The Performance and Development of Teams
The focus of this (and the next) chapter is on teams, which can be regarded as a type of small group. Although the terms "team" and "group" are often used interchangeably (even within the same study), it is useful to distinguish between these concepts to provide the boundaries for the review in this chapter. Our implicit working definition of teams is not so broad as to encompass all kinds of small groups, nor so narrow as to exclude important insights from the literature on groups.
In distinguishing between groups and teams, Hare (1992) notes that "group" is the more general term and refers to a set of individuals who have some common characteristicwithout actually interacting with one another. "Team" is more specific term: joint action is implied, with sports teams being a very visible example. Dyer (1987:24-25) defines a team as "a collection of people who must collaborate, to some degree, to achieve common goals.''
Hare suggests that teams can be placed along a continuum according to the amount of collaborationintegration and role differentiationrequired. In sports, golf would be considered low in both integration and differentiation, football high in both, with track (high differentiation, low integration) and synchronized swimming (low differentiation, high integration) high in one and low in the other (Hare, 1992:Fig. 2.2). Concluding that a team is more than a collection of individuals, Francis and Young (1979:6-7) describe an effective team as one that "combines high morale, effective task performance, and clear relevance to the organization. Bassin's (1988) requirements for high performance teams include vision (a shared purpose), perceived dependent needs, leadership, coordination, and the skillful use of feedback to adjust and adapt to changes. These criteria are similar to those suggested in other writings on team building (see review in Hare, 1992). Bassin
classifies teams by such functional characteristics as discovery driven (research and development teams), rule driven (sports teams), product driven (business teams), and technology driven (plane crews).
Freeberg and Rock (1987) define a team in terms of such distinct features as a goal or mission orientation, formality of structure, a requirement for member interaction stemming from task interdependence, and the assignment of special roles to members. Citing work done by Hall and Rizzo (1975), Freeberg and Rock (1987:5) conclude that the "hallmarks of team interaction processes appear to draw upon the behavioral dimensions of collaboration, coordination, and communication" (see also Rizzo, 1980). Within these broad features, there are variations in emphasis from one study to another. One important difference is whether the team interacts with machines or people. Another is the extent of task interdependence in problem-solving situations. Teams also vary in terms of the rigidity of their structures. Contrast, for example, mission-oriented military teams with research teams: the former are likely to exhibit a higher degree of member specialization and coordination in task performance, as well as a clearer designation of positions or assignments, than the latter. Concentrating on work teams, Sundstrom et al. (1990:120) offer the definition as "small groups of interdependent individuals who share responsibility for outcomes of their organizations." Their analysis examines teams that function in the context of organizations.
Team performance is defined usually in terms of outcomes, although process measures are sometimes included. In addition, most of the experimental investigations have construed team performance in linear ("inputoutput") terms, although some recent work, largely case studies, provides a basis for reconceptualizing the process in nonlinear terms. Furthermore, the production process may be more complex. For example, Hackman (1987) suggested that teams evaluate their collective performance as they work, and evaluations affect team processes, which influence subsequent performanceas self-reinforcing spirals of increasing or decreasing effectiveness (see also Hall and Watson, 1971; Steiner, 1972).
This chapter reviews the work from both linear and nonlinear views of team performance. We cover both the internal interactive processes emphasized by Freeberg and Rock (1987), and the external (contextual and boundary variables) influences highlighted by the Sundstrom et al. (1990) framework. We begin with what is known about team performance from laboratory research, then turn to the broader frameworks for understanding performance, including contextual influences. In the main part of the chapter, we examine team developmental processes, addressing issues about how teams learn and the effectiveness of team-building interventions. Finally, issues of performance and development are considered in relation to the use of games as vehicles for training. (Related issues of team training are treated in the next chapter.)
DETERMINANTS OF TEAM PERFORMANCE
A Meta-Analysis of Laboratory Studies
Freeberg and Rock (1987) conducted a meta-analysis of many studies dealing with input, mediating (throughput), and output (performance outcome) variables in team research. Emphasizing internal team processes, they organized their meta-analysis in terms of 12 dependent outcome and process categories:
1. team accuracy
2. time required to achieve a solution
3. amount of product produced
4. extent of task transfer
5. solution agreement among members
6. originality of solutions
7. perceived satisfaction
8. trials to acquisition
9. performance proficiency
10. team cohesiveness
11. coordination among members
12. suitability of interaction or communication among members.
There were 25 independent variables used in the meta-analysis, grouped in the broad categories of team member characteristics (e.g., sex, prior experience with task, time worked together), team task characteristics (e.g., complexity, task fidelity, feedback of results), and team organization variables (e.g., communications structure, assigned roles and role stratification, coordination of member functions). Studies chosen for the analysis were those that dealt exclusively with task-interacting groups in which "team" was a unitary entity (also referred to as an ad hoc laboratory team). Most were laboratory studies (79 percent) conducted with college student populations; the average team size was 3.2 members, the average sample size (number of teams in the analysis) was 39. However, due to the strict statistical requirements that must be met to perform the analysis, only 21 percent of the available literature (117 of 547 papers) could be used in the meta-analysis. The strongest effect sizes 1 were obtained for studies of highly rated laboratory (rather than field) studies, studies with smaller team sizes, and those rated highest in data reporting.
The strongest relationships were obtained for three outcome measures: accuracy of team task performance, time required to achieve a solution to
the task, and quantity of the team product. The seven strongest independent variables, by rank, were task complexity, task structure, performance over time (practice), interaction/communication, task load, cooperation /competition, and coordination. (These relationships are depicted in Figure 6-1.) Negligible effect sizes were obtained for prior task experience, team members' individual skills, feedback, and cohesiveness. With regard to those variables most relevant to trainingprior task experience, practice, and feedbackpractice showed the strongest effect sizes in relationships with quantity of output and accuracy of product. Moderate effect sizes were obtained for feedback with solution time and interaction/communication.
A major contribution of this analysis is the development of three "minimodels" in which the authors connect input to output variables through each of three mediating variables, interaction/communication, coordination, and cohesiveness. Each model highlights one of these mediating variables, showing how relationships between particular inputs and outputs may depend on the way they influence that process. Model 1 indicates that the mediating influence of communication results from the effects of three heavily taskdependent team conditions: load, feedback, and practice. These inputs influence the outputs of time, proficiency, and accuracy through a communication process. Of the various relationships specified by this model, the relationship between feedback and performance accuracy has received the least attention in the research literature. Model 2 highlights the importance of cooperation/competition as an influence on team output, both in terms of their direct effects on team accuracy and their indirect effects through the mediating variable of coordination. The model also calls attention to the
importance of task fidelity (comparability to real-world tasks) as an influence on outputs. Both cooperation/competition and fidelity affect team coordination which, in turn, affects the accuracy and quantity of team output. Both are central to the committee's appraisal of the team literature. Model 3 further underscores the importance of cooperation/competition: in this model they operate through team cohesiveness to produce effects on the outcome variables of quantity and proficiency. These results suggest that intrateam conflict has a strong effect on team performance. They provide a basis for recommending research into the way that conflict is manifest in the team process, for example, by probing relationships between types of intrateam conflict and modes of settlement or resolution. More generally, the analysis shows that team variables exert stronger effects on outputs than do member characteristics.
The Freeberg and Rock (1987) meta-analysis provides an important cumulation function for research on team performance. Similar to metaanalyses performed on other topics, this work distinguishes among variables that have relatively strong and weak influences on team outputs across many studies. The authors also distinguished among variables that had direct versus indirect effects on outputs. For example, in one model, feedback was found to relate to interaction or communication which, in turn, was linked to several performance outcomes. In other words, knowledge of results is likely to exercise some of its effects through the patterns of member communication established within a team. Further specification is achieved when the same variablesfor example, task load and practiceare shown to have strong indirect effects on such output variables as time and direct effects on such other measures of team performance as quality of product. Identifying possible causal paths through which variables operate to produce effects is a major addition to the direct bivariate (two-variable) relationships uncovered in the meta-analysis. They highlight the importance of team process and, by so doing, uncover effects of some input variables shown to have relatively weak direct effects on team output (e.g., feedback, cooperation and competition, task fidelity). They also make research gaps evident and, therefore, provide an analytical basis for further research on the topic. One direction for further research is to trace the linkages suggested by the models over time in different group tasks.
The Freeberg and Rock (1987) analysis is limited by the nature of the sample of studies included. Those studies may not be representative of the universe of studies on team performance. The sample is small and biased toward laboratory experiments, although it is of high quality in terms of methodological criteria. It may well be that these studies represent the only body of work that merits the sort of cumulative analysis performed by the authors. Related to the issue of biased sampling is the problem of overlooking certain variables rarely included in laboratory exercises. One set of
variables, referred to as "context," consists of aspects of larger organizations within which teams function, including relations among different teams operating within the same or different organizational environments. Another set of variables are those that refer to team-building and identification processes that occur over a longer period of time than is available for laboratory experiments. These processes have been examined primarily in field studies of particular cases. Both types of variables are discussed in the sections to follow.
Broader frameworks for analyzing factors that influence team performance include contextual variables. Emphasizing the context within which teams perform, Sundstrom et al. (1990:122) define performance as the "acceptability of output to customers within or outside the organization who receive team products, services, information, decisions, or performance events (such as presentations or competitions)." This customer-driven definition is complemented by more specific criteria of effectiveness, which include "quality, quantity, downtime, satisfaction, group stability over time" (from Goodman, 1986: 145).
Sundstrom et al. (1990) distinguish between context and process in their attempt to organize the literature of work teams. For them, factors external to the group may have stronger effects on performance than internal processes. By context, they refer to the culture of the larger organization within which the group operates and its physical environment. But they also include such factors as technology, mission clarity, autonomy, performance feedback, rewards and recognition, and training and consultation. Process is referred to as team development and consists of interpersonal processesFreeberg and Rock's communication and interactionnorms, cohesion, and roles. Between context and process are boundaries that differentiate a work unit from others, as well as the barriers that prevent access to information, goods, or people, and serve as points of external exchange with other teams, customers, competitors, and so on. Sundstrom et al. are vague about causal and temporal dynamics, preferring to treat these factors as interrelated processes, but "output" is team effectiveness, which they define as performance (acceptability of output to customers) and viability (members' satisfaction, participation, and willingness to continue working together).
Particularly notable in the Sundstrom et al. list are those factors that have received little attention in the experimental literature reviewed by Freeberg and Rock. These factors include organizational cultures, the physical environment, and the integration and differentiation aspects of group-organization boundaries. This set of variables defines the relation of a work team to its organization and determines what constitutes effectiveness
in its particular context. The key point is that teams vary in the extent to which their performance is affected by the larger organization in which they are a part.
The more that team performance depends on synchronizing with counterpart units, the less that performance is a function of internal group processes. Externally oriented teams are probably less differentiated, less autonomous, and more integrated into the larger system "through coordination and synchronization with suppliers, managers, peers, and customers" (Sundstrom et al., 1990:124). They are also more influenced by the culture of the organization than more differentiated or autonomous units and arrange the physical environment in ways that foster external exchanges. Whether externally oriented teams pass through the same developmental sequences as the more autonomous "inward-looking" teams remains an issue. The intertwining of process and context, suggested by the Sundstrom et al. framework, suggests that team development varies with differences in organizational contexts. Experimental research has done little to elucidate this issue since laboratory teams are largely without context and interact over short periods of time. Field research may, however, provide the needed clarification.
Support for the central role of contextual variables in team performance is provided by Gladstein (1984). Data from members of 100 sales teams in the communications industry were used to test a comprehensive model of group effectiveness, defined as satisfaction, self-reported effectiveness, and sales performance. Interestingly, the satisfaction and effectiveness ratings did not correspond to actual performance. Team members attributed sales to their internal interactions and experience rather than to the key contextual variable of market growth. Intragroup processes, leadership behavior, training, and experience influenced self-reported effectiveness and satisfaction, but actual performance was related to the way teams managed their interactions with outside groups and other aspects of their organizational context. One implication of these findings is that attempts to foster internal processes such as cohesiveness may not improve team performance as much as negotiating favorable objectives or by promoting group products to top management. It may even lower performance, as when cohesive groups enforce group standards that restrict output (Schachter et al., 1951; Berkowitz, 1954). Another implication is that process changes alone are unlikely to improve performance when structural (contextual) factors are not also taken into account. Interventions may encourage members to attribute the source of problems to their own behaviore.g., lack of skill in conflict managementand prevent them from considering the source as organization-level phenomena. For example, team building is unlikely to boost sales in a stagnant market over which the team has little control. However, it may boost morale, which sustains the team through a downward cycle, keeping it intact to take advantage of an upturn in the market.
When context is taken into account, teams are seen as part of larger structures. The connections between team and organizational performance may be quite complex. Successful interventions at the level of teams may not translate into improved organizational performance: structures and processes that may serve as facilitators or inhibitors of change from one level to another. To date, methods have not been developed to follow changes through a systemfrom individual to group to organization. Nor is there a framework that can guide research on linkages between levels of analysisfor example, the effect of team productivity on organizational productivity or other types of performancealthough progress toward developing such a framework is being made (see Harris, 1994).
Context also refers to organizational culture. Cultural analysis is common when one thinks of ethnic or national entities but it has recently been applied to the study of groups and organizations within a society. Although "culture" is still an elusive concept, the rapid development of a literature on the subject attests to the likelihood of cultural influences on the way that people in organizations think, feel, and act. Put another way, the need for a concept of culture reflects the judgment that behavior cannot be accounted for only in terms of structure and processes.
Less consensus exists on a formal definition. Schein argues (1985:6):
the term culture should be reserved for the deeper level of basic assumptions and beliefs that are shared by members of an organization, that operate unconsciously, and that define in a basic 'taken-for-granted' fashion an organization's view of itself and its environment.
Others have included in their definition observed behavioral regularities, norms that evolve in working groups, dominant values espoused, a guiding philosophy, the rules of the game, and a feeling or climate conveyed in an organization. For Schein, these various meanings are reflected in leader behavior and are, in fact, regarded as the primary functions of leadership. Just how cultures are conveyed to group members and the effect of those cultures on behavior are demonstrated in case material used by Schein and others. (Relying also on case material, in Chapter 12 we illustrate the role played by organizational cultures in affecting decisions about training programs.)
Another contextual variable is the relationship between organizations during times of change (see Gladstein, 1984). Restructuring in response to market changes consists both of internal changes in structure and mode of operation and external changes in relations with other organizations. Critical to this process are the decision makers in boundary rolesthose who negotiate relationships between departments within an organization or between autonomous organizations. A substantial literature on boundary roles provides many insights into the constraints, opportunities, and tactics used by negotiators in these roles (for a recent review, see Kahn, 1991). The
management of interdependence, joint ventures, and the establishment of new "regimes" are responses to change that can be orchestrated through a negotiation process (see Kremenyuk, 1991). The research literature on organizations contributes to the understanding of these processes.
Teams are dynamic entities. According to Tannenbaum et al. (1992:2): "roles and norms evolve, members develop new skills and attitudes, tasks are modified, communication patterns unfold, goals are revised, personnel may change and hopefully, progress occurs." The inevitability of change makes it useful to focus on developmental phases, transition points, and team building. These concepts are discussed in this section in three parts: team learning processes in which cognitive variables are emphasized; team building, which highlights the role played by motivation in performance; and, third, the implications of team building for interteam relations. Each of these parts focuses on process (throughput) variables rather than outcome or output measures.
Team Learning, Developmental Phases, and Metacognition
Recent conceptual work has concentrated less on the determinants of team performanceconstrued in input-processoutput terms-than on the way teams learn through time and repeated interactions among their members. These investigators examine the details of group processes and ask questions about the way that teams learn or acquire new insights that contribute to performance. By merging cognitive and social phenomena, these studies explore the "mechanisms by which people actively shape each other's knowledge and reasoning processes" (Resnick, 1991:2). Rather than viewing context as an exogenous influence on performance (as does Sundstrom et al., 1990), this view argues either that the social context in which cognitive activity takes place is an integral part of that activity or that context and process are intertwined, with effects occurring in both directions.
Shared perspectives, reframing, shared meaning, changed schemas, and metacognition are some of the concepts used to describe team learning. The process is often conceived in terms of a sequence of developmental stages. Whether teams follow a fixed sequence or show different temporal patterns in varied organizational contexts has not yet been determined. For Dechant and Marsick (1992), team learning evolves through four phases. In the first phase, "contained learning," the individuals in a group become aware that they are part of team, but there is little exchange of perspectives or reframing based on the perspectives of others. In the second phase, "collected learning," members share their understandings of the problem,
but there is no collective reframing. In the third phase, "constructed learning," members develop a language of shared meaning, including a lexicon containing models and metaphors. A consensus begins to develop through a process of exchange, criticism, and role reversing. In the fourth phase, "continuous learning," the group extends its consensual understanding to other parts of the organizationreferred to as "boundary crossing''and, by so doing, broadens its own perspective. Proceeding in a linear fashion, members become, over time, more like a team in the sense of acquiring shared meaning around language, roles, mission, and purpose. This conception takes on added significance when the phases are construed by the authors as intervening processes between influences and learning results at each of three levels of analysisindividual, group, and organization. Although developed largely on the basis of observations made in one company, and therefore subject to bias, the authors have developed a "Team Learning Survey" used for data collection in other large companies.
In a study of the lifespans of eight naturally occurring teams, Garsick (1988) challenges the view of a stage-like process through which groups develop. Her concept of "punctuated equilibrium," borrowed from evolutionary biology (Gould and Eldredge, 1977), refers to "progress through an alternation of stasis and sudden appearancelong periods of inertia, punctuated by concentrated revolutionary periods of quantum change" (Garsick, 1988:16). She is less interested in slow evolutionary learning processes, as depicted by Dechant and Marsick and elsewhere (see, e.g., Tuckman and Jensen, 1977), preferring to concentrate on the transitions that occur, usually at the temporal midpoint of a team's calendar, when a "major jump in progress" takes place. Parallel concepts from other areas of research include midlife transitions in adult development (Levinson, 1986), framebreaking changes in career development (London, 1985), and turning points in negotiation (Druckman et al., 1991; Druckman, 1986). This work directs attention to periods of stability and change, noting the importance of situational contingencies that can influence significantly the path a team takes. More important, perhaps, are implications for intervention activitiesactivities that affect a team's development. Knowing when these transitions are likely to occur can facilitate the timing of interventions: interventions that occur either "too early" or "too late" can prevent a team from turning a crisis into an opportunity for growth and development. Interventions or feedback that help members reframe their problems during transition periods can make the difference between effective and ineffective teams. Garsick provides a useful description of the process. Further work is needed to develop indicators of transitions and to distinguish between effective and ineffective interventions.
Transitions or turning points are not imposed on teams. They typically result from activities in which members monitor their own progress and use
resources to accomplish objectives. This process can be aided by feedback provided by observers. Recent work by Klein et al. (1992) highlights the strategic role of feedback in facilitating the way that teams handle information. Although the authors are less concerned about assessing the effect of the feedback on team performance, they provide categories that define what to look for. Referred to as "behavioral markers," the categories are based on a theory about the ways in which teams change and improveby developing a sense of identity, by moving toward goals, by learning how to perceive the world, by achieving a higher level of cognitive complexity, and by learning to monitor themselves, "metacognition." Examples of the categories are anticipate/confirm, clarify/compensate, detect and fill information gaps, and share mental and time management activities. Incidents are translated into these categories as observers learn ''to use specific conversations and behaviors as the basis for feedback delivery, rather than global ratings or general evaluations of team effectiveness" (Klein et al., 1992:21). This framework appears to be useful for observing, and then correcting, performance at a microlevel of specific interactions among team members. However, it does have shortcomings. The strong emphasis on cognitive factors may leave out important motivational variables likely to affect a team's performance. An exclusive focus on internal group processes omits attention to context and boundary processes (discussed above). Moreover, their preoccupation with usefulness apparently leads them to overlook the role of research in documenting the impact of different kinds of feedback.
Taken together, the studies reviewed in this section underscore the relevance of social and cognitive processes that can only be understood at the level of groups. Group-level explanations for behavior have been a conceptual and methodological challenge for social psychologists since the founding of the field. The team-development perspective advanced by Dechant and Marsick (1992), Garsick (1988), and others (e.g., Glickman et al., 1987; McGrath et al., 1986) describes a process in which members become increasingly self-conscious of their team identification. They acquire shared perspectives that transcend, and differ from, the meanings brought to the group by its members. As such, cognition becomes metacognition as this term is used by Klein et al. It is a property of teams in the sense that social processes are internalized in a manner similar to Mead's (1934) idea of "conversations with the generalized other" (see also Vygotsky, 1978). These studies seem to be struggling with concepts that situate group processes between the earlier "collectivist" notions of a "group mind" (e.g., see Allport, 1969) and the modern-day cognitive social psychologists' emphasis on individual thinking that is influenced by others with whom he or she interacts (e.g., Nisbett and Ross, 1980). The middle ground may well be found in a focus on the characteristics and products of the interaction process (Sherif and Sherif, 1956:342), including "developing reciprocities among individu-
als, organizational structures, and group products, like social norms." It is also found in Wegner's (1986) recent work on transactive memory, which takes into account both the diversity and uniformity of group members' thought patterns. But, even at this middle level, the phenomena have eluded measurement because of uncertainties about how to combine performances of team members: Are they to be combined by using linear (additive or multiplicative) or nonlinear algorithms?
The usefulness of emergent concepts is illustrated by Orasanu and Salas' (1993) attempt to organize the literature on team decision making. Their concept of "shared mental models" seems to account for research findings on the distinguishing features of effective and ineffective teams operating in field (not laboratory) settings. It refers to organized knowledge shared by team members who work together over relatively long periods of time. According to Orasanu and Salas (1993:7): "such knowledge enables each person to carry out his or her role in a timely and coordinated fashion, helping the team to function as a single unit with little negotiation of what to do and when to do it." In many of the studies they reviewed, effective teams were those that developed "shared models of the game and their roles in it so that much of their teamwork was habitual and that minimal language served a guiding or correcting role" (p. 10). When asked to account for a team's success, coaches will often say that they play well together or that they anticipate each other's moves. What the coaches may be observing is the effect on performance of having shared mental models or of developing a "team mind" (see Klein and Thordsen, 1989).
Central to this process, according to Orasanu and Salas, is the role played by team leaders. For example, in a study of cockpit simulations, Orasanu (1990) found that leaders of high-performing crews explicitly stated more plans, considered more options, provided more explanations, and sounded more warnings or predictions. Supported by results obtained in other studies (e.g., Chidester and Foushee, 1988), the Orasanu finding suggests that future research should focus on the role of leader communication in team performance. The research should also distinguish between decisions and performance: better decisions made by leaders may not translate into improved performance by members or a leader's poor decisions may be well implemented by the team. In stratified teams, such as in the military, decision making may be a different process than team performance or operation. Overall, however, the link between team decisions and performance has received little attention. This lack of attention may be due, at least in part, to separate traditions of research: one tradition focuses on decision processes, the other on measuring performance outcomes.
The literature on teams illustrates considerable breadth in terms of the variety of concepts, methods, and settings used. The meta-analysis performed by Freeberg and Rock (1987) call attention to the large number of
laboratory studies using quantitative methods to evaluate influences on performance. Field studies have been both quantitative (e.g., Gladstein, 1984) and qualitative (e.g., Garsick, 1988; Argyris et al., 1985). Each of these types of studies has contributed in important ways to an understanding of team performance. The contributions can be regarded as being complementary. The quantitative research has been useful in modeling the sequence of influences on performance. The qualitative studies have made evident some advantages of regarding teams as entities for study in their own right. Dechant and Marsick's (1992) work illustrates an attempt to merge the approaches. Their qualitative work, performed in the context of one company, produced a theory of team development. Their quantitative work, using large samples of teams in the field, serves to test hypotheses derived from the theory. The advantage of combining the approaches is that concepts developed from the qualitative work become the variables that are evaluated for their effects in the quantitative studies. The committee endorses this multimethod strategy for further research on team learning and development.
Team Building and Performance
Many of the investigations of team learning overlook the role played by motivation in performance. Focusing primarily on cognitive processes, the studies have described team development in terms of the acquisition of new concepts that contribute to problem solving or task products. But team development can also be described in terms of acquiring an identification or enhancing cohesion. These processes are heavily influenced by motivational variables. Insights into those processes derive largely from the literature on organizational development and team building. There are different approaches to team building, but all accept the assumption by Tannenbaum et al. (1992:3):
a team's active involvement in planning change is more likely to result in favorable consequences than imposing change and that the people closest to the task situation can solve their own problem if provided with an appropriate structure and process for doing so.
Four general approaches to team-building interventions emphasize goal setting, interpersonal relations, role clarification, and problem solving. There is much enthusiasm for these approaches among practitioners and consultants, but it is not matched by strong empirical support for their effect on team performance.
Results from reviews of studies that evaluated effects of team-building interventions (DeMeuse and Liebowitz, 1981; Woodman and Sherwood, 1980; and Buller, 1986) do not find strong effects. Although positive results for team building were found in many of the early studiesfor example,
80 percent of the studies reviewed by DeMeuse and Liebowitz (1981) and 63 percent of the studies examined by Woodman and Sherwood (1980)few of the effects were obtained on performance variables. Rather, effects were obtained primarily on perceptions and attitudes. It may well be that the interventions exert a stronger influence on perceptions and attitudes than on performance, but it is also possible that methodological problems prevented the detection of effects. One problem is that many studies did not measure performance or effectiveness: of the 52 team-building studies identified by Buller (1986), only 9 used performance criteria. Another problem is that the team-building interventions were weak or obscure. A third problem is that the study designs were flawed by confounding variables contained in the intervention package. Of the nine studies that measured performance, only five used experimental or quasi-experimental designs with some control over extraneous variables. Interestingly, an inverse relationship was found between degree of rigor and outcome success for the team-building intervention: the more rigorous the design, the weaker the outcomes.
These evaluation studies highlight a discrepancy between the generally accepted assertion that the way a team interacts influences its effectiveness and the weak relationships obtained in the studies. Three interpretations are possible: the assertion is incorrect; the interventions did not improve team interactions; or the studies did not detect relationships that exist because of inadequate designs.
More recent reviews of studies designed to avoid some of the methodological problems have clarified some of the effects. In a review of 17 empirical team-building studies reported since 1980, Tannenbaum et al. (1992) found many improvements in methodology in comparison with the earlier work. Eleven of the 17 studies used a quasi-experimental designin particular, a pretest, posttest, nonequivalent control group design. However, the sample size of teams examined remained small, although the incidence of single team studies decreased.
Improvements in team members' perceptions or attitudes were shown more consistently than behavioral changes. Four of the ten studies that assessed behavioral change reported mixed or nonsignificant results. A similar pattern was obtained for the studies that used the strongest research designs: most of these studies showed positive results for perceptual or attitudinal change; only one study found a significant behavioral change. Other changes in team-building interventions are a trend toward using multiple approaches and the inclusion of more than self-report indicators. Over 75 percent of the more recent studies evaluated interventions targeted at more than one obstacle to team development. It is not clear, however, why any particular approaches are used: specific team concerns are not coupled in a clear way with particular interventions. Team building may not be an
appropriate intervention in some circumstances: a mix of types of interventionsteam building, counseling, trainingmay work better. Such a mix is captured in the problem-solving approach to team building. This consists of a self-diagnosis followed by interventions designed to focus on the diagnosed problems. Most studies do not include a thorough diagnosis prior to the treatment. Those that do report more effective problem solving for the groups that diagnose the problem before beginning solutions (e.g., Hirokawa, 1983). By withholding proposed solutions until after a diagnosis, a group is more likely to suggest several alternatives for consideration (Maier and Maier, 1957). They are also less likely to become enamored of particular solutions leading to decisions based more on persuasiveness than on solution quality (Maier and Hoffman, 1960). These findings have implications for practice. Following Tannenbaum et al. (1992), we believe that team building should be considered part of a larger improvement strategy that includes multiple interventions.
The preference of most investigators is to assess team processes rather than performance. Yet as the studies reviewed above found, improved process may not translate to improved team performance. Effects of team building interventions may differ for different dependent variables: they appear to have a stronger effect on perceptions and attitudes than on behavioral changes. Although process measures are valuable in explaining why an intervention succeeds or fails, changes in process do not mean that the intervention influenced performance (see Gladstein, 1984). The relationship between process and performance is likely to be complex. Insufficient time lags used in most of the studies may have masked the relationship: process changes may be manifest in performance only after a period of time has passed; conversely they may dissipate through time. Long-term effects are rarely studied; most interventions are essentially one-time events. Further work should use causal modeling techniques to detect reciprocal relationships between process and performance over time. It should also investigate effects of task and developmental stage as variables that may moderate the process-performance relationship. The type of assessment also matters. Tannebaum et al. (1992:26) concluded that "the further removed the dependent variable is from the immediate control of the team ... the less likely that the team building intervention will demonstrate improvements." Team building is most likely to improve those aspects of performance that are in control of the team and less likely to affect aspects of performance determined by external factors. Needed are more precise connections between the specific procedures used in an intervention and the particular variables affected by those procedures, with proper control for assessments of casual paths.
Some progress toward better designed evaluations is reflected in two studies reported by Buller and Bell (1986) and Wolfe et al. (1989). Buller and Bell (1986) designed a field experiment to examine the effects of two
interventions, team building and goal setting, on the performance of hardrock miners working in an underground metal mine. Using a quasi-experimental design, the study assessed the independent and interactive effects of these variables (before and after the interventions) on several measures of productivity over a period of 3-1/2 months' quantity of production (tons per man-shift) and quality of production (grade of silver). Although care was taken to ensure a relatively unambiguous interpretation of the results, limitations still existed. The slight improvements that occurred for the teambuilding intervention are inconclusive due to a lack of control over several factors in the field setting.
The limitations in the mine study may be symptomatic of field research on this topic. One limitation is that team-building interventions are diffuse. The intervention "packages" contain many elements that, together, can produce either strong or weak effects on performance. For example, some aspects of the package that enhance performance may be offset by other aspects that interfere with performance. Due to their complexity, teambuilding interventions do not focus specifically on the performance variables (which, in this study, were tons per man-shift and quality of ore). In this sense, team-building packages are similar to the packages designed for accelerated learning, stress reduction, or influence as described in the committee's first report (Druckman and Swets, 1988): although it may be possible to demonstrate effects, it is difficult to ascertain which part of the package accounts for those effects. Another limitation is that relatively short-term evaluations may not capture the full effects of the intervention. French and Bell (1984) noted that team building may be a relatively slow process of developmental change that is manifest only over the long term. The Buller and Bell (1986) evaluation period (3-1/2 months following the intervention) may not have allowed for detection of a slow process.
Two other problems characteristic of field research are referred to as selectivity biases and reactive effects of the experiment. Pre-experimental differences between the treatment groups (intervention and control) make it difficult to compare effects on performance. In many field settings these differences are unavoidable. Although it is often possible to adjust these differences statistically (correcting for different baseline performances), the problem is not entirely solved in this way. More informative data on performance would be those on changes over time within treatment groups. But there are problems with this measure, too: changes may be a result of the increased attention given to those employees in the experiment. The increased motivation that may result from such attention, "Hawthorne effects," could account for improvements. Many packages intended to improve performance may well be capitalizing on these effects (Druckman and Swets, 1988). In fact, Buller and Bell (1986:325) argue "that the objective of (team-building) interventions is a sustained Hawthorne effect." If this is
the case, then team-building effects on performance, when they occur, are due primarily to the increased attention to employees and their work.
More recently, Wolfe et al. (1989) evaluated effects of team building on the performance of a simulated company. Following Buller and Bell, these authors comment on the weakness of the team-building interventions used in earlier experiments. Their experiment was an attempt to strengthen the effect of team building on performance by having "a practiced interventionist implement a team development effort within an intensive format" (Wolfe et al., 1989:393). The strongest effects occurred on team cohesion. Cohesion was measured on an index consisting of three correlated parts emphasizing belongingness and commitment to the team (Seashore, 1954): extent of perception of self as a member of the team; extent of preference to remain on the team; and extent of perception that the team is better than other teams with regard to the way members get along, the way they help each other, and the way they stick together (see also Norris and Niebuhr, 1980). Teams exposed to the intervention were more cohesive in this sense than control groups throughout the simulation; they also expressed higher levels of self-disclosure during the interactions. However, the intervention produced only marginally better economic performance during the early stages of the simulation, due largely to the increased cohesiveness. They were able to maintain their advantage but not increase it through the course of the simulation. Nor did the "treated" teams indicate that they learned more or express higher levels of satisfaction with the experience than the "untreated" teams.2They also underestimated their performance during the early phases of the experiment, but by the end they overestimated their team's performance. These results support findings obtained in earlier studies (e.g., McKenney and Dill, 1966; Deep et al., 1967; Hand et al., 1975) showing that the major contribution of team building is its effects on morale, cohesion, cooperation, and mutual trust. These effects do not translate into improved team performance in a simple or direct way. For example, a cohesive team may not improve its performance due to a lack of resources, poor intergroup relations, technical problems, or adverse conditions in the environment. The effects on team process, however, do have implications for relations between teams, a topic to which we now turn our attention.
Team Building and Interteam Relations
Team-building interventions not only foster positive intrateam relations, but also produce negative interteam relations. The enhanced cohesion resulting from team-building exercises may be a source of biased images and negative attitudes toward other teams. The well-known hypothesized relationship between ingroup amity and outgroup enmity (see, e.g., LeVine and Campbell, 1972), has been supported by results obtained in numerous laboratory
and field studies on intergroup relations (for reviews Stein, 1976; Tajfel, 1982). Less clear, however, is the direction of causality:. Does internal cohesion cause external conflict or is the cohesion a reaction to conflict? The cohesion produced by team-building interventions is due largely to such internal group processes as cooperative problem solving and conflict management within the team. External competition is rarely included as part of the package. At issue is whether this source of team identification, developed in the absence of interteam competition, leads to conflict.
With regard to biased perceptions, results from many experiments show that the bias can be aroused by mere categorization; results from other experiments link the bias to competitive situations. Together, the studies suggest that competition is not a necessary condition for ingroup bias, although it can result from competition and is probably stronger in competitive situations (see reviews by Tajfel, 1982; Brewer and Kramer, 1985). Focusing on mechanisms, Messick and Mackie (1989) discuss alternative theories intended to explain the results of these studies. Most promising perhaps is Turner's (1987) self-categorization theory, which posits that evaluative bias ("our group is better than yours") is a function of perceptual bias (categories that distinguish among groups in terms of similarities and dissimilarities). Without perceptual distortions, the evaluative biases would not occur. Implications for team building turn on whether the "treatment" produces the sorts of categorical distinctions that lead to biased evaluations.
More relevant, perhaps, are the studies that examined behavior in intergroup settings. Most of these studies showed that groups play games more competitively than individuals and that people who are representatives of groups are more competitive bargainers than those who are not (Messick and Mackie, 1989). At issue in these studies is whether the observed competitiveness is a function of group identification per se or other factors in the situation. Results obtained by Insko et al. (1988) suggest the other factors explanation for competitive behavior. They showed that the increased competitiveness of groups was due to intragroup consensus about the group's strategy; when groups acted in lock step, enhanced competition occurred. A similar finding was previously obtained by Druckman (1968): the most competitive groups in his study were those that agreed in prenegotiation sessions on the relative importance of the issues under discussion. Thus, it may be that a consensual strategy, rather than group identification per se, is responsible for the observed increase in competitive behavior.3
Further support for the importance of strategy development comes from the results of a recent meta-analysis of findings on compromise behavior in bargaining situations (Druckman, 1994). The analysis addressed the issue of the relative importance of variables that are hypothesized to influence interteam bargaining. By including the variables of strategy preparation
and group representation in the analysis, it was also possible to address the specific issue of the relative importance of group identification and strategy development on bargaining behavior. The results showed that group representation or accountability was not a strong influence on compromise behavior; representatives compromised only somewhat less than nonrepresentatives. Stronger effects were obtained for the way teams prepare for negotiation, their orientation toward the negotiation, and the size of the conflict defined in terms of initial position distance. Specifically, Druckman (1994) found the following effect sizes4for the nine variables: bargainer's orientation as competitive or cooperative (.42), prenegotiation experience as strategy versus study (.37), time pressure as deadline or no deadline (.37), initial distance between positions (.35), opponent's strategy as tough or soft (.32), team representation as bargaining for a team or for self (.30), accountability to the team (.27), visibility of the negotiation as audience present or not present (.21), and whether the issue was large or small (.18). An implication of these findings is that bargaining competitiveness is more likely to be increased or decreased by "affecting" the way team representatives prepare for the discussions or their orientation toward the negotiation than by the extent of their "loyalty" or their accountability to the team.
These results highlight the importance of strategy development as an influence on the extent of conflict between teams; it may also be a positive influence on team performance. The Buller and Bell (1986) study found that strategy development was a key variable that intervened between the team-building interventions and performance. According to Buller and Bell (1986:326): "team building may have influenced the development of strategies which in turn may have improved quality of grade." This is consistent with earlier models of factors that influence the behavior of individuals in organizations (e.g., Hackman, 1976). While serving to focus the team effort, careful planning sets in motion processes that heighten team identity, which, in turn, may hinder relations with other teams. The intrateam dynamics would seem to foster the kinds of perceptual distinctions that lead to evaluative biases, consistent with Turner's (1987) theory (discussed above). One sequence of events is as follows:
within-team strategy formation ? intrateam consensus on goals ?enhanced sense of member identity with the unit ? perceptual distinctions between "us" and "them" ? evaluative biases and related distortions ? heightened competitiveness, win-lose dynamics ? reduced willingness to collaborate, compromise, and so on.
Although this sequence captures some of the implications of the research completed to date, it is presented primarily as a set of hypothesized relationships that should be subject to further research. It calls attention to linkages between internal, intrateam development and external, interteam
relations. Those connections provide a conceptual bridge between developmental sequences and boundary-role processes (discussed above). The committee endorses attempts to design new studies that explore these relationships.
The Freeberg and Rock (1987) meta-analysis makes evident the importance of such training variables as prior task experience, practice, and feedback or knowledge of results. The case studies of team development emphasize the importance of the active learner who controls the pace and content of his or her own learning. Neither tradition of team research deals with the issue of transfer of skills from learning to performance settings. That issue is addressed with regard to simulations in Chapter 3.
The terms simulation and games are often used interchangeably. In this section, we focus primarily on games that are role-playing exercises involving groups. This is considered to be a type of simulation although it differs from the kind of mechanical simulators or training simulations discussed in Chapter 3. Most of the games discussed in this section are concerned less with the training of specific mission-oriented operational skills and the transfer of those skills than with learning general and social skills in educational settings. Because of its popularity as a training and research device, particularly in the military, the "technology" of game design and evaluation has been the subject of considerable research and conceptual work.5In this section, we discuss issues related to the use of games and provide a summary of the evidence obtained to date on their effectiveness.
Games are frequently used as exercises in team development packages designed by organizational consultants. One of the more popular games is "Pumping the Colors" created by Gary Shirts for facilitating team building. Rarely, however, are the exercises evaluated in terms of whether they accomplish their objectives. Even when used as tasks in studies that are comparing different team-building interventions, the game exercises are not evaluated (e.g., Wolfe et al., 1989; Miesing and Preble, 1985; Hsu, 1984; Norris and Niebuhr, 1980). Thus, implications for the effectiveness of games used for team development must be derived from more general research on educational games. That research has addressed both cognitive and motivational effects of games; by so doing, it has implications for their effect on team-learning and team-building processes. However, measures of individual learning and motivation, used in most of these studies, may not translate into team outcomes. At issue are the effects of team members' development for a team's performance, and few studies to date have addressed this relationship (see Druckman and Bjork, 1991:Ch.12).6
The kinds of exercises used most often for training take the form of games played by students or trainees to discover new concepts or to develop new
skills. At issue is whether the intended learningdefined either as acquiring concepts or skillsoccurs. This issue has been addressed by research designed to evaluate outcomes of the learning experience. A second issue is whether the new knowledge or skills can be used effectively in other environments. Beginning in the early 1960s, an active network of gaming researchers devoted their careers to finding answers to these questions. Many of these studies have been reported in the journal, Simulation & Gaming, as well as in a number of edited books that cover a variety of types of uses and applications; see also Crookall and Oxford (1990) for an extensive bibliography of general and specialized sources. Our review draws on this literature, emphasizing, in particular, work reported during the past 15 years. We concentrate on gaming technology in general, rather than specific simulations intended to develop particular operational skills (see Chapter 3).
Games have a number of features that should facilitate learning. Among the features highlighted in the literature are involving students in an active learning situation (Glenn et al., 1982), enhancing their control over the learning environment (Boocock and Schild, 1968), focusing on learning principles and referents for concepts (Greenblat, 1975), rapid feedback and the learning of strategies (Schild, 1968), enhancing motivation to learn (Bredemeier and Greenblat, 1981), and providing an opportunity to encounter problems in ways analogous to the way they are encountered in realworld contexts (Van Sickle, 1978). The key question is whether these features contribute to better learning.
Cherryholmes (1966) is often credited with the earliest evaluation of learning through game playing. On the basis of only a few studies completed to that date, he concluded that only interest in the material being learned improved significantly; negligible changes occurred on cognitive and attitudinal variables. Somewhat more optimistic conclusions were reached 10 years later. Pierfy (1977) reviewed studies, reported during the 1960s and 1970s, that compared learning through games versus other educational experiences. With regard to learning, 15 of 21 studies showed no significant difference between experimental and control groups, indicating that the games were not more effective than conventional instructional techniques: of the other 6, 3 studies showed games to be better, 3 showed conventional methods to be better. With regard to retention, 8 of 11 studies reported significant differences in favor of games, indicating that students retained information longer than those trained with more conventional approaches. With regard to attitude change, 8 of 11 studies showed that games had a greater effect on attitudesin terms of increased realism and approval of real-life personsthan conventional methods. For interest, 7 of 8 studies reported significantly more interest in the simulation activities than in the more conventional classroom activities, a finding that supports Cherryholmes' earlier conclusion.
Pierfy goes on to indicate that deficiencies in research design render these conclusions tentative. He lists a number of sources of possible confounding factors in many of the studies, including:
• unintended biases from game designers who also conduct the studies;
• unintended effects of instructor variables when matched classes are taught by different instructors;
• ''Hawthorne effects" due to the difference between one group receiving a new method (game) while the other group is exposed to a familiar, conventional method;
• for some studies, administration of the posttest after a debriefing, allowing for the possibility that the posttest responses were influenced by the debriefing discussion;
• the techniques used in control classes may be regarded by students as vague, dull, and incomplete, so that any gains shown for the simulation classes are not strongly biased;
• use of only a pretest-posttest designnot adding groups without the pretestallowing for the possibility that the pretest interacted in different ways with one or another method of instruction.
These flaws are not limited to game evaluations, but also characterize much of the evaluation literature in general; many of them can be remedied.
Stronger studies have appeared in more recent years due, in part, to attempts by designers and users to routinely incorporate evaluations in their packages. In addition, comparability from one study to another can be increased if the same categories of learning are used in constructing dependent variables, including knowledge of facts, analogies, game structure, skills needed for playing the game, knowledge of outcomes of various strategies used in the game, perceptions of the game, and attitudes toward the game (see also Fletcher, 1971). A variety of methods can be used to measure any particular learning objective, as Anderson and Lawton (1992) illustrate with respect to the objectives of basic knowledge, comprehension, application, analysis, synthesis, and evaluation. Furthermore, replication of studies in a variety of settings would help to distinguish between findings that hold across situations from those that are specific to particular situations. Such replication can reduce the impact of confounding variables; see, especially, Campbell and Stanley's (1963) discussion of a "heterogeneity of irrelevancies."
Another methodological concern is that even the best evaluations may not uncover causal mechanisms. Most of the studies reviewed by Pierfy were demonstration experiments that simply show effects of the instructional packages: Does it work? Few attempts are made to "unpack" the parts in order to determine what may account for the observed effects on learning or motivation: How does it work? This distinction was recognized
in early appraisals of learning in simulation and games. In their wideranging survey of the issues, Boocock and Schild (1968) distinguish between the "engineering" and the "science" approaches to understanding: the former consists of demonstrating that a social technology produces gains; the latter is the identification of the mechanisms responsible for the gains. Few of the studies they examined searched for mechanisms; nor has there been any trend toward explanatory studies in more recent years. Investigators seem to have largely ignored the Boocock and Schild distinction.
Also missing in the research on simulations and games are issues raised by Bredemeier and Greenblat (1981) in their update of the Pierfy review. Bredemeier and Greenblat (1981) divide learning into three parts: subject matter, attitudes, and learning about oneself. With regard to subject matter, the available evidence suggests that games are at least as effective as other methods and are more effective aids to retention. With regard to attitudes, the evidence suggests that games can be more effective than traditional methods of instruction in facilitating positive attitude change toward the subject and its purposes. With regard to learning about oneself, they cite the results obtained by Johnson and Nelson (1978) showing that subjects who played games (versus those who did not) showed greater positive change on willingness to communicate. The positive effects on self-awareness may, however, depend on the extent to which the game was experienced in a positive way. Many of these conclusions support those reached in the earlier review by Pierfy. They do not illuminate reasons for why the effects do or do not occur and, therefore, make only small contributions to the development of theory in this area.
More recently, Randel et al. (1992) updated the earlier reviews in an examination of 69 studies that had been conducted over a 28-year period. Overall, they found that 56 percent of the comparisons between simulation games and conventional instruction showed no difference, 32 percent found differences favoring games, 7 percent favored games but their controls were questionable, and 5 percent found differences favoring conventional instruction. Dividing the studies into six subject-matter areas, the authors found that the greatest percentage of results favoring games were in mathematics (seven of eight studies) and language arts (five of six studies). Although the largest number of gaming evaluations have been in the area of social science, the majority of these studies (33 of 46) showed no differences in performance between games and conventional instruction. A meta-analysis of social science simulations (Van Sickle, 1986) reported a small effect size.
On the basis of these findings, Randel et al. (1992:269) concluded that games are likely to be more beneficial for topics "where very specific content can be targeted and objectives precisely defined." They also reaffirmed conclusions reached in earlier reviews that games show greater
retention for students over time and elicit more student interest than more conventional instruction. As did the earlier reviews, Randel et al. (1992) call attention to many of the design and measurement problems typical of the studies. They add to the earlier lists the possible confusion between effects produced by the game and those produced by the debriefing sessions. They also emphasize the importance of distinguishing between preferences expressed by players and what it is that they learn from the games.
The reviews make evident that there is much yet to be understood about the effects on learners of participation in games or simulations. Progress toward richer theory and application may depend on providing answers to the following questions:
• What accounts for the discrepancy between learners' impressions and subjective reports, and the weak evidence on performance?
• What is the relationship between motivational and learning variables? How does involvement in the game affect learning?
• Why are motivation and interest stimulated by games?
• To what extent are effects due to instructor variables? For example, those instructors who are amenable to using games may be people who stimulate relaxed classrooms as well as facilitate later changes in classroom atmosphere.
• What aspects of games are expected to have what sorts of distinct effects on what sorts of participants?
• Does what one learns in a particular game transfer to other situations? To what range of situations do the lessons learned in a particular game apply?
The promise of interesting games as vehicles for learning skills and concepts have been largely unrealized to date. That promise is based on the assumption that role-playing activities may do more than stimulate interest; they "may also involve students in an active learning situation that may teach them specific skills" (Glenn et al., 1982:209). Although the evidence to date is inconclusive, the problems may rest not with the technology but with the way it is implemented and evaluated. A clearer definition of what is to be accomplished by the experience, how to accomplish it, and, then, how to evaluate effects would help. So too would a theoretically based taxonomy of games that distinguishes among games used for different purposes (see Bredemeier and Greenblat, 1981). Game designers need to be guided by conceptual frameworks. Game evaluators need to increase their sensitivity to relevant methodological issues. Advances along both these lines will, almost certainly, strengthen the state of the art. They will also clarify the distinction between cognitive and motivational effects on participants in gaming exercises used for team development.
The research to date on teams provides a base of knowledge for improving team performance. A variety of input variablessuch as task complexity, structure, and task loadhave been shown to relate to the output variables of quantity, accuracy, and efficiency. These relationships depend on the way that the inputs influence such mediating processes as interaction, coordination, and cohesiveness. For example, effects of intrateam cooperation or conflict on the quantity and proficiency of team output depend on the way that cooperation or conflict affects cohesion. These findings derive from the path models developed by Freeberg and Rock (1987), which deserve greater attention in the development of theories of team performance.
Claims have been made for the performance-enhancing effects of teambuilding interventions. However, such interventions appear to have limited effects on performance. Further research is needed to determine if consultant and practitioner enthusiasm is warranted. The research should follow the multimethod approach of Dechant and Marsick (1992) in examining teams both qualitatively and qualitatively.
Recent studies have identified certain processes or activities that could enhance the effect of team-building interventions. One of these is timing: interventions are likely to be most effective during transition periods. Another is to elicit from a team a self-diagnosis of its problems before proceeding with an intervention. A third consists of developing shared mental models among members during periods when teams prepare to perform. The development of shared models among team members deserves further study.
One format for implementing these interventions is through gaming exercises. Games are popular training vehicles used widely in the military. Their popularity is based mostly on judgments made by participants, rather than on carefully designed evaluation studies. Although the literature on games is large, much of it consists of demonstration experiments or poorlydesigned studies. The few well-designed studies suggest that while they are effective in instilling positive attitudes toward and interest in the subject matter, they are not more effective than other methods as aids to learning. By producing stronger effects on motivational variables, games may be more useful for team-building than team-learning exercises. Further research is needed to establish the value of games in team training. In fact, both team building and gaming interactions need to be "unpacked" in order to determine what works.
Improved team performance may not translate into improved organizational performance. Factors external to the organization and largely out of its controlsuch as market growthmay account for its performance (Gladstein, 1984). Internal factors also influence organizational performance. The
enhanced cohesion and morale resulting from team-building activities may increase intraorganizational conflicts between teams. By strengthening the ties between members within teams, interventions can weaken relationships with members of other teams. This effect is heightened to the extent that team-building programs include strategy formation as part of the procedure. The effect of team building on interteam relationships requires further examination.
Interteam relations involve negotiation and other boundary-spanning processes that require harmonious relations for coordination to occur. They also call attention to the concept of linkages between levels of an organization. Research is just beginning about the way that improvements at one level of an organization affect performance at other levels. Integration of the literature on organizations with the literature on team effectiveness would help clarify relationships between micro- and macrolevel processes and organizational effectiveness.
1The effect sizes used in the Freeberg and Rock analysis are correlation coefficients calculated from reported t or F ratios, according to conversion formulas given by Wolf (1986).
2Evaluations of educational or training simulations focus on learning gains and satisfaction with the experience. Different results often occur: while expressing high levels of satisfaction with the experience, players do not usually show the expected gains in learning (see below).
3Whether strategy formation also produces evaluative biases that favor one's own group in noncompetitive situations is not known. It is possible that intragroup consensus contributes to the perceptual discrimination that seems to precede evaluative biases; this, too, remains to be explored.
4The effect sizes are averages, expressed as correlation coefficients. Each coefficient is highly significant based on the Stouffer method of adding Zs (see Wolf, 1986).
5A detailed survey of uses of simulations and games in the military is reported by Shubik and Brewer (1972). They identified approximately 135 active military simulations in use at the Department of Defense. To our knowledge, this survey has not been updated, nor have we been able to locate a similar survey of games developed for other uses. However, it is possible to get a rough estimate of popularity from various published sources. In the section on "Newly Available Simulations," appearing in each issue of the journal, Simulation & Gaming (formerly Simulation and Games), a wide variety of packaged games are made available. The games are distributed directly by the designers, by the institutes or councils that sponsored the gaming activity, by small companies formed to market particular games, or by book publishers. Sales for most games number in the thousands, but a few, such as the well-known crosscultural game BAFA BAFA, have sold well over 100,000 worldwide. However, sales may not be a good indicator of use for at least three reasons. First, the same packaged materials are used by many people: Gary Shirts, the designer of BAFA, estimates that players of his games number in the millions (private communication). Second, many games are designed and used for relatively idiosyncratic purposes, such as classroom adjuncts to other teaching techniques, experimentation in specialized areas, or training of highly specialized skills. Third, many games are distributed for free or are available in texts that contain role-playing exercises.
6 Highly technical simulations have also been designed for military training. Combat exercises have been simulated in the form of board games, computerized virtual realities, and field exercises that provide soldiers with broad experiences of many facets of combat over relatively long periods of time; see Oswalt (1993) for a review of current military applications. Unfortunately, few of these exercises have been evaluated systematically in terms of their effect on training goals. Most evaluations reported in the published literature have concentrated on games designed to improve skills in such areas as business management, language learning, negotiation, medical education and hospital administration, environmental management, and social science concepts. Thus, implications for the effectiveness of military simulations must be derived from a literature on nonmilitary applications.