas are then empirically testedin an experiment in which teams were required tostructurally reconfigure in opposing directions.

STRUCTUAL CONTINGENCY THEORY

The applied behavioral and social sciences arereplete with contingency theories (Miner, 1984).The core proposition underlying all contingencytheories is that there is no one best way to solve allorganizational problems. Instead, the proponents ofcontingency theories argue that an approach thatmight be suitable under one specific set of circum-stances may be unsuitable under a different set ofconditions (Dill, 1958). This new set of conditionsmay demand an approach that is the exact oppositeof what was formerly appropriate.

For example, with respect to allocation of re-wards social interdependence theory suggests thatcompetitive rewards should be employed whentask interdependence among job incumbents is lowbut that cooperative rewards should be used whentask interdependence is high (Beersma, Hollen-beck, Humphrey, Moon, Conlon, & Ilgen, 2003;Deustsch, 1949). In Hackman and Oldham’s (1976)theory of job design, tasks should be designed oneway if the job incumbents are high in “growth needstrength,” but a different way if they are low ingrowth need strength. In the area of decision mak-ing, Vroom and Yetton (1973) called for autocraticdecision-making styles under one set of conditions,consultative styles under a different set of condi-tions, and group-based consensus procedures un-der yet a third set of conditions. In the area ofleadership, the path goal theory categorizes leaderbehavior into four different categories (directive,supportive, participative, and achievement-ori-ented) and lays out a whole host of factors thatdetermine which set of behaviors to avoid or en-gage in, depending upon the circumstances (House,1971). Similar contingency-based theories can befound in countless other topic areas, including so-cialization (Van Maanen & Schein, 1979), conflictmanagement (Conlon, Moon, & Ng, 2002), commu-nication structure and group performance (Shaw,1976), and executive compensation (Balkin & Go-mez-Mejia, 1987).

Structural Contingency Theory: The Types ofDepartmentation

Structural contingency theory, which has thesame form as these other contingency theories,grew from early observations of different organiza-tional structures and different external environ-

682 OctoberAcademy of Management Journal

ments. Archival analyses of how well various orga-nizations performed under different conditions ledto the conclusion that long-term viability was con-tingent upon a proper match between an organiza-tion’s structural design and the particular environ-ment it faced. Subsequent research employingcross-sectional, between-subjects designs has pro-vided some support for these proposed relation-ships (Drazin & Van de Ven, 1985; Hambrick, 1983;Miller & Friesen, 1983).

Although the notion of asymmetric adaptabilitycould be applied to any of the contingency theorieslisted above, the focus of this experiment was struc-tural changes in work groups. Many organizationsare adopting team-based structures, and thus thereis a renewed interest in work groups. Ilgen (1999)noted the relevance of group structure as a centraldeterminant of processes and outcomes, if structureis conceptualized in terms of how large numbers ofpeople are differentiated into smaller groups, aswell as how the roles of members within thesegroups are differentiated and coordinated (Pen-nings, 1992).

One of the most critical dimensions of structureis “departmentation” (Wagner, 2000). Departmen-tation deals with the division of labor and refers tothe degree to which work units are grouped on thebasis of functional similarity or on the basis ofgeographic/product-market differentiation. In func-tional departmentation schemes, grouping is basedon the similarity of the work people perform,whereas in divisional departmentation schemes,grouping is based on either the type of product theyproduce or the geographic region they serve. At theteam level, functional departmentation tends tocreate narrow, specialized roles, in which an in-cumbent has low personal discretion and a strongneed to coordinate with others. By contrast, divi-sional departmentation creates broader, more gen-eral roles, in which an incumbent has wide per-sonal discretion and low needs to coordinate withothers (Burns & Stalker, 1961).

For example, imagine a team of four mutual fundresearch analysts who have to present four compre-hensive company case analyses to a large institu-tional client. One way they could divide up theassignments would be to make one team memberresponsible for all the research that will go into allfour papers (and thus serve as a research specialist),make another team member responsible for all thedata analysis associated with all four papers (a dataspecialist), make a third team member responsiblefor writing the four reports (a writing specialist),and finally, make the last team member responsiblefor delivering the physical presentations (a pre-senter). This approach to task decomposition

would establish a purely functional structure. Al-ternatively, the four could decide to let each indi-vidual do one of the four projects independently,carrying out each of the four operations—research,data analysis, writing, and presenting—on his orher own. This approach to task decompositionwould establish a purely divisional structure. In theend, the exact same mission (preparing and pre-senting four case studies) will be accomplished, butclearly in two very different ways.

Structural Contingency Theory: Departmentationand Task Environment

The structural contingency theory answer to thequestion, “What task decomposition scheme isbest?” is that a group’s structure interacts with itstask environment to influence performance. In rel-atively predictable and stable environments, struc-tures that employ functional departmentation tendto perform better than those that employ divisionalstructures. Functional structures are effective inthis type of environment because they promote ef-ficiency. Efficiency is created because redundancyacross subunits is minimized and high levels offunctional expertise can be developed.

Returning to our example of the four case studies,it seems clear that if the analysts choose a divi-sional structure, they will be less efficient becauseeach has to physically access many informationsources individually. Since no one does this re-search repeatedly, each is a relative novice, andwill therefore probably not be as efficient in carry-ing it out as the “research specialist” in the func-tional team who does this one task over and overagain. The same novice versus specialist differencemight arise in all of the other functional responsi-bilities, leading the divisional team to be less effi-cient than the team of specialists. Moreover, divi-sional structures sometimes lack coordination, inthe sense that “the right hand may not know whatthe left had is doing,” and two analysts may use thesame information source to make incongruentpoints in their presentations.

Although functional structures are efficient inrelatively stable and predictive environments,these same structures tend to perform poorly inunstable and unpredictable environments. Unsta-ble and unpredictable environments create chang-ing and complex contingencies that poorly matchthe specialized skills of individual team members.Returning to our running example, if the individu-als choose a functional structure, they will proba-bly be more efficient at the assigned task; however,if there were a sudden change in the task environ-ment, this structure would provide less flexibility

2004 683Moon, Hollenbeck, Humphrey, Ilgen, West, Ellis, and Porter

than the divisional structure. For example, supposeone of the members became ill and the remainingteam members had to do the fourth report on theirown. Because each divisional team member haddone each aspect of the task, any one of them coulddo the fourth report alone. Alternatively, theycould share the work in a number of ways, becauseeach had had at least some experience with allaspects of the work. This change in the task envi-ronment would create more fundamental problemsfor the functional team, however, because none ofthem would have had any experience in the sub-task the absent team member performed.

This example shows the clear efficiency-flexibil-ity trade-off the two structural choices involve andillustrates why there is no one best way to structureteams. According to structural contingency theory,groups should be structured functionally in stableand predictable environments, but divisionally inunpredictable and unstable environments.

Recently, Hollenbeck and coauthors (2002) dem-onstrated the application of structural contingencytheory, which has mostly been studied at the orga-nizational level of analysis, to the team level ofanalysis. This experiment showed, in line withstructural contingency theory predictions, that noone structure was best at all times or under allconditions, but instead, that divisionally structuredteams performed better in unpredictable task envi-ronments, whereas functionally structured teamsperformed better in predictable environments. Onepurpose of this experiment was to replicate (Eden,2002) findings regarding structural contingencytheory at the team level (Hollenbeck et al., 2002) inorder to provide a base from which to develophypotheses about asymmetric adaptability. Specif-ically:

Hypothesis 1. The nature of a task environ-ment moderates the effect of team structure onperformance in such a way that functionalstructures will perform better in predictableenvironments, whereas divisional structureswill perform better in unpredictable environ-ments.

ASYMMETRIC ADAPTABILITY

Characteristics of Starting Points andDestinations

In view of the theory and research discussedabove, many have advocated that in the face ofenvironmental change, groups need to be able tochange their structures so that they are alwaysaligned with their environments (Allred et al.,1996; Levitt et al., 1999; Townsend et al., 1998).

Although this inference may logically follow fromthe existing data, it needs to be noted that thiscontingency has only been established via cross-sectional studies, and/or studies with between-groups or between-organizations research designs.No one has ever directly documented that teamscan actually switch back and forth from one struc-ture to another, moving in either direction equallywell, without encountering unforeseen difficulties.

The concept of asymmetric adaptability impliesthat reconfigurability is directionally dependentand that it is easier to move social systems in somedirections than it is to move them in other direc-tions. In structural terms, an organization may startin a functional structure, and then need to adapt toa divisional structure. For example, a team maystart out in a stable and predictable environment,but because of some change in the competitivelandscape (the introduction of a new technology ora new set of competitors), its members find thatthey do not have the flexibility required to competeeffectively. According to structural contingencytheory, this team should then adapt and changefrom their functional structure to a divisional struc-ture (that is, carry out functional to divisional ad-aptation).

On the other hand, a team may start out in adivisional structure and then need to adapt to afunctional structure. For example, say a team’s ini-tial task is development of a new product for whichuser requirements and demands are at first uncer-tain. Adoption of industry-wide standards thenmakes the task environment more predictable andstable. This team may find then that it lacks theefficiency needed to compete in the new environ-ment, and structural contingency theory wouldsuggest that this team should adapt and changefrom the divisional structure to a functional struc-ture (a D-F adaptation). Although on paper, it is nomore difficult to redraw an organization chart toshow functional to divisional (F-D) adaptation thanD-F adaptation, in operational reality it may bemuch more difficult for actual teams to shift in onedirection than in the other.

Entrainment theory. In terms of group norms,each of the two different structures places differentdemands on a team’s members that could affect thegroup’s habits with respect to group coordinationprocesses—especially the frequency of communi-cation and mutual support. For example, the nar-row roles and the relatively high levels of interde-pendence created by structuring a groupfunctionally will result in high levels of coordina-tion (that is, communication and support) amonggroup members. In contrast, the broad and complexroles experienced by team members in divisionally

684 OctoberAcademy of Management Journal

structured groups force incumbents to concentrateon their relatively high-scope jobs. This need forconcentration, when combined with the relativeindependence of divisionally structured teammembers, makes coordination less critical.

According to entrainment theory (Ancona &Chong, 1996), once a set of norms and habitualactivities becomes routine in a social system, thesenorms become self-reinforcing—or entrained—sothat they often persist over time, even after theiroriginal operational value is gone. Indeed, there isdirect empirical support for the notion that normsestablished early in a group’s existence often con-tinue even after their value is no longer evident(Bettenhausen & Murninghan, 1985).

Thus, a team that starts out in a functional struc-ture will develop norms for high levels of coordi-nation behaviors (communication and support),and when this team shifts to a divisional structure,these norms may persist. Although the new struc-ture may not require high levels of coordination,the persistence of these norms may not harm theteam’s effectiveness. In fact, they may even be ben-eficial in the sense that members can share theexpertise they developed as functional specialistswith each other as they enact their new, expandedroles.

On the other hand, a team that starts out in adivisional structure will not develop the samenorms for high levels of communication and mu-tual support, but instead will develop norms forindependent activity. When this group shifts to afunctional structure, persistence of its initial normsand habits will be dysfunctional because the func-tional structure demands high levels of coordina-tion, communication, and mutual support. Theirabsence will result in performance deficiencies thatprobably would not be experienced by a team thathad simply started out with a functional structurein the first place. More formally, our second twohypotheses are:

Hypothesis 2. Teams that experience a divi-sional to functional structural adaptation willperform worse upon realignment than teamsthat experience a functional to divisionalstructural adaptation.

Hypothesis 3. Coordination behaviors mediatethe difference in performance between teamsengaged in divisional to functional and func-tional to divisional structural adaptation.

In contrast to what might be expected from en-trainment theory, other arguments could be raisedthat might suggest that a divisional to functionalshift may be easier to manage than an functional to

divisional shift. For example, a divisional to func-tional shift allows individuals to first gain a senseof the “big picture,” which may subsequently helpthem see how their functional tasks, after the struc-tural change, will fit into the overall scheme of thegroup’s task. Moreover, starting out in a divisionalstructure allows individuals to develop a broadrange of knowledge and skills that may make for aneasier transition into narrower roles. Thus, al-though the predictions based upon entrainmenttheory are plausible, they are certainly not self-evident, and hence require empirical testing.

Team composition. Although entrainment the-ory may imply that functional to divisional adap-tation may be easier than divisional to functionaladaptation, the nature of a team could also affectthe adaptation process as it moves in different di-rections. For example, the shift to a divisionalstructure from a functional one may place in-creased cognitive demands on a team. In terms oftask scope (Hackman & Oldham, 1976), teams thatadapt in the functional to divisional direction startout performing relatively narrow tasks, and thenswitch to a system in which they perform a moreholistic role. In terms of worker empowerment(Thomas & Velthouse, 1990), teams that adapt inthe functional to divisional direction experience anincrease in personal discretion and choice, whereaspersonal discretion is reduced for those changingin the opposite direction.

Both of these changes in the nature of a job in-cumbent’s task make it more complex, and researchon task complexity has shown that high levels ofcognitive ability are more critical for complex jobsthan for simple jobs (Hunter & Hunter, 1984). In-deed, in an experiment using data on teams withstatic structures (fixed structures that did notchange), Hollenbeck and his coauthors (2002)showed that teams structured divisionally per-formed better the higher the average general cogni-tive ability of their members.

Thus, although in general, it might be easier forteams to shift in the functional to divisional direc-tion than in the opposite direction, the averagegeneral cognitive ability of team members may af-fect the ease of the functional to divisional transi-tion. Teams that are relatively high in general cog-nitive ability may respond much more positively tofunctional to divisional shifts than teams that arelow in this characteristic, because the former havethe cognitive capacity to manage the enlarged role.Teams that are low in cognitive ability and experi-ence a functional to divisional shift may perform atthe same levels as divisional to functional teams,which should perform poorly regardless of theirlevels of cognitive ability. This argument leads to a

2004 685Moon, Hollenbeck, Humphrey, Ilgen, West, Ellis, and Porter

moderation hypothesis involving the average levelof cognitive ability and the nature of structuralrealignment. Specifically,

Hypothesis 4. A team’s composition and thenature of the structural adaptation moderateperformance: the positive effects of functionalto divisional transitioning will be stronger inteams that are high in general cognitive abilitythan they will be in teams that are low in thischaracteristic.

METHODS

Research Participants and Task

Research participants were 252 upper-level stu-dents, each of whom was in 1 of 63 four-personteams. All individuals were randomly assigned toteams, and all teams were randomly assigned toexperimental “conditions.” In return for their par-ticipation, each individual earned class credit aswell as eligibility for performance-based prizes.The task was the modified version of a simulationdeveloped for the Department of Defense, Distrib-uted Dynamic Decision Making (DDD; Miller,Young, Kleinman, & Serfaty, 1998). A depiction ofthe computer screen, which was divided into fourequal geographic quadrants, appears in Hollenbecket al. (2002). The major elements of the task wererelated to asset management and geography.

Asset management. Each individual participant(manager) controlled four vehicles (assets) thatcould be launched and then moved to any area onthe screen, including those areas monitored byother team members. These vehicles were semi-intelligent agents that could automatically performfunctions such as tracking, returning to base torefuel, launching, and so forth. There were fourtypes of vehicle (1) AWACS planes, (2) tanks, (3)helicopters, and (4) jets, and each vehicle had dif-ferent abilities in terms of range of vision, speed ofmovement, duration of operability, and weaponscapacity. The managers were tasked with usingtheir four assets to identify and properly engagevarious tracks.

There were 12 unique tracks, 3 of which wereconsidered friendly and 9 of which were consid-ered unfriendly. These tracks differed in variouscapabilities such as speed, requirements to disable,and ease of identification. For example, the trackidentified as “A5” was an unfriendly track thatmoved rapidly through the screen and could onlybe effectively engaged by the tank. A track identi-fied as “G0” was friendly, moved slowly, and couldbe engaged by any asset with weapons capability.Some tracks (labeled with a starting designation of

“U”) needed to be further identified via a trial anderror learning process. Hollenbeck et al. (2002) out-lined the specific abilities of all of the assets andtracks. Each team used its assets to engage an arrayof tracks on a single, networked simulation grid.

Geography. The simulation screen was parti-tioned in several ways using a grid/coordinate sys-tem. First, each quadrant (NW, NE, SW, SE) wasassigned to one member of each team. These quad-rants were further divided into three regions thatvaried in terms of the extent to which they neededto be protected from penetration by unfriendlyforces. The regions were labeled “neutral” (all areasoutside the restricted areas around the outmost pe-rimeter of the screen), “restricted” (a 12 by 12 gridin the center of the screen), and “highly restricted”(a 4 by 4 grid in the very center of the screen). Eachteam’s mission was to keep unfriendly tracks frommoving into the restricted and highly restrictedareas, while allowing friendly tracks to freely movein and out of the same areas.

All tracks originated from the edge of the screenand proceeded inward. Once a track came withinthe identification range of either the base or a ve-hicle, the manager had the opportunity to identifythe track. Once a track invaded a restricted zone,the team was required to expeditiously engage thetrack with the appropriate assets. The team wouldlose one point for every second the track resided inthe restricted zone and two points for every secondthe track was in the highly restricted zone (seeHollenbeck et al. [2002] for a full description of thecapabilities of all the tracks and vehicles). The teamwas also informed that it would be penalized forengaging any friendly track.

To monitor the geographic space, each managerhad a base (a small black box labeled “DM1-4” inthe center of the four quadrants) with a detectionring and an identification ring. If the track wasinside the detection ring (the outside ring), the in-dividual could see that something was there, but heor she could not determine the nature of the track(friendly or hostile) until it entered the more prox-imal identification ring (inside ring). Any track out-side the detection ring was invisible to the manag-ers, and therefore they had to rely on theirteammates to monitor regions of the space thatwere outside of their own quadrants. In sum, four-person teams were tasked with the management of16 total assets in a networked computer simulationwherein they addressed a wide array of tracks mov-ing on a partitioned geographic space.

Training. All individuals, and all teams, regard-less of experimental condition, received the sametraining. This considered of two separate modules.The first module (which lasted approximately 30

686 OctoberAcademy of Management Journal

minutes) introduced the participants to the simu-lation. The participants were then given a shortquestionnaire to ensure that they understood therudimentary operation of the simulation. The sec-ond module (which lasted approximately 60 min-utes) allowed the participants hands-on practice ofthe simulation, primarily to learn the basic mousemovements and operations.

Manipulations and Measures

Nature of structural adaptation: Functional todivisional versus divisional to functional. Struc-tural adaptation was manipulated between-teamsvia the task. In the functional structure, vehicleswere grouped to create narrow competencies inwhich each manager was in charge of four assets,all of the same type, taking on the role of tankcommand, helicopter command, jet command, orAWACS command. These roles had relatively lowtask scope because a manager was limited to theparticular capability of that one type of asset. Peo-ple working in these structures also were relativelylow in discretion, in the sense that once a trackbecame an issue, there might be one possible orlogical person to execute the task (for instance, if anA5 or G5 track appeared, the only person whocould successfully engage this track was the tankcommander). Finally, the fact that each managercould only perform a subset of all the tasks thatmight be required of a team meant that there was ahigh degree of interdependence in the functionalstructure.

In the divisional structure, vehicles weregrouped to create broad and versatile competen-cies. Each manager in the divisional structure alsomanaged four assets, but in this case, they were allof different types. Because of the four vehicles’complex array of strengths and weaknesses, oper-ating the four different vehicles created a job withrelatively high task scope. People working in thedivisional structures also were relatively high indiscretion, in the sense that once a track became anissue, any one of the four team members couldexecute the task, and there was no set demand forany one person to do any one task. Finally, the factthat each team member in a divisional structurecould manage any task him- or herself meant thatthere was a lower interdependence in the divisionalstructure (see Hollenbeck et al. [2002] for a graphicdisplay of the different team structures).

Each team operated under both structures, withthe order of their use manipulated. Half the teamsstarted out in the functional structure in stage 1,and then shifted into the divisional structure instage 2. These teams were in the functional to di-

visional condition. The other half of the teams,which used the divisional structure in stage 1, andthen shifted into the functional structure in stage 2,were in the divisional to functional condition). Tocapture the type of shift a team experienced, wedummy-coded a variable labeled “structure” (“di-visional to functional,” 1; “functional to divi-sional,” 2).

Task environment. Each team experienced twotypes of task environment. We varied the sequencein which the teams experienced the environmentsto control any order effects or effects for experi-ence. In the unpredictable task environment, a ran-dom number generator determined the entry andexit point of each track, and each track changedcourse once over the course of its life. In the pre-dictable task environment, each track proceeded ina diagonally straight line, originating in the NWquadrant and exiting in the SE quadrant. Thus, theorigin and exit points of tracks in the predictableenvironment were easy to anticipate.

Both task environments created their own uniquechallenges. The unpredictable environment waschallenging because of the uncertainty as to wheretracks were originating, but the predictable envi-ronment was also challenging because of the heavyconcentration of tracks flooding into a single area.Each stage contained 100 separate tracks and lastedroughly 30 minutes. Structural adaptation and taskenvironment were completely crossed, in a puretwo by two factorial design in which all the inde-pendent variables were orthogonal. Like structure,environment was dummy-coded (“random envi-ronment,” 1; “predictable environment,” 2).

Cognitive ability. Individual cognitive abilitywas measured via Form IV of the Wonderlic Person-nel Test (Wonderlic & Associates, 1983). We aggre-gated individual team members’ cognitive ability intoan average team level of cognitive ability.

Team coordination. Team coordination was cal-culated by using two computer-generated outputsof team behavior. Supportive behavior was mea-sured as the frequency with which one team mem-ber engaged enemy tracks in another team mem-ber’s zone of responsibility. For example, if themanager in the NW quadrant sent an asset into theSW quadrant to engage a track, this action countedas a coordination behavior, because it helped themanager in the SW quadrant clear his or her zone.Communication behavior was the frequency withwhich one team member transferred informationregarding the nature of a track to another teammember. That is, if the manager in the NW quad-rant learned that a certain track headed south wasfriendly, he or she could electronically share thisinformation with the SW manager. Doing this re-

2004 687Moon, Hollenbeck, Humphrey, Ilgen, West, Ellis, and Porter

quired some effort on the part of the NW manager,but it spared the SW manager from having to iden-tify the track him- or herself.

Team performance. Each team started the sim-ulation with 50,000 points and lost 1 point for eachsecond that any unfriendly track was in the re-stricted zone and 2 points per second for eachunfriendly track in the highly restricted zone. Theteams also lost 300 points for disabling any friendlytrack. These point calculations were the same thatHollenbeck and his coauthors (2002) used. Perfor-mance was measured twice, both when a team wasin stage 1 (and in its origin structure) and when theteam was in stage 2 (in its destination structure).Because all hypotheses were directional, we em-ployed one-tailed tests of statistical significance.

RESULTS

Table 1 presents the means, standard deviations,and correlations of all the variables measured ormanipulated in this experiment. Table 2 presentsthe results of our regression analysis of the differ-ence in stage 1 and stage 2 performance and theresults for performance at each stage regressed onstructure, task environment, and the interaction be-tween structure and task environment.

For the model shown in the first column of Table2, structure and environment were entered in thefirst two hierarchical steps, followed by their inter-action in a third step. This regression showed thatat stage 1, there were no “main effects” for structure(�R2 � .02 or environment �R2 � .01). Thus, in linewith structural contingency theory, no one struc-ture was best across environments.

Consistent with Hypothesis 1, however, the re-sults in the third step of this regression equationshow that the interaction between structure and theenvironment was statistically significant (�R2 �.08) and accounted for an appreciable amount ofvariance. Teams that were working in predictable

environments performed better when structuredfunctionally, whereas teams confronting unpredict-able environments fared better when structured di-visionally. This pattern of findings is a direct rep-lication of Hollenbeck et al. (2002), and it is justthis type of robust result that has led many tosuggest that organizations should change theirstructures in order to stay aligned with their envi-ronments.

The second column of Table 2 shows the resultsfor performance in stage 2, when each teamchanged its structure from functional to divisionalor from divisional to functional. We again saw astatistically significant interaction between struc-ture and environment (�R2 � .07), similar to whatwas discovered in stage 1, in that, all else beingequal, it was better for a team when its structurematched its environments in the manner pre-scribed by structural contingency theory. However,unlike what was found at stage 1 (where no onestructure was better), at stage 2, we saw a statisti-cally significant main effect for structure that ex-plained 6 percent of the variance in performance(�R2 � .06). Regardless of their environment, theteams that changed from a functional to a divi-sional structure (x� � 33,797) tended to outperformthose that changed from a divisional to a functionalstructure (x� � 31,815).

We tested the robustness of this finding using amore rigorous test. Because the ultimate dependentvariable of interest was a difference score reflectingease or difficulty in switching from one structure toanother, we used the regression decompositiontechniques described by Edwards (1995). That is, inaddition to the stage 1 and stage 2 performancescores, we also regressed the change scores be-tween stage 1 and stage 2. These results are re-ported in the third column of Table 2. We foundthat only structure was significantly related to thedifference in performance scores between the twostages (�R2 � .11). Neither the main effect for en-

TABLE 1Descriptive Statistics and Correlationsa

Variable Mean s.d. 1 2 3 4 5 6 7 8

1. Structure 1.51 0.502. Environment 1.55 0.50 .013. Cognitive ability 24.76 2.56 .00 �.064. Average support 26.03 6.89 �.01 �.17 .135. Average communication 60.00 11.90 �.12 �.19 .27* .25*6. Performance, stage 1 27,556 3,984 �.14 .09 �.01 .21 .197. Performance, stage 2 32,822 4,081 .25* �.07 .02 �.09 �.13 .31*8. Change, stage 1 � stage 2 �5,266 4,734 �.32* .14 �.03 .25* .27* .57* �.60*

a n � 63.* p � .05

688 OctoberAcademy of Management Journal

vironment (�R2 � .02) nor the structure by envi-ronment interaction (�R2 � .00) was significant.Thus, there was asymmetrical adaptation, support-ing Hypothesis 2.

Hypothesis 3 predicts that team coordinationwill mediate the relationship between structureand team performance at stage 2. Testing mediationis a multistep process (Baron & Kenny, 1986). Me-diation is supported when the independent vari-able being tested significantly relates to the medi-ator. Here, both team communication (�R2 � .30)and team coordination (�R2 � .10) were positivelyrelated to stage 2 structure. Further, the indepen-dent variable must significantly relate to the depen-dent variable in the absence of the mediator; thisrelationship was shown in our test of Hypothesis 2.Finally, the influence of the independent variableon the dependent variable must shrink upon theaddition of the mediator to the model. When teamcoordination and team communication were con-trolled, the main effect for structure at stage 2 wasno longer significant, with the effect size (�R2)dropping from 6 to 0 percent of variance explained.

Table 3 gives the results of our test for mediation,with the statistics in the second column of Table 3demonstrating the influence that controlling forteam coordinating activities had on the relation-ship between structure and performance at stage 2.Thus, we found support for Hypothesis 3.

Finally, Hypothesis 4 predicts that team cogni-tive ability will interact with type of structuralchange, with low cognitive ability attenuating theotherwise beneficial effects of the functional to di-visional transition. Table 4 reports the results of ahierarchical regression analysis testing this moder-ation hypothesis (Stone & Hollenbeck, 1984). Look-ing at the results for stage 2 performance, we find asignificant interaction between a team’s level ofcognitive ability and the type of change the teamengaged in (�R2 � .07). Figure 2 illustrates theinteraction between cognitive ability and type ofchange, with the relationship between cognitiveability scores and performance plotted separatelyfor teams in functional to divisional and divisionalto functional structures. As predicted in Hypothe-sis 4, teams high in cognitive ability responded

TABLE 2Results of Hierarchical Regression Analysisa

Independent Variable

Stage 1 Stage 2 Difference

� �R2 � �R2 � �R2

Structure �1,066 .02 1,982* .06* �3,048* .11*Environment 738 .01 �592 .00 1,330 .02Structure � environment �4,346* .08* �4,415* .07* 68 .00

Total R2 .11* .13* .13*

a n � 63 teams.* p � .05

TABLE 3Mediation of Structure by Team Coordination

Independent Variable

Stage 1 Stage 2 Difference

� �R2 � �R2 � �R2

Help 1,392 .08* �209 .15* 253 22**Communication �497 �1,031 1,485Structure �1,418 .03 401 .00 890 .01Environment 3,539 .08* �1,583 .02 �1,277 .01Structure � environment �7,148 .18* �4,464 .07* �2,462 .01

Total R2 .37* .24* .21*

n � 63 teams.* p � .05

** p � .01

2004 689Moon, Hollenbeck, Humphrey, Ilgen, West, Ellis, and Porter

better to the functional to divisional transition thanteams that were low; however, the high-cognitive-ability teams also responded worse in the divi-sional to functional shift, which was not techni-cally a formal prediction of Hypothesis 4.

DISCUSSION

The field of organizational behavior is dominatedby many different contingency theories whose gen-eral import is that there is no one best way to solveall organizational problems. Instead, contingencytheories imply that people or organizations, in or-der to sustain excellence over time (Pulakos, Arad,Donovan, & Plamondon, 2000), need to engage inone set of behaviors when confronted with one setof conditions, but engage in a different set of be-haviors under an alternative set of conditions(Smith & Nichols, 1981).

Theoretical Implications

Asymmetric adaptability and team structure.We introduced the concept of asymmetric adapt-ability to capture the idea that adaptation can onlybe fully understood by directly analyzing the pointof origin and the destination point associated withspecific types of changes. Using structural contin-gency theory as an exemplar, we replicated thecommon, cross-sectional contingency finding atstage 1 that indicated that functional structures per-form better in predictable environments, whereasdivisional structures perform better in unpredict-able environments. The dynamic implication ofthis finding is that if a team’s task environmentchanges, to stay in a fit with its environment, theteam should change structures.

Unlike most tests of contingency theories, in thisexperiment, we tested directly to see if teams could

FIGURE 1Effects of Cognitive Ability and Structural Shift on Performance at Stage 2

TABLE 4Moderation Effects of Team Cognitive Ability on Structure

Independent Variable

Stage 1 Stage 2 Difference

� �R2 � �R2 � �R2

Structure �1,066 .02 1,982* .06* �3,048* .11*Environment 738 .01 �592 .00 1,330 .02Structure � environment �4,346* .08* �4,415* .07* 68 .00Average cognitive ability 34 .00 69 .00 �35 .00Average cognitive ability � structure 540 .03 856 .07* �315 .01

Total R2 .14* .20* .14*

n � 63 teams.* p � .05

690 OctoberAcademy of Management Journal

actually adapt in the manner implied by the theory.Although both divisional to functional and func-tional to divisional shifts can be considered “adap-tive” under the right circumstances, in fact, one ofthese adaptations was much easier to negotiatethan the other. Specifically, it was much more nat-ural for teams to shift from a functional to a divi-sional structure than it was for them to switch inthe other direction.

We speculated that the functional to divisionaladaptation was easier to make because the normsthat are established in the first stage of a functionalto divisional transition support performance in thesecond stage, whereas the norms established in thefirst stage of a divisional to functional transition arecounterproductive for stage 2 performance. The evi-dence supported this conjecture, in the sense that ifcoordination behaviors—that is, the frequency of mu-tual support and communication—were controlledfor, the superiority of the functional to divisionalteams to the divisional to functional teams at stage 2disappeared. Although this experiment was certainlynot a direct test of entrainment theory, we do supportthe idea embodied in that theory that norms can oftenpersist over time, even when changes in a group’sstructure or task environment make them obsolete(Bettenhausen & Murnighan, 1985).

Third, with respect to group composition issues,we also found evidence that team composition mod-erated the effect of structural change. The averagelevel of general cognitive ability in a team amplifiedstructural change differences, in that teams that werehigh on this characteristic responded better to changethat enhanced the amount of task scope and auton-omy than did teams that were low in ability. Teamswith high cognitive ability also showed a much morepronounced negative response when task scope andautonomy were reduced.

We did not formally predict that groups that werehigh in cognitive ability would ever do worse thanlow-ability teams, and hence the negative relation-ship between cognitive ability and performance fordivisional to functional teams was unanticipated.This unexpected finding, at first blush, challengesresearch (Hunter & Hunter, 1984) that has shownpositive relationships between cognitive abilityand a wide array of tasks. However, the idea thatcognitive ability could be a liability for simple tasksis not unprecedented. In fact, on September 2,1999, the U.S. district court of Connecticut decidedthat the city of New London had a right not to hirean applicant for the job of police officer if his/hercognitive ability was too high. The city contended(Allen, 1999) that the candidates of high abilitywould become bored during the course of their joband would not react well to the rigid procedures

associated with the position. Our findings and theNew London court case are consistent with re-search (Johnson & Johnson, 2000) that has demon-strated a positive link between perceptions of over-qualification and job dissatisfaction. An interestingimplication of our findings regarding cognitiveability is that one needs to consider not only the fitof a team’s structure to its environment, but also thefit between a team’s structure and the characteris-tics of its members.

Practical Implications

The results of this study suggest several impor-tant considerations for those who are either design-ing team structures or thinking about changingteams from one structure to another. First, in termsof designing an initial structure, since it is appar-ently easier to adjust a structure in the functional todivisional direction than vice versa, managers maywant to establish initial structures that err on theside of being too functional. If subsequent adjust-ments in a divisional direction are needed, thesewill be easier to manage than those that go in theopposite direction. Thus, team designers shouldinitially show a bias in favor of functional struc-tures or structures that lean in that direction. Initialerrors that require a divisional-functional readjust-ment will be more difficult to overcome and have tobe avoided at all costs.

Second, for managers who are consideringchanging a team’s structure to match an environ-mental change, the criterion for triggering thechange may need to be set at different levels de-pending upon the direction of the change. If a teamstarts out in a functional structure, even a smallamount of evidence that the structure needs to beadjusted (for instance, the team appears to be lack-ing sufficient flexibility) may be enough to trigger aadaptation toward divisional structure. On theother hand, if a team starts out in a divisionalconfiguration, the burden of evidence needed toinitiate change in the functional direction (for in-stance, the team appears to lack sufficient effi-ciency) may need to be set much higher. Sinceexecuting this latter reconfiguration is more diffi-cult, the value of structural adaptation would needto be higher to offset the higher transition costsassociated with this type of restructuring.

Third, if the evidence strongly suggests that ateam needs to reconfigure from a divisional struc-ture to a functional structure, managers need tosupply external supports in order to ease this tran-sition. Supports might include training programsaimed at increasing the amount of communicationbetween team members, reward systems (team-

2004 691Moon, Hollenbeck, Humphrey, Ilgen, West, Ellis, and Porter

based bonuses) that stress that value of collabora-tive behavior, or goal setting–feedback systems thatexplicitly monitor and manage communication andhelping behaviors among team members. Leadersof teams making this type of transition may need toassume the role of communication and support of-ficer (or directly assign such a role to a team mem-ber) to insure that some individual is responsiblefor assuring that the types of difficulties that teamsencounter when making adjustments in the func-tional direction are addressed.

Finally, in regards to our moderation findingsrelated to cognitive ability, a “war for talent” iscurrently being waged among organizations as theypursue the most talented workers. Trank, Rynes,and Bretz (2002) recently found that individualswith high ability prefer organizations that are morechallenging and selective. This finding is consis-tent with Ganzach’s (1998) finding that job com-plexity influenced the link between intelligenceand job satisfaction. Our findings regarding cogni-tive ability establish the converse of the need forchallenge among teams of high ability, in that theydemonstrated a negative reaction to loss of chal-lenge. Therefore, managers must not only attracttop talent by increasing challenge and complexity,but also guard against repulsing top talent with alack of challenge and complexity.

Asymmetric Adaptability in Alternative Domains

Although it was beyond the scope of this oneexperiment to establish asymmetric adaptabilitythroughout all contingency theories, we think itinteresting to speculate on how various contin-gency theories might stand up to the same kind oftest that was applied here to structural contingencytheory. For example, the Vroom-Yetton (1973)model of leadership is a contingency theory accord-ing to which the decision-making process that aleader uses should depend upon characteristics ofher or his followers and situation. One decisionprocess recommended by this model in one set ofcircumstances is “GII,” in which the leader sharesthe problem with subordinates, and together theygenerate and evaluate alternatives. The goal wouldbe to work slowly and attempt to reach consensuson a solution. The leader serves as chairperson,coordinating the discussion and keeping it focusedon the problem. The leader makes sure the criticalissues are discussed but does not try to influencethe group to adopt his or her solution. In the end,the leader needs to be willing to accept and imple-ment any solution that has support from the group.

Alternatively, within a different set of circum-stances, the model might recommend that the same

leader use a process labeled “AI,” in which theleader solves the problem himself or herself, usingpersonal information and not involving the subor-dinates in any way. Although accepting the staticlogic that might lead this theory to recommendsuch different styles under different circumstances,the concept of asymmetrical adaptability makesone question the dynamic logic involved when, ina real operational setting, the social system tries tomove from one state to the other.

Specifically, if the original circumstances dictatea series of initial decisions in which the AI style isappropriate and executed, but then circumstanceschange, it may be quite natural for the group toadapt from an AI to a GII style because the groupmembers’ roles and influence are expanded in thenew, adapted situation. However, if the originalcircumstances allow for a long series of GII styledecisions, but then circumstances change—de-manding the leader adopt a new AI style—will thisshift be as easy to execute as the other? If groupmembers are asked to sacrifice influence and dis-cretion, will their reactions to the AI style be thesame as the reactions of those who experiencedonly the AI style at stage 1 and have never experi-enced GII? There may very well be asymmetricaladaptability in this situation, in which it is easierfor a team to evolve from a series of AI to GIIdecision rules than it is for the same team to adaptfrom a series of GII to AI decision rules.

Another contingency theory that it might be in-structive to analyze in this same manner is socialinterdependence theory (Deutsch, 1949). Social in-terdependence theory contrasts competitive andcooperative reward allocations and, like all contin-gency theories, suggests that there is no one bestway to allocate awards. According to this theory, iftask interdependence among team members is low,then an organization should employ competitiverewards that pit members against each other. How-ever, when team members’ task demands high lev-els of interdependence among them, the organiza-tion should employ cooperative rewards, in whichall team members experience the same outcomeregardless of their individual contribution. Thegeneral, static logic underlying this theory is thatcooperative rewards promote collaboration andteamwork under conditions of high interdepen-dence, whereas competitive rewards prevent socialloafing among group members when interdepen-dence requirements are low.

Again, one can accept the static logic underlyingthis theory and at the same time question the dy-namic logic if the implication is that a group shouldchange from one reward allocation structure to an-other if the level of task interdependence changes.

692 OctoberAcademy of Management Journal

For example, a group might originate in a situation oflow task interdependence and work under a compet-itive reward structure in which every person is look-ing out for himself or herself. Over the course of time,this will affect the behaviors and interpersonal rela-tionships of these people in a particular direction. Ifthe level of task interdependence changes, however,and this group has to adapt to a new, cooperativereward structure, it is not clear that the level and typeof cooperative behavior that its members would showwould in any way resemble what would be seen in agroup working under a cooperative structure withoutever having experienced the competitive structure.Indeed, one might see “cutthroat cooperation” in theformer group that would differ dramatically from theheartfelt cooperation seen in the group that had neverexperienced competition in this work context.

Alternatively, a group that started its work undera cooperative reward allocation structure might ex-perience less disruption if asked to switch to a new,competitive structure. Having established support-ive relationships in its early days (stage 1), compe-tition in this group at stage 2 might be “friendly”and might not create the difficulties that might beexperienced in the “cutthroat cooperation” group.This pattern would again imply that adaptability isasymmetrical, in the sense that it is more naturalfor a group to evolve from cooperation to competi-tion than it is for it to go from competition tocooperation. There is potential for research toblend asymmetric compensation strategies withmoderators such as trust and individual differencesamong team members.

Limitations

We realize that what we present is an initialexperiment of a new concept tested in a laboratorysetting. Although there are clear limits to what canbe accomplished in laboratory settings, one needsto keep the nature of the research question in mindwhen assessing the relevance of external validity(Berkowitz & Donnerstein, 1982). Since no formalaspect of this theory implies it would not work inthis specific context, this context provided a legit-imate venue within which to test the theory. AsIlgen (1986) noted, this is precisely the type ofquestion that is well suited to laboratory contexts.

Future Research

Although this experiment called into questionthe desirability of divisional to functional shifts inteam structure, it did not explore all the possiblealternatives short of a total divisional to functionalshift that might be available to these teams. Thus,

even though structural adaptation may be a one-way street, as always with navigation, there is morethan one way to get from point A to point B. Forexample, in the experiment we conducted, whenthe environment became more predictable, it mighthave been better for the teams we had switch fromdivisional to functional to go to something lessextreme than a strictly functional structure. That is,operationally speaking, instead of shifting from asituation in which they controlled all four types ofvehicles to a situation in which they controlledonly a single vehicle, the divisional to functionalteams may have been better off shifting to a “com-promise structure” that allowed them control overtwo vehicles rather than just one. This changewould have been a shift in the functional direction,but not a total shift to a functional structure. Sucha compromise structure might make a better desti-nation for such teams, given their point of origin.

Alternatively, this structure could serve as a“transition structure” as a team moved from divi-sional to functional in two or three steps ratherthan one. Within organizations, attention towardthese hybrid structure shifts might mitigate some ofthe problems inherent in change. In addition, cer-tain types of “robust” organizational structures mayexist that, although not optimal for any one envi-ronment, will be serviceable in a number of differ-ent environments, thus obviating the need tochange structures. Future research needs to exam-ine whether these types of “compromise struc-tures,” “transition structures,” or “robust struc-tures” can ameliorate some of the problemsmanifested in teams that are trying to navigate adivisional-functional adaptation.

Finally, research might also explore “proceduralworkarounds,” that is, adaptations in processesthat may substitute for structural changes—espe-cially changes in the divisional to functional direc-tion. Thus, instead of changing a structure fromdivisional to functional, one might be able tochange a team’s work processes in ways that en-hance efficiency within the existing structure. Forexample, competitive reward structures have beenfound to increase the speed of operations in teamssimilar to those studies here (Beersma et al., 2003).Since increased processing speed enhances effi-ciency, such a use of competitive reward structurecould serve as an alternative to changing teamstructure. Future research needs to examine thisand other types of process adaptations that mightserve as alternatives to structural changes in teamsthat are poorly matched with their environments.

The focus of this paper was on structural contin-gency theory, and although any one of a number ofother contingency theories could be analyzed in a

2004 693Moon, Hollenbeck, Humphrey, Ilgen, West, Ellis, and Porter

similar manner, the overall point that should bedrawn from this discussion should be clear. Thestatic logic that provides the basis for many contin-gency theories needs to be complemented by a dy-namic logic. This dynamic logic needs to addresswhether the changes that are dictated by sequentialapplications of different behavioral routines lead tothe type of asymmetrical adaptability that was doc-umented in this experiment. It is a myth to thinkthat individuals, teams, and organizations can beinfinitely flexible, and hence future research needsto document other asymmetries, as well as inter-ventions to offset them.

REFERENCES

Allen, M. 1999. Help wanted: The not-too-high-Q stan-dard. New York Times, September 19: wk 3.

Allred, B. B., Snow, C. C., & Miles, R. E. 1996. Charac-teristics of managerial careers in the 21st century.Academy of Management Executive, (10)4: 17–27.

Ancona, D., & Chong, C. 1996. Entrainment: Pace, cycle,and rhythm in organizational behavior. In L. L. Cum-mings & B. M. Staw (Eds.), Research in organiza-tional behavior, vol. 18: 251–284. Greenwich, CT:JAI Press.

Balkin, D. B., & Gomez-Mejia, L. R. 1987. Toward a con-tingency theory of compensation strategy. StrategicManagement Journal, 8: 169–182.

Baron, R. M., & Kenny, D. A. 1986. The moderator-medi-ator variable distinction in social psychological re-search: Conceptual, strategic, and statistical consid-erations. Journal of Personality and SocialPsychology, 51: 1173–1182.

Beersma, B., Hollenbeck, J. R., Humphrey, S. E., Moon,H., Conlon, D. E., & Ilgen, D. R. 2003. Cooperationcompetition and team performance: Toward a con-tingency approach. Academy of Management Jour-nal, 46: 572–590.

Berkowitz, L., & Donnerstein, E. 1982. External validity ismore than skin deep. American Psychologist, 37:245–257.

Bettenhausen, K., & Murnighan, J. K. 1985. The emer-gence of norms in competitive decision-makinggroups. Administrative Science Quarterly, 30: 350–372.

Burns, T., & Stalker, G. M. 1961. The management ofinnovation. London: Tavistock.

Conlon, D. E., Moon, H., & Ng, K. Y. 2002. Putting the cartbefore the horse: The benefits of arbitrating before me-diating. Journal of Applied Psychology, 87: 978–984.

Deutsch, M. 1949. A theory of cooperation and competi-tion. Human Relations, 2: 129–152.

Dill, W. R. 1958. Environment as an influence on mana-gerial autonomy. Administrative Science Quar-terly, 2: 409–443.

DiMaggio, P. J., & Powell, W. W. 1983. The iron cagerevisited: Institutional isomorphism and collectiverationality in organizational fields. American Socio-logical Review, 48: 147–160.

Drazin, R., & Van de Ven, A. H. 1985. Alternative forms offit in contingency theory. Administrative ScienceQuarterly, 30: 514–539.

Eden, D. 2002. Replication, meta-analysis, scientificprogress, and AMJ’s publication policy. Academy ofManagement Journal, 45: 841–846.

Edwards, J. R. 1995. Alternatives to difference scores asdependent variables in the experiment of congruencein organizational research. Organizational Behaviorand Human Decision Processes, 64: 307–324.

Ganzach, Y. 1998. Intelligence and job satisfaction.Academy of Management Journal, 41: 526–539.

Hackman, J. R., & Oldham, G. 1976. Motivation throughthe design of work: Test of a theory. OrganizationalBehavior and Human Performance, 16: 250–279.

Hambrick, D. C. 1983. High profit strategies in maturecapital goods industries: A contingency approach.Academy of Management Journal, 26: 687–707.

Hollenbeck, J. R., Moon, H., Ellis, A., West, B., Ilgen,D. R., Sheppard, L., Porter, C. O. L. H., & Wagner,J. A. 2002. Structural contingency theory and indi-vidual differences: Examination of external and in-ternal person-team fit. Journal of Applied Psychol-ogy, 87: 599–606.

House, R. J. 1971. A path-goal theory of leadership effec-tiveness. Administrative Science Quarterly, 16:321–338.

Hunter, J. E., & Hunter, R. F. 1984. Validity and utility ofalternative predictors of job performance. Psycho-logical Bulletin, 96: 72–98.

Ilgen, D. R. 1986. Laboratory research: A question ofwhen, not if. In E. A. Locke (Ed.), Generalizing fromlab to field settings: 257–267. Lexington, MA:Heath.

Ilgen, D. R. 1999. Teams embedded in organizations:Some implications. American Psychologist, 54:129–139.

Johnson, G. J., & Johnson, W. R. 2000. Perceived over-qualification and dimensions of job satisfaction: Alongitudinal analysis. Journal of Psychology, 134:537–555.

Lawrence, P. R., & Lorsch, J. W. 1967. Organization andenvironment. Boston: Harvard University Press.

Levitt, R. E., Thomsen, J., Christiansen, T. R., Kunz, J. C.,Yan, J., & Nass, C. 1999. Simulating project workprocesses and organizations: Toward a micro-contin-gency theory of organizational design. ManagementScience, 45: 1479–1495.

Miller, D. 1988. Relating Porter’s business strategies toenvironment and structure: Analysis and perfor-

694 OctoberAcademy of Management Journal

mance implications. Academy of ManagementJournal, 31: 280–308.

Miller, D., & Friesen, P. H. 1983. Strategy-making andenvironment: The third link. Strategic Manage-ment Journal, 4: 221–235.

Miller, D. L., Young, P., Kleinman, D., & Serfaty, D. 1998.Distributed dynamic decision-making simulationPhase I Release notes and user’s manual. Woburn,MA: Aptima.

Miner, J. B. 1984. The validity and usefulness of theoriesin an emerging organizational science. Academy ofManagement Review, 9: 296–306.

Pulakos, E. D., Arad, S., Donovan, M. A., & Plamondon,K. E. 2000. Adaptability in the workplace: Develop-ment of a taxonomy of adaptive performance. Jour-nal of Applied Psychology, 85: 612–624.

Pennings, J. M. 1992. Structural contingency theory: Areappraisal. In L. L. Cummings & B. M. Staw (Eds.),Research in organizational behavior, vol. 14: 267–309. Greenwich, CT: JAI Press.

Shaw, M. E. 1976. Group dynamics: The psychology ofgroup behavior. New York: McGraw-Hill.

Smith, D., & Nichols, T. 1981. Change, standardizationand contingency theory. Journal of ManagementStudies, 18: 73–88.

Stone, E. F., & Hollenbeck, J. R. 1984. Some issues asso-ciated with the use of moderated regression. Orga-nizational Behavior and Human Performance, 34:195–213.

Thomas, K. W., & Velthouse, B. A. 1990. Cognitive ele-ments of empowerment: An “interpretive” model ofintrinsic motivation. Academy of Management Re-view, 15: 666–681.

Townsend, A. M., DeMarie, S. M., & Hendrickson, A. R.1998. Virtual teams: Technology and the workplaceof the future. Academy of Management Executive,12(3): 17–29.

Trank, C. Q., Rynes, S. L., & Bretz, R. D. 2002. Attractingapplicants in the war for talent: Differences in workpreference among high achievers. Journal of Busi-ness and Psychology, 16: 331–345.

Van Maanen, J., & Schein, E. H. 1979. Toward a theory oforganizational socialization. In B. M. Staw (Ed.), Re-search in organizational behavior, vol. 1: 209–264.Greenwich, CT: JAI Press.

Vroom, V. H., & Yetton, P. W. 1973. Leadership anddecision-making. Pittsburgh: University of Pitts-burgh Press.

Wagner, J. A. 2000. Organizations. In A. E. Kazdin (Ed.),Encyclopedia of psychology, vol. 6: 14–20. NewYork and Washington DC: Oxford University Pressand American Psychological Association.

Wonderlic & Associates. 1983. Wonderlic personnel testmanual. Northfield, IL: Wonderlic & Associates.

Henry Moon ([email protected]) is an assis-tant professor of management at the Goizueta BusinessSchool at Emory University. He received his Ph.D. inmanagement from Michigan State University. He enjoysresearching organizational behavior.

John R. Hollenbeck received his Ph.D. in managementfrom New York University in 1984, and he is currentlythe Eli Broad Professor of Management at the Eli BroadGraduate School of Business Administration at MichiganState University. Dr. Hollenbeck has published over 60articles and book chapters on the topics of team decisionmaking and work motivation.

Stephen E. Humphrey is a doctoral candidate in organi-zational behavior at Michigan State University. His cur-rent research interests include virtual teams, brokeredultimatum games, and the relationship between time andbehavior.

Daniel R. Ilgen is the John A. Hannah Professor of Psy-chology and Management at Michigan State University.Professor Ilgen received his M.A. and Ph.D. degrees fromthe University of Illinois, Urbana-Champaign. His researchinterests are in work motivation and team behavior.

Bradley West is a doctoral candidate in industrial andorganizational psychology at Michigan State University.His current research interests include team development,team learning and decision making, employee attitudes,diversity, and organizational justice.

Aleksander P. J. Ellis ([email protected]) earned hisPh.D. at Michigan State University in industrial/organi-zational psychology. He is an assistant professor in theDepartment of Management and Policy at the Eller Col-lege of Business and Public Administration. His currentresearch focuses on team processes such as team learn-ing, back-up behavior, transactive memory, and sharedmental models. He is interested in determining whethercertain situational and personal characteristics such astraining, stress, and personality have the potential tobolster or damage the efficiency and effectiveness of ateam’s collective information-processing system.

Christopher O. L. H. Porter ([email protected]) is an as-sistant professor of management at Mays Business Schoolat Texas A&M University. He earned his Ph.D. in busi-ness administration at Michigan State University. Hisresearch interests focus on team composition and team-work, performance appraisals and feedback interven-tions, and police performance management.

2004 695Moon, Hollenbeck, Humphrey, Ilgen, West, Ellis, and Porter