Team mental model: construct or metaphor?

Team mental model: construct or metaphor?

Richard Klimoski

In the area of team performance, a number of authors (e.g., Hackman, 1992; Guzzo & Shea, 1992) offer a generalized framework for team/group performance, which delineates inputs (e.g., resources), process (e.g., collective effort), and outcomes (e.g. team performance). By and large, in this paper we discuss team mental models as reflecting team processes. However, an argument can also be made for considering them as inputs.

To illustrate the potential of how the team mental model construct might lead to a better understanding of team performance, consider trying to model individuals engaged in the real-time accomplishment of a team task. This would be analogous to people working in a production team on an assembly line There has been a recent resurgence of interest in group cognition in the field of organizational science. However, despite the apparent enthusiasm for the notion of the group mind in some modern guise, important conceptual work is needed to examine the concept critically. We attempt to do this in our treatment of the content, form, function, antecedents, and consequences of team mental models. In addition, we illustrate how the construct can bring explanatory power to theories of team performance and offer other implications for research and practice.

From the outset, writers who have had interests in the group as a social entity have tended to refer to the existence of something like the “group mind”. Generally, this was used to imply an aspect of the mental state of the members of a group collectively (e.g., Durkheim, 1938). While this mental state was based on individual members’s perceptions, thoughts, beliefs, and expectations, it was more than just the sum of such individual properties. It was a group level phenomenon.

Recently, there seems to be a resurgence of interest in this notion as writers and researchers in such fields as business policy and strategy (e.g., Bonham, Shapiro & Heradstveit, 1988; Floyd & Wooldridge, 1992; Reger & Huff, 1993), human resources (e.g., Cannon-Bowers & Salas, 1990; Kraiger, 1993), and OB (e.g., Fiol, 1993; Innami, 1992; Walsh & Fahey, 1986; Weick & Roberts, 1993) have all used or implicated “group mind”-like constructs. Terms such as shared mental models, common cause maps, shared frames, teamwork schemas, transactional memory, and sociocognition are being offered by investigators in these areas to explain variance in team development, team performance, strategic problem definition, strategic decision making and even organization performance.

Despite this enthusiasm in the field of organizational science for the notion of the group mind in some modern guise, we maintain that it is premature. More specifically, important conceptual work must first be done to examine the concept critically. We offer this article as a response to this observation.

The purpose of this review is to highlight the confusion in the literature relating to group cognition and to provide a conceptual foundation for the notion of a team mental model by addressing a number of pivotal issues (e.g., content, form, function, antecedents, consequences). Therefore, an inductive as opposed to deductive approach is adopted.

The plan for the review will be as follows. First, we will briefly examine certain concepts derived from cognitive psychology, with special reference to the concept of a mental model. In doing this, we are supporting the claim that “the same type of processes that occur for individuals are conceptually involved in the information processing by the group” (Hinsz, 1990, p. 12). This will be followed by a more elaborate and critical review of several domains where investigators are using mental model-like concepts to explain group level phenomena in organizational contexts. We will attempt to fairly characterize each domain by showing how these concepts are being treated, both conceptually and operationally. We will end up by arguing that there is much vagary here, but we will also conclude that the concept has potential as a psychological construct.

We will demonstrate that this potential can best be realized if the notion of a group mind is examined relative to what we know about mental representations and cognitive architecture at the level of the individual. Our review will highlight the usefulness of conceptualizing mental models in terms of key features. In particular, we will emphasize the importance of such things as content. That is to say, among other parameters, understanding what is shared or held in common (what does the model model) will be important to the understanding of shared cognitions. Finally, we will explore what is gained from our analysis and offer research implications based on our understanding of team mental models.

Because we will continually be referencing the concept of team, we would like to distinguish between teams and groups at the outset of this review. Groups are collections of individuals whose tenure together and division of responsibilities can vary considerably. However, following the lead of Cannon-Bowers, Salas, and Converse (1993), Orasanu and Salas, (1993) and others, we feel that a team consists of differentiated and interdependent members. Put another way, while all teams are groups, the converse is not necessarily so.

This distinction not withstanding, a lot of what we know about teams derives from research on groups (e.g., Orasanu & Salas, 1993; Sundstrom, DeMeuse & Futrell, 1990). At a minimum they share similar dynamics, and group and team performance probably have similar antecedents.

The Mental Model Construct

Individual belief structures play a powerful role in the development of socially shared cognition (e.g., Damon, 1991; Resnick, 1991); thus, in a real sense, we will only come to understand the notion of team mental models if we first focus on individual cognitive processes.

A plethora of cognitive terminology has been employed to help explain the process by which individuals make sense of their surroundings. Categories (e.g., Rosch, 1978), cognitive maps (e.g., Axelrod, 1976; Ford & Hegarty, 1984; Neisser, 1976; Weick & Bougon, 1986), belief structures (e.g., Fiske & Taylor, 1991), mental models (e.g., Johnson-Laird, 1983; Rouse & Morris, 1986), schemas (e.g., Anderson, 1980; Neisser, 1976; Rumelhart, 1984), and scripts (e.g., Abelson, 1976) are all concepts relating to “knowledge structures”.

Such structures are postulated to aid interpretive processes by enabling individuals to screen out information in order to prevent information overload and intolerable levels of uncertainty. On the negative side, these cognitive structures may cause individuals to ignore discrepant information and may inhibit creative problem solving. Because they are the bases upon which one relates knowledge, attributes meaning, and fashions understanding, these structures are central to the sensemaking process (Gioia, 1986) and much of the work relating to group cognition.

The phrase “team mental model” will be the terminology of choice in this paper; thus, the rest of this discussion will focus on the mental model concept.

The existence of mental models is widely asserted in the literature. They are often used in a way that is synonymous with “knowledge” in general (Rouse & Morris, 1986). However, according to Holyoak (1984, p. 193), a mental model is a “psychological representation of the environment and its expected behavior.” Adopting a functional approach, Rouse and Morris (1986) state that the role of mental models is to provide a conceptual framework for describing, explaining, and predicting future system states. They allow individuals to understand phenomena, make inferences, and experience events by proxy (Johnson-Laird, 1983). Our use of mental models refers to a general class of cognitive constructs that have been invoked to explain how knowledge and information are represented in the mind.

Mental model measurement continues to be a challenging undertaking, given the fact that models often take multiple forms and are dynamic in nature. Nevertheless, verbal protocol analysis, analytical modeling, and experimental methods are some techniques that have been used to review mental representations (Rouse & Morris, 1986). Measurement issues associated with mental models will be woven throughout this review. However, to anticipate one of our conclusions, the measurement of mental models, especially at the group level, is so complex and problematic that it deserves a separate review.

Researchers in organizational studies are increasingly embracing a cognitively-oriented perspective and incorporating the notion of mental models (representations) into their theories (e.g., Gioia, 1986; Poole, Gioia & Gray, 1989). For example, Reger and Huff (1993) demonstrated that strategists cognitively group their competitor’s strategies. Similarly, adopting a cognitive approach to the problem of competitor definition, Porac and Thomas (1990) used the notion of a mental model as a framework for understanding how strategists interpret their competitive environments.

Although mental models and other cognitive constructs have traditionally been considered at the individual-level of analysis, there has been a renewed effort to expand consideration of these phenomena to the group level.

The Team Mental Model Concept

The Group Mind Revisited?

The idea that groups of people can retain information through sharing in a way that transcends the cognitive facilities of individuals has been around for quite some time. The notion of a group mind (collective unconsciousness, group metaphor, group mentality) enjoyed a period of widespread popularity in psychology and sociology, with historical proponents including Durkheim (1938), Fleck (1935), and McDougall (1920). However, the once powerful concept declined in influence over time, and scholars such as Floyd Allport (1925) were quick to point out deficiencies. Ironically, this, in part, was due to the failure of researchers to develop a “collective mind” concerning how the notion of a “group mind” should be defined conceptually and operationally. As a result, the concept lost much of its intended meaning and became too amorphous to be a driving force in theory development or practice.

Recently, however, there has been a resurgence of interest in the general notion of a group mind and, generally, in cognition at the group-level of analysis. The current interest in “shared cognition” can be partially attributed to the increasing popularity of group research in organizational science. Indeed, Levine and Moreland (1990) assert that the torch of group research has been passed from social psychology to organizational psychology. However, many social psychologists (e.g., Levine) continue to make important contributions to research in this area.

Increasingly, organizational research is reflecting the fact that teams are the cornerstone of American modern industry (Cummings, 1981; Hackman & Morris, 1975) and that strategic decisions are often made by groups instead of individuals acting alone (e.g., Axelrod, 1976; Orasanu & Salas, 1993). Thus, several constructs which have been traditionally considered at the individual-level of analysis are now being recognized as group-level phenomena. Specifically, group affect (George, 1990), collective efficacy (Bandura, 1986), group-level problem identification (Larson & Christensen, 1993; Moreland & Levine, 1992), transactive memory (Wegner, 1987), group-level integrative complexity (Gruenfeld & Hollingshead, 1993), and group prosocial behavior, as well as group turnover (George & Bettenhausen, 1990) are relatively recent developments.

Another reason for the resurgence of interest in group mind-like constructs has to do with the growing belief that cognition is almost always a social phenomenon. “Reality” is jointly constructed by individuals acting in a social context (Levine, Resnick & Higgins, 1993; Resnick, 1991). In other words, the individual-level focus in the study of cognition is no longer perceived as adequate for capturing even individual level thinking in all of its variety.

Not surprisingly, various versions of the notion of shared cognition have emerged in the literature. Reflecting the view that mental processes can be meaningfully understood at the group-level of analysis, Larson and Christensen (1993, p. 6) re-define social cognition as the “social processes that relate to the acquisition, storage, transmission, manipulation, and use of information for the purpose of creating a group-level intellectual product.” In addition, terms referring to the collaborative nature of memory have been proposed, including joint (Edwards & Middleton, 1986) or group remembering (Clark & Stephenson, 1989), and transactive memory (Wegner, 1987). Furthermore, a number of researchers are postulating the existence of information processing effects at the group-level (e.g., Cannon-Bowers & Salas, 1990; Innami, 1992; Resnick, 1991; Walsh & Fahey, 1986). Even terminology like “teamthink” is being bandied about in the popular press (Martin, 1993).

A recent review of group research described the development of “shared understanding” as an essential group process (Bettenhausen, 1991, p. 350). Moreover, there have been numerous calls for research to investigate socially shared cognition and collective sensemaking activity at the group-level (e.g., Dukerich & Milliken, 1993; Hastie & Pennington, 1991; Innami, 1992; Isabella, 1990; Levine et al., 1993; Panzano, 1992). In addition, the notion of shared cognitions in a group is proposed as one of the most promising areas in sociocognition research (Levine et al., 1993).

Despite this interest and activity however, we argue that the notion of a shared reality must undergo further development before the field can move ahead. First, the very popularity of group mind-like terms in a variety of different domains in organizational science creates problems. There is a need to integrate these views, perhaps with an overarching framework. Second, neglected definitional and conceptual issues regarding the content, form, and development of shared cognition must be addressed. In fact, given the confusion that exists regarding the mental architecture of shared cognition, we might be well on our way to repeating the errors of the past. Although the concept of a team mental model may have potential, it will most likely suffer the same fate as the historical notion of a group mind unless such definitional, conceptual, and methodological issues are addressed.

Specific Applications

The shared mental model construct has been applied in a number of domains relevant to organizational science. Table 1 provides a listing of some of the works in the stream of research dealing with shared cognition and highlights the wide array of terms that have been attached to this concept. Because it provided a useful way in which to organize the literature, Table 1 is divided into two sections, namely strategic decision making and team dynamics/performance.

One of the patterns that is evident from this selective review of the literature is that few researchers clearly define or even explicitly describe what is meant by their terminology. Most authors do not go much beyond describing (or ascribing) a collectivity of beliefs, shared understanding, or some similarity in TABULAR DATA OMITTED the way information is processed. Conceptual issues regarding the content, form, and development of shared cognition remain largely unanswered.

Strategic Decision Making Domain

Using a shared mental model construct. A number of scholars share the position that the aggregation of individuals’ knowledge structures creates a context for efficient group decision making. Specifically, it has been argued that collective belief structures affect the speed, flexibility, and implementation of a decision (Walsh & Fahey, 1986), as well as facilitate problem definition, alternative generation, evaluation, and choice (Walsh, Henderson & Deighton, 1988). Innami (1992) hypothesized that an inclusive, balanced group mental structure would increase the quality of group decisions, whereas a biased group structure would decrease the quality of decisions. In addition, the development of shared meanings is believed to give form and coherence to the experience of group members, which facilitates coordinated action (Smircich, 1983). Furthermore, Daft and Weick (1984) hypothesize that organizational interpretation, which involves developing shared meanings, precedes learning and action.

The decision making literature cited in Table 1 share the common premise that group knowledge structures concerning strategic issues exist in some sort of a collective representation. However, the varying descriptions provided differ with regard to what aspects of knowledge are shared. While some researchers suggest beliefs (Axelrod, 1976; Bonham et al., 1988; Innami, 1992; Walsh & Fahey, 1986) or understanding (Daft & Weick, 1984; Floyd & Wooldridge, 1992) as the entities which decision makers hold in common, others refer to shared frames of reference (Isabella, 1990; Panzano, 1992) or shared categories (Fiol, 1993; Panzano, 1992). Gioia and Sims (1986) seem to indicate that similarity in cognitive processes leads to consensuality, which differs from the cognitive content focus of the other references.

Although many of the authors cited in Table 1 are equivocal about exactly what is held in common among group members, some have articulated different components of sharing. Weick and Bougon (1986) and Gray, Bougon and Donnellon (1985) suggest that agreement can center around concepts and/or relationships among concepts. Gray et al. (1985) also recognize that coincident meaning may involve values and ideologies. In addition, Floyd and Wooldridge (1992) and Wooldridge and Floyd (1989) emphasize that commitment in addition to understanding must be shared. They argue that consensus is strongest when a common understanding and common commitment to strategy exist.

Also relevant to the sharing issue, Fiol (1993) breaks the notion of consensus into two interpretive dimensions: content and framing. First, consensus around interpretations is embedded in the content of communications, which is reflected in the categories or labels that define what is expected (e.g., threats and opportunities). Second, meaning also resides in the framing of communications, which refers to the way in which viewpoints are expressed, regardless of content (e.g., rigid or flexible perceptions of an issue). Therefore, there can be agreement not only with regard to what is said, but how it is said.

Apart from the issue of what is shared, other differences can also be noted among the references listed in the first part of Table 1. For example, Bonham et al. (1988) as well as Walsh and colleagues (1986; 1988) explicitly incorporate political processes (e.g., power) into their respective notions of group cognition and negotiated belief structures. In addition, Axelrod (1976), Bonham et al. (1988), Weick and Bougon (1986), and Bougon, Weick and Binkhorst (1977) all discuss some form of collective cognitive mapping. Therefore, the latter authors have a measurement focus absent in many of the other articles which are more conceptually oriented.

The empirical record. Given the conceptual appeal of collective belief structures in the decision making literature, what empirical evidence supports the existence of shared group interpretations? Unfortunately, there is only a small amount of extant research. In a field study investigating how top decision makers in Ohio mental health boards interpret the Mental Health Act of 1988, Panzano (1992) found mean rater reliability statistics which supported the notion that organizational frames exist as shared perceptions among members. Interestingly, convergence was higher for framing variables such as threat and opportunity than other perceptual measures (e.g., slack). According to Panzano, these results suggest that managers in mental health boards share interpretations about how key issues impact their organization.

In another empirical study, Walsh et al. (1988) used realized coverage (the breadth of perspectives voiced during a discussion) and realized consensus (shared representations) to operationalize their notion of a negotiated belief structure. The analysis of 713 product decisions made by 29 groups of graduate business students in a complex, simulated business environment indicated that coverage and consensus were systematically related to product and firm performance. Specifically, shared agreement around a few schematic dimensions was associated with superior brand performance. Walsh and colleagues concluded that the linkages between the coverage and consensus variables and decision performance point to the validity of the negotiated belief structure construct.

Other studies have simply assumed the existence of a collective frame of reference (Isabella, 1990) or incorporated the metaphoric concept of a group belief structure into a theoretical framework (Innami, 1992) without attempting to demonstrate its existence.

The methodology. The primary measurement technique that has been identified to capture shared group interpretations is collective cognitive mapping (Axelrod, 1976; Bonham et al., 1988; Weick & Bougon, 1986; Bougon et al., 1977; Langfield-Smith, 1992). Axelrod (1976) points out that a collectivity can have a cognitive map either through the mechanical combination of beliefs or through a collectivity speaking for itself. Other researchers have identified various forms of collective cause maps including assemblage, composite and average maps (Weick & Bougon, 1986).

In addition to cognitive mapping at the aggregate level, other measurement approaches have also been employed. Fiol (1993) utilized a combination of content, structural, means/confidence interval, and discriminant analysis in order to explore the emergence of collective patterns of interpretation in the communication logs of a new-venture team.

Team Dynamics/Performance Domain

Using the shared mental model construct. Group mind-like notions have not only been hypothesized to affect the decision making process and decision outcomes of organizational elites, but team dynamics and performance as well. To illustrate, the model proposed by Bettenhausen and Murnighan (1985) posits that group interaction will proceed in a well-coordinated manner if group members’ definitions of the situation are also alike. Similarly, shared situation models assure that all participants are solving the same problem and help exploit the cognitive capabilities of the entire crew (Orasanu, 1990).

In addition, shared mental models are assumed to enhance the quality of teamwork skills and team effectiveness (Cannon-Bowers & Salas, 1990; Cannon-Bowers et al., 1990, 1993; Orasanu & Salas, 1993). Specifically, it is hypothesized that the greater the overlap or commonality among team members’ mental models, the greater the likelihood that team members will predict the needs of the task and team, adapt to changing demands, and coordinate activity with one another successfully (Cannon-Bowers et al., 1993; Converse, Cannon-Bowers & Salas, 1991). Teams who share mental models are expected to have common expectations of the task and team, allowing them to predict the behavior and resource needs of team members more accurately (Cannon-Bowers & Salas, 1990).

While the decision making references in Table 1 primarily discuss a shared understanding/belief/frame of some external phenomenon (e.g., strategic issue), many team dynamics/performance researchers refer to commonality in the manner in which group members work together. Cannon-Bowers et al. (1993) argue that models of the team (e.g., knowledge, skills, and abilities of teammates), as well as models of the task (e.g., familiarity with equipment and strategies), can be shared. In a similar way, Levine and Moreland (1991) divide socially shared knowledge within work groups into knowledge of the group, knowledge about group members, and knowledge about work. According to Orasanu (1990), shared situation models are composed of shared understandings of the problem as well as team member roles. Researchers such as Rentsch and colleagues (1993) focus exclusively on how members view the way that the team operates together.

As with the decision making references, there are differences among authors regarding just what it is that team members hold in common. Converse et al. (1991) state that the two aspects of mental models that should be shared are accuracy and the degree of detail or specificity. Cannon-Bowers et al. (1993) are very explicit about stating that it is expectations rather than the mental models themselves that are held in common, while Orasanu and Salas (1993) posit that organized knowledge is what is shared among team members. In addition, Weick and Roberts (1993) propose that the interrelating of social activities embody the collective mind.

The empirical record. Unfortunately, there is only a small amount of relevant empirical research relative to team performance, most of which is indirect and merely suggestive at best. In fact, in the team performance area, the concept of shared mental models is often merely invoked post hoc, to help describe and interpret team phenomena. While researchers have admitted the need for direct experimental evidence (Converse et al., 1991; Orasanu & Salas, 1993), few have provided any.

An analysis of the behavior of cockpit crew members as they coped with emergencies in a simulated flight by Orasanu (1990) found that effective crews managed their resources in anticipation of difficulties and better articulated their plans and strategies compared to ineffective crews, which lacked plans and behaved reactively. According to Orasanu (1990), the communication patterns of effective crews (e.g., explicitly defining the problem, articulating plans and strategies) as opposed to ineffective crews reflected their building of shared models for the specific problem. Similarly, in a study by Kanki, Lozito and Foushee (1989), the conversational patterns of high performing crews was interpreted post hoc as evidence of better shared team models on the part of the good crews (Orasanu & Salas, 1993).

Using two methods to assess teamwork schemas, namely multidimentional scaling and free hand concept maps, Rentsch et al. (1993) found that high experience individuals showed great consistency across the two different schema representations. According to Rentsch and colleagues (1993), this congruity suggested that high experience participants had well articulated teamwork schemas which they generalized to new group situations.

Mitchell (1986) found that working relationships among team members improved when internal frames of reference were shared among team members. Specifically, work teams whose members understood and respected one another’s “alignments” or internalized frameworks seemed to be able to accurately predict and explain other’s beliefs and actions, as well as develop expectations concerning other team members. Although the notion of a team mental model was not explicitly mentioned by Mitchell, this study can be added to the list of studies implicating a shared mental model concept.

Stronger evidence concerning shared cognition was provided by Hutchins (1991). Using computer simulation modeling and a connectionist framework for thinking about cognitive phenomenon at the group level of analysis, Hutchins’ (1991) results show that the cognitive properties of groups can differ from those of their participating members. Specifically, the simulations suggested that, even when holding the cognitive properties of individuals constant, groups may display different cognitive properties, depending on how communication is organized within the group over time. The cognitive properties of groups were produced as a result of an interaction between structures internal and external to individuals. It should be noted, however, that computer models are abstractions from real systems, and to the extent that important aspects of group phenomena are not modeled, the results are less informative.

The methodology. Researchers in the team performance area have articulated the need to develop techniques for measuring shared mental models. According to Converse et al. (1991), determining the overlap between team member’s mental models is the most difficult measurement problem imposed by the shared mental model hypothesis. These difficult measurement issues will no doubt require that researchers be creative in considering the realm of possibilities. For example, Rentsch et al. (1993) operationalized core team work schemas by using both multidimensional scaling and free hand (free of researcher influence) concept maps. Utilizing a variety of different techniques would most likely offer the best hope of capturing the team and task dimensions of collective mental models.

Conclusions

Several points can be highlighted regarding the references listed in Table 1 from both the decision making and group dynamics/performance domains. First, the frequent usage of numerous group mind-like terms from a variety of areas relevant to organizational science speaks to the popularity and widespread acknowledgment of the potential usefulness of the team mental model concept. Second, in addition to different terminology, the meanings attached to the notion of a shared reality are quite different. Although similarities can be argued, it is clear that researchers interpret the concept in diverse ways, especially with regard to the pivotal issue of what is shared among team members.(2) These conclusions reflect the need for integration.

Third, it is evident that empirical research of shared mental models has substantially lagged behind descriptive and theoretical work. This discrepancy may, in large part, be due to the methodological problems of measuring cognition at the group-level. In addition, given the very recent resurgence of interest in this area, one would not expect a plethora of empirical data at this stage of the research agenda.

However, it is also likely that the absence of a clear understanding of key definitional and conceptual issues has also contributed to the lack of empirical investigation. If studying shared cognition in groups is to be one of the most promising areas in sociocognition research as it is predicted to be (Levine et al., 1993), then theoretical clarity must be achieved. The following section begins to lay a foundation for a team mental model construct.

An Analysis of the Team Mental Model Concept

The reader might notice that we favor the term “team mental models” over alternatives to be found in the literature. We prefer this term for several reasons. First, it conveys that the locus of interest is on team dynamics and team functioning. We are not attempting to deal with other assemblies of individuals (e.g. groups, crowds, mobs, cultures). In a related manner, our view of team mental models presumes what is being shared and operating is among team members as a collectivity. That is, it is not enough to envision shared cognitions among dyads (something that the alternative phrase “shared mental models” allows for). Finally, it does allow for the notion of multiple sets or levels of shared knowledge. As described in our definition, this is a necessary part of the concept.

We do view the notion of a team mental model as a construct rather than a metaphor. Metaphors are statements in which some less clearly understood object or event (A) is compared to a more clearly expressed object or event (B) (Lakoff & Johnson, 1980). Some of the most popular metaphors used in organizational science include the notions of organizations as “machines” (e.g., Koch & Deetz, 1981) and “garbage cans” (Cohen, March & Oldsen, 1972). On the other hand, constructs are “terms which, though not observational either directly or indirectly, may be applied or even defined on the basis of the observables” (Kaplan, 1964, p. 55). Although metaphors are useful as heuristic devices and may serve as precursors to theories, a higher standard is required of constructs in that they “must sufficiently tap the domain of the phenomenon in question” (Bacharach, 1989, p. 507). Indeed, constructs are the building blocks of theory construction.

When analyzing emerging concepts and novel ideas, distinguishing between metaphors and constructs helps to better clarify the phenomenon under question. In our view, team mental models go beyond a figure of speech used to capture similarities among team related dynamics. Rather, they can be better classified as a “broad mental configuration of a given phenomenon” (Bacharach, 1989, p. 500).

Our selective review of shared mental model research in the organizational literature and the variety of definitions found there demonstrates clearly that investigators do not agree on its essential nature. This is often felt to be a “fatal flaw” for constructs used in developing or testing theory (e.g., Osigweh, 1989). However, we would prefer to interpret this state of affairs as a reflection of both the elusive nature of mental phenomena generally (Rouse & Morris, 1986) and the novelty of this approach in the study of organizational processes and outcomes. Thus, we wish to treat the apparent confusion regarding team mental models as an “opportunity” rather than a “threat” for the field. However, for investigators to make the most of this opportunity, several issues of conceptual definition and operationalization must be resolved.

We have selected several problematic issues that became salient to us as we conducted our review of the literature. These are outlined below. As will become evident, some of these will be more difficult to resolve than others.

The Content of Mental Models

It is the goal of any model to be a valid and efficient representation of some phenomenon (Klimoski, 1991). When it comes to the team mental model (TMM) construct, an important issue to address is just what is being modeled and shared by team members. In this regard there is no shortage of options; but there is no consensus either.

Several writers appear to focus on the interpretation of the stimuli that are impinging on individuals in a team context. In this view, what is being modeled is the latter’s attempts to make sense out of them. Specifically, TMMs are felt to be mental representations of: the nature (demands) of the tools/technology the must be employed by the individual or team (e.g., Cannon-Bowers et al., 1993); the team task that is to be performed (e.g., Cannon-Bowers et al., 1993); a particular problem facing the team as it is performing its task (e.g. Fiol, 1993); the knowledge, skills, abilities or other attributes relative to team functioning that is possessed by individual group members or the behavior patterns of team members, especially as they interact with one another (e.g., Rentsch et al., 1993). In addition, team mental models may also reflect representations of environmental events and projected future states (Wellens, 1993).

All the possibilities noted above have in common a theme of team members trying to conceptualize what is going on about them. They all imply an assessment of stimuli in situations. In the language of cognitive psychology, team members are developing and using categories in their interpretation. For instance, if a model references a tool to be used, the category “difficult to use” might be invoked. If it references a problem facing the team, the category of “threat” might come to mind.

In contrast, other writers appear to emphasize what team members are to do about what they see or apprehend. They write in terms of models of action (Brehmer, 1972). In particular, team mental models are often treated as internalized decision-choice/behavioral routines or “scripts” for action. For example, Weick and Roberts (1993) refer to the collective mind in terms of patterns of heedful interrelating in ongoing social processes. In addition, Cannon-Bowers et al. (1993) describe the content of team mental models as “task work” and/or “team work”. Models of task work refer to an understanding of the activities and action sequences that both the individual and the team collectively must carry out to perform the team task. In contrast, models of team work would include a conceptualization of the need to communicate with one another, compensate for one anothers’ weaknesses, how much mutual performance monitoring to carry out, and the type of internal coordination strategies generally needed for this team to function effectively.

More systematically, when thinking about the possibilities of team-relevant mental models for action, one can envision several possibilities. Clearly, an effective team has individuals who can perform their particular assignment well. This means that each individual needs a well developed response pattern to be carried out relatively independently in that assigned domain. Similarly, team work usually requires the coordinated actions of team member dyads of different composition from time to time. A particular team member must have a conceptualization of what is expected of him or her by each team member for each to jointly succeed. Finally, aggregating to the level of the team, we would assert that a mental model must also exist for those occasions where all members must work together in coordinated fashion. In this regard, Rentsch et al. (1993) used a multidimensional scaling technique to get at the mental models of novice and experienced individuals. She found that the latter had a knowledge structure (“team work schema”) that reflected fewer and more abstract dimensions. She also noted that there was more commonality among experienced individuals in their structures as if each had evolved a similar “theory” of effective team work.

Some writers imply that the content of mental models for action would also subsume how team members should interact outside of strictly task work. In this regard, mental models are assumed to cover the whole gamut of what would be considered appropriate behavior in work place relationships (Gabarro, 1990). For example, one might have an internalized structure for responding to on/off the job social opportunities or demands.

Phrased this way, the potential content of team mental models seems to include what has been traditionally termed “role expectations” (Katz and Kahn, 1978) or group norms (Feldman, 1984). This is an observation that has not been acknowledged in the team mental models literature. To illustrate, Bettenhausen and Murnighan (1985) studied newly formed decision making groups and were able to document the development of what they called team norms. Starting with past experience outside of the group and depending on the degree of similarity, behavior routines were either taken for granted or negotiated as group members interacted. All groups quickly reached what the authors call a collectively produced frame of reference. They felt however, that this definition of the situation (and how they as new team members should behave in it) was distinct from what they might have used for developing this interpretation. The latter would be a tool for sensemaking (e.g., how can I discover what is appropriate in this situation). It is easy to see from this example how mental models are related, but not identical, to the concept of norms. They go beyond consensus around actions and involve a conceptualization of the bases for that action.

It should also be noted that we view team mental models as being similar, but distinct from the notion of culture. Team mental models are emergent characteristics of the group which reflect organized knowledge and the tendency of individuals to categorize what they “know.”

The discussion above might be interpreted as a mandate for investigators to make a choice as to whether mental models represent theories of situations or of actions. This would be wrong. There is no reason why they can’t be conceptualized as involving both, as long as we are clear about it. We would accept the notion of multiple mental models (Young, 1983).

The Form of Mental Models

Aside from content, an issue that needs to be resolved is the form that team mental models take. Once again there is no shortage of views on how investigators have conceptualized this “mental architecture”.

Most simply, it is possible to view such models as collections of declarative and procedural knowledge. Thus a shared model is shared knowledge (Resnick, 1991). Such knowledge could be about traditions, customs or other aspects of culture as well (Levine & Moreland, 1991; Sternberg, 1979).

However, most writers seem to require that a model involve (at least) organized knowledge (Orasanu & Salas, 1993), involving meaningful patterns (Cannon-Bowers et al., 1993). As highlighted earlier, research on organized knowledge structures is fairly well developed in social/cognitive psychology, so it is not unusual for concepts like “schemas” and “scripts” to be used as a way of characterizing team mental models. To illustrate, Cannon-Bowers et al. (1993) see team mental models as knowledge that is organized into meaningful patterns based on some decision rule, whether it be categorical membership, time to execute or cause/effect.

This convergence notwithstanding, there is a great deal of uncertainty as to what such structures look like (Rouse & Morris, 1986). Are they organized spatially or verbally? Are they representational or abstract? In the case of the former, this may be in terms of exemplars. The latter might implicate concepts (e.g., “good co-worker”) or principles (e.g., “avoid”).

A great deal of work on mental models has been done by those interested in studying human performance. In the work of Rasmussen (1979) for example, potential forms of a mental model (of the task/system to be monitored and manipulated for good performance) have included physical form, physical function, functional structure, abstract function and functional meaning or purpose. These appear to represent a mixture of spatial/verbal and concrete/abstract types.

One might argue that the form that a mental model takes may be a function of the methods used by an investigator to measure it. In this respect, what we think we know about form might be an artifact of method. For instance, Rentsch et al. (1993) had subjects sort into categories 100 adjectives potentially relevant to “team-work”. They then had subjects rate the meaning of each adjective in each category relative to the other categories. Their data were represented in terms of a diagram with distances between categories implying perceived similarities. A similar approach (pair comparison judgements) was used by Kraiger (1993) to map experts’ understanding (mental models) of the potential stimuli and responses of an air traffic control task. In both cases, useful graphic representations were obtained which appear to convey how subjects organized (conceptualized) the material given to them as stimuli.

As interesting as these may be, the question remains: is this portrayal a valid “picture” of how these same individuals characteristically store and mentally process the domains of interest? We don’t know. The same normative question might be asked of the results of other investigators using such diverse mapping tools as the Kelly Reperatory Grid (Kelly, 1955; Reger, 1990), cause maps (e.g., Fiol & Huff, 1992; Weick & Bougon, 1986), or verbal protocols (e.g., Martin & Klimoski, 1990).

It may be that shared mental architecture changes over time. Thus, members of a newly formed group start out with an abstract, diffuse or general model and specificity increases with experience. McClure (1990) takes this view and argues that a “collective mind” emerges in all groups but the form that it takes depends on the group members’ experiences with one another. Therefore, the manifestation or character of a shared mental model may reflect the state of group development as well as investigator methods.

A special aspect of the team mental model form issue is that of hierarchy. As noted earlier, work in cognitive psychology generally, and in person perception in particular, has been convincing in establishing the fact that knowledge structures are themselves organized into higher order (more abstract or general) or aggregate (more inclusive) structures (Rosch, 1978). Brewer (1988) suggests that individuals tend to use the most specific level of abstraction that is available and adequate for effective categorization (interpretation). Rosch (1978) refers to these as the basic level. Categories that are more concrete are then referred to as subordinate. Those at a higher level of abstraction would then be termed superordinate. This suggests that, operationally, an investigator must try to uncover what level of abstraction is involved in the mental models that team members typically use to approach a situation (Porac & Thomas, 1990). For example, Wellens (1993) discusses the need to measure the degree of overlap in immediate, intermediate, and long-term situation awareness zones held by group members.

The concept of hierarchy also allows for the notion of a “Meta-mental model” concept (analogous to meta-cognition) to explain how team members “go” from one level of abstraction (lower level category) to another (higher level category). For example, the notion of “framing”, may not be capturing the use of a particular mental model so much as the application of information processing “rules” that encourage a decision maker shift to more abstract levels of a hierarchy of models (Fiol, 1993).

The Function of Mental Models

The resolution of the issue of form may ultimately be tied to a better understanding of why team mental models come about. Writers dealing with person perception and attitudes (e.g., Fiske & Taylor, 1991) stress that categorical processing comes about because it is fast and efficient. It just takes less mental effort. Referring to the research in the area of human performance, a convincing case can also be made for taking this functional perspective. According to this literature, mental models come about in order to allow individuals to predict future events, to determine the causes of events and phenomena, and (especially) for choosing appropriate courses of action (Rouse & Morris, 1986). They are basically, in the service of sensemaking and appropriate action.

It should be noted that although shared mental models are often viewed as functional (e.g. facilitating coordination, fostering efficiency, promoting predictability), they have their “dark side” as well. Over reliance on shared information, with all group members processing the same knowledge in the same ways may lead the group to under-utilize the resources of the team (Levine et al., 1993). Echoing the same concerns, Cannon-Bowers et al. (1993) state that the knowledge or expectations of team members may overlap to the point where shared mental models become a liability, and the potential for individual contributions is lost. In fact the concept of “groupthink” has been coined to reflect the circumstances where there is too much similarly in team members’ thought processes or when they refuse to abandon consensually validated but essentially incorrect views of the word. According to Fiol (1993), group members must simultaneously agree and disagree in order to maintain a balance between unity and diversity. Similarly, Wellens (1993) suggests that optimal group situation awareness is achieved when enough overlap occurs to maintain group coordination while allowing enough division to maximize coverage of the environment. In other words, completely overlapping team mental models are viewed as dysfunctional with regard to team performance.

Adopting such a pragmatic perspective implies that the form of mental models would be tied to the context or task domain involved. That is to say, how elaborately we conceptualize something may well depend on the complexities of the phenomenon itself. It might also be affected by the complexities of responding to or otherwise dealing with it as well. In this regard, Rouse and Morris (1986) make a case for decision maker “discretion” as a driving factor. How much discretion an individual has in responding to an event would provide a clue as to how (complexly) it is conceptualized. When there is little discretion, that is, where the task compels a limited range of behavior, mental model formation and use would be a function of the properties of the task. Thus, if it is simple and actions are highly scripted (e.g. as in assembling a gun), one’s mental model should reflect this in its directness and simplicity. In contrast, where there is a lot of discretion (such as when one must decide if the situation calls for the application of work values), mental models are likely to be quite complex in structure and may involve many levels of conceptualization.

This task based rationale is implied by Orasanu and Salas (1993) in their discussion of team mental models. While their work focusses on teams that must make decisions in real time or action situations (e.g. command and control teams), they too favor a contingent answer to the issue of form. In routine situations, they imply that team mental models exist to facilitate the efficient interplay of team members. Thus, they would take the form of scripts, emphasizing procedural routines. Models would be fairly direct and concrete. For effective performance in non-routine situations however, they feel that effective teams must have well developed/rehearsed “shared problem models”. In most cases, these would be complex and embedded in a hierarchic (contingent) system. Duffy’s (1993) emphasis on the team goal as providing a context or “processing objective” for team member information processing activities is tacit support for such a contingent view as well.

Accepting this view would require that we tolerate a contingent answer to the question of mental model form. It would depend (largely) on the function (mission/task/subtask) of the team and its members.

Not only may task characteristics be instrumental in affecting the form of mental models, but the importance of the task or mission to individuals in the group might also influence how elaborately the team mental model is conceptualized. That is to say, individual needs and aspirations can color and affect the complexity of team mental models. One might expect, for example, that members who view a group task or mission as important may have a more integrated and detailed model than members who perceive the task as personally unimportant. Experience with the group may also have implications for the complexity or abstractness of individual and team mental models.

In addition, individuals who have internalized the group’s goals as their own may develop more complex mental models than those with low levels of commitment. Furthermore, personality characteristics (e.g., high need for integration; cognitive complexity) may affect how elaborate an individual’s mental model might be, with implications for how complex and structured a team mental model can become. The point is that person issues, as well as task characteristics, need to be considered in addressing the issues of the form and function of team mental models.(3)

What Does it Mean to “Share” a Mental Model?

In developing an understanding of what is shared among group members, it is important to recognize the ambiguities inherent in the term “sharing.” Specifically, sharing can mean “having in common,” as when group members share the use of equipment, or “dividing up” as when group members share the workload (Cole, 1991; Resnick, 1991). In addition, Cannon-Bowers et al. (1993) discuss the notion of sharing in terms of “overlapping” knowledge and expectations.

Based on these different usages of the term “shared,” it appears that team mental models can refer to a cognitive representation that is identical among team members (e.g., common knowledge), a distributed configuration of representations (no overlap), or to a configuration of overlapping representations among group members. Cannon-Bowers et al. (1993) state that shared mental models do not imply identical mental models; rather, team members hold compatible models that lead to common expectations. However, much of the literature relating to team mental models is generally vague in specifying which meaning of “shared” is being discussed. It may be the case that newly formed groups begin with sharing as the division of cognitive labor and then evolve to a state in which sharing consists of overlapping information, with some knowledge held in common by all team members.

In their review of the social foundation of cognition, Levine et al. (1993) identify many ways in which we might construe mental models as “shared”. For example, people in groups frequently verbalize their thoughts to others. In this sense, conceptualizations are shared. On the other hand, individuals often come to rely on a group to help interpret a given situation. This is another kind of sharing. Thus, through these and other mechanisms, cognition is a result of collaboration with others. Levine et al. use the term “intersubjectivity” to reflect the shared understanding of what is being discussed or worked on, irrespective of the precise mechanism that brings it about.

On one level, “to be shared” might be based on an actuarial assessment. In this case, an investigator is describing the state of nature with regard to how mental model information gathered from different team members is or is not similar. Thus, a team with a shared mental model is one where most, if not all, of the people involved think about a phenomenon or situation in a very similar manner (e.g., Cannon-Bowers et al., 1993). This seems straight forward enough. But if this view is taken, we would still have to deal with such operational questions of what is the mental model, how completely must the mental models of two individuals overlap to be considered the same, or what proportion of members of a team must be congruent in their thinking for it to be considered having a shared model.

On the other hand, one could take a phenomenological approach to the notion of sharing. This implies that there must be some level of awareness by and among group members regarding how they interpret tasks, situations and events. A statistical similarity is not enough.

To be more specific, a model would be shared only if one or more team members believed this were so, and acted on this belief. Assuming the validity of this belief, we are now talking about the existence of a team mental model only to the extent that most of its members share the same belief. To the extent that there was ignorance of or disagreement regarding the distribution consensus, we couldn’t speak of a team mental model.

But there is still an additional complexity. A phenomenological approach might further require that each group member be aware of the level of awareness of other group members regarding the similarities and differences of their mutual thought processes. This would require that mental models must be consciously and explicitly negotiated and agreed upon to be shared (Levine et al., 1993; Fiol, 1993). If team members are ignorant of similarities and/or if they do not know what their colleagues “know”, there is not a team mental model. Consistent with this view, Larson and Christensen (1993) assert that group level problem identification only occurs when members become aware that others in the group also perceive the problem. Just having every member independently recognize a problem is not sufficient for identification at the aggregate level.

Of course the correctness of each approach might be assessed empirically, perhaps examining how different operationalizations would relate to our ability to predict such things as team member behavior and satisfaction and team performance. However, it would also be important to resolve this at the conceptual level.

In this regard, we take the position that actuarial similarity is a necessary but not sufficient condition for saying that a shared model exists. Given that the postulated beneficial effects of team mental models stem from the way that they facilitate coordination and the allocation of team resources, we feel that some level of awareness is necessary. This is because it seems highly unlikely that a team member would behave toward another in any performance setting of consequence without thinking first, and even validating his or her assumptions. It is only after explicitly raising expectations and/or shared knowledge, that operative mental models would prevail.

The Antecedents and Development of Team Mental Models

Up to now, we have been talking about the essential nature of team mental models. However, in the next two sections, we enter into a discussion on the construct’s antecedents and consequences.

Throughout our examination of the concept, we have alluded to the path that the development of team mental models might take. By way of recap and elaboration, we see the following possibilities.

Just being made a member of a team and being given some understanding of why the team has come about is likely to start the process of mental model development. In spontaneously formed teams, diffuse and nebulous formative factors (e.g., resonating to a “cause”, perceived similarity based on gender or style of clothing, etc.) would predispose team members to invoke and use patterns of thought organization (e.g., “we are egalitarian in our relations toward one another”). On the other hand, in deliberately formed teams, the method of team member recruiting and selection might insure shared perspectives. Indeed, in many work organizations, selection criteria are such as to insure that there would be similar team member orientation toward work and working, based around values and work habits (Klimoski & Jones, 1994)

New groups typically go through a “forming” stage (Tuckman, 1965). By this we mean, there is a period of time when team members spend energy to elicit from and share with others how they wish to work together. This usually involves the discussion and shaping of perceptions of the tasks to be performed by the team as well as the problems to be encountered by the team. Sometimes this sharing and the efforts to bring about a consensus leads to conflict (“storming”). Diverse views and preferences often clash. But most models of team development ultimately refer to this as a transition phase. Thus, there comes a point where such turmoil ultimately gets resolved. Through mutual adjustment and some member attrition (Schneider, 1987), team members, usually reach, at a minimum, some understanding of the nature of the team, its task and the rules that will govern team-member behavior (“norming”). Perhaps, it is only at this later stage that we might say that there exists some deeper sense of shared mental models. According to McClure (1990), groups in the forming stage of development will manifest a less detailed or refined “collective mind” than groups with a longer history together. Specifically, the group mind in regressive or immature groups is postulated to be only equal to or less than the aggregate of its parts, while it assumes an identity greater than the sum of its parts in generative or mature groups.

In most work organizational settings, the origin or refinement of mental models occurs in training. New employees, singly or in cohorts, are exposed to formal instruction designed to inculcate declarative or procedural knowledge. Among both public and private sector organizations in the U.S., the investment in such training is enormous, and consequently the approaches used vary considerably (Goldstein, 1986). At times, training is likely to focus explicitly on conceptual understandings or causal relationships. That is, it is probably creating mental models (even if not shared). Thus, instructions on how individual team members are to conceptualize a problem to be overcome or on how to work together to accomplish an objective may be part of the instructional design. However, these outcomes are usually not emphasized. Instead, promoting individual learning or individual or team performance might be the goal.

Of course there are any number of factors that would affect the speed and course of team mental model development (Feldman, 1984). For instance, as noted, pre-selection (by the company) and self selection (or anticipatory socialization; Louis, 1980) will have the effect of accelerating the process. So would experience of early team successes or failures. These serve to clarify how people think about issues. And of course, the experience of formal training as supplied by the organization would play its role.

Teams usually experience a change of membership during its life span. In a growth scenario, new persons are added. But team members might come and go based on rules of personnel rotation or termination or attrition. All this will also affect shared mental model development. In effect, team life cycle dynamics will make a difference.

This difference might not only affect the probability of a mental model coming into existence and its rate of development; it is likely to influence the content and form of such models as well. Thus, in stable teams, one would expect very wide spread, complex but team-specific models. In those units experiencing frequent turnover, such models might be more delimited (e.g., dealing with only really important aspects of group life) and more parsimonious (are relatively simple). As a result of the infusion of new participants, such team models might be more generic (and perhaps more functional). It is interesting to note that in studying the effects of longevity on R & D teams, Katz (1982) found that communication patterns (perhaps guided by shared world views) ceased to be functional for team effectiveness and survival. The right people just were not talking to one another in order to get critical information needed by the team.

This notion of communication patterns also relates to the development of shared models. For instance, Orasanu (1990) contends that what she calls “shared situation models” evolve through specialized communications. Similarly, Innami (1992) argues that the emergence of group belief structures is only possible through verbal communication, and Donnellon, Gray and Bougon (1986) identify communication mechanisms related to shared meaning and organized action.

Cohesion may also contribute, directly or indirectly, to the emergence of mental models. Research indicates that members of cohesive groups are more likely than others to participate actively in conversations and engage in self-disclosure or collaborative narration (e.g., Owen, 1985). This increased communication may facilitate the development of team mental models. In addition, Langfield-Smith (1992), in a study designed to elicit shared cognitive maps, concluded that her experimental group was not cohesive enough to cultivate shared understandings. Therefore, the implication is that cohesion may be an important antecedent of team mental model development.

Finally, several authors have implicated other forms of social interaction in the production of team mental models. These include social influence tactics (Hustle & Pennington, 1992), information sharing (Stasser, 1992), participation (Walsh et al., 1988) and negotiation (Bettenhausen, 1991, Walsh and Fahey, 1986; Walsh et al., 1988).

The Consequences of Team Mental Models

On one level, we seem to have already dealt with this issue. Shared team models “work” by virtue of their capacity for allowing individual members to anticipate and predict the behavior of individual members and the probable behavior of the group (when there is occasion to operate as a group). This capacity, in turn, allows for the efficient and effective use of team member inputs.

While this rationale is reasonable, we also feel that it is worthwhile to consider additional consequences that may be involved in the operation of shared mental models.

Our concerns relate to the wealth of evidence that our cognitions vis-a-vis the cognitions of other individuals (especially close team members) potentially can set up important forces affecting such things as our self concept, our willingness to take risks and our confidence (Levine et al., 1993; Resnick, 1991).

This view might be illustrated by considering how the perception of shared mental models might serve to energize and motivate team relevant behavior. In this scenario, a team member believes in the fact that he/she shares a similar world-view with teammates. This perception produces an appreciation of similarity and commonality with the team. This in turn causes two things to happen. First, there is positive affect. A team member comes to like his or her colleagues. Second, there a greater propensity to trust other team members. Positive emotions and trust implicate group cohesion. Therefore, in addition to being an antecedent, cohesion may also be a consequence of team mental models. The existence of a well developed model, of positive affect and of trust all bode well for individual functioning in a team context. They allow us to predict high effort, coordinated actions, spontaneity, assertiveness, risk-taking, etc. Ultimately, team performance would be the beneficiary of such behaviors. The shared team model goes beyond influencing the predictability of others. It sets up a chain of effects influencing multiple determinants of team effectiveness.

While the above scenario is speculative, it is based on what we know about the effects of perceived similarity as investigated in other contexts (Berscheid, 1985). It also is consistent with the dynamics of the social foundations of cognition (Levine et al., 1993). Thus we have every reason to believe that similarity in world views (mental models of important aspects of the world) should work to set up many of the same processes. If this is the case, the operation of the team mental model concept would be powerful indeed.

Our View of Team Mental Models

We take the position that the concept of a team mental model is a useful one. It is more than a metaphor, but it is also more than is currently being portrayed in the literature.

Our definition of a team mental model has been implied in the analysis provided above. To be more explicit however, we will recapitulate its properties:

a. It is a hypothetical construct. By this we mean that what it refers to need not be identified or labeled as such by the individuals affected. But it captures a phenomenon in a way that is useful to social scientists.

b. It is an emergent characteristic of the group, which is more than just the sum of individual models. Although the measurement techniques used to index team mental models may be at the individual level, we are essentially dealing with a group-level phenomenon.

c. It reflects the tendency of individuals to categorize what they “know”. Team mental models represent efforts to simplify events or responsibilities in order to make them more tractable.

d. Mental models reflect organized knowledge. Usually this is in the form of a set of concepts stored and retrieved from memory in relationship to one another. Such organization may derive from presumed cause and effect linkages or they merely may reflect learned patterns. More to the point, while the organized patterns may be strictly spatial in nature and origin, in all likelihood, such knowledge is organized semantically.

e. It implies a variety of content. While life would be more simple if we could conclude that team models always implicate a specific domain, we feel that this is not the case. Further, it would be too delimiting a use of the construct. More specifically, we allow that the content of shared mental models might reference representations of tasks, of situations, of response patterns or of working relationships.

f. Team mental models reflect internalized beliefs, assumptions, and perceptions. That is, allowing for the impact of method and circumstances of measurement, they are really how the group members as a collectivity think or characterize phenomena.

g. Team mental models exist to the extent that they are apprehended (at least at some level of awareness) by team members. Coordinated action among team members does occur with out such awareness, but we would not use this construct to attempt to explain such phenomena. Although we recognize the possibility of highly automated and routinized behavior (Gersick & Hackman, 1990), we view this on the high end of the continuum of team mental model development.

Implications

As we noted in the outset of this article, team mental model-like concepts are being used with some enthusiasm by scholars in a variety of lines of investigation. These include the areas of team performance, strategic decision making, and training. Thus, we can hardly claim to have invented the notion. However, we do feel that our review and analysis does make a contribution in several important ways.

Highlighting Confusion in the Literature

If we have done nothing else, we have established that despite the popularity of the concept, many authors have been very casual in its application. Most disconcerting, many writers do not really define what they mean by a shared or team mental model. There is a surprising lack of definitional or conceptual clarity. Also problematic, when attempts at definition have been made, different authors have defined things in alternative (usually in incomplete) ways (e.g., Floyd & Wooldridge, 1992; Innami, 1992; Walsh & Fahey, 1986). In our view, part of the problem is that writers in a particular area (e.g., strategic decision making), often do not cite the literature in other areas that may be referring to the same concept of interest (albeit with a different name). To put it another way, heretofore there has not been much “cross-fertilization” (e.g., Weick & Roberts, 1993). Various writers seem to be “re-inventing the wheel”. By pointing out and reviewing these diverse literatures, we hope that this study will ameliorate this situation.

It is also true that when the team mental model is used somewhat systematically (e.g., Cannon-Bowers et al., 1993), there is not much in the way of empirical evidence to substantiate its nature. In most cases, the construct is implicated by inference. While it may be used as an explanatory construct, it is usually done after the fact.

Much of this may stem from the difficulty of operationally defining and measuring team mental models. As pointed out in our review, however, its not that there are no options available. To the contrary, there appear to be several possibilities. These range from multidimensional scaling (Rentsch et al., 1993) to cause map analysis (e.g., Weick & Bougon, 1986). Its just that the research needed to sort things out has not yet been done. Alternative measures or approaches have not been contrasted in the same study. On the other hand, given the conceptual confusion as to just what team mental models are, this lack of measurement work is not surprising.

This leads us to a second area where we feel that this paper has made a contribution.

Isolating the Nature of the Construct

When we started our review of the literature on the team mental model notion, we really were not certain whether we were dealing with a construct or a figure of speech (a metaphor) that various writers have found useful to convey their impressions of team related phenomena. It’s not that a metaphor has no place in scholarly writing. We feel that it does (see Lakoff & Johnson, 1980), and recognize that using a metaphor can be particularly effective in teaching or in communicating with non-specialists. It’s just that we have different (and typically more demanding) standards when concepts are to be used as constructs (Schmitt & Klimoski, 1992).

In any event, our own reading and analysis has convinced us that “there is a construct there”. Beyond reaching this conclusion, we feel we have isolated the nature of the team mental model construct.

Unlike the use of the term in a lot of the literature found in the organizational sciences, we clearly define what we mean. More importantly, we conclude that it is a multi-faceted construct (with aspects of content, form, etc). This paper has developed its theoretical roots in cognitive and group psychology. Finally, we have placed the construct in a nomological network of antecedents (developmental causes) and consequences.

If the notion of the team mental model is indeed a construct as we have asserted it to be, what then is its actual or potential contribution to particular domains within organizational science? That is to say, what is the unique value of team mental models and, how do they fit into existing theories of strategic decision making and team performance? Of course, these questions will ultimately have to be addressed empirically, but we offer some initial thoughts below.

In the area of strategic decision making, team mental models most likely have their greatest impact, not on the decision phase, but the implementation phase. Decisions can be made in the absence of a team mental model and despite differences of interpretation among individuals (Donnellon et al., 1986; Weick, 1979). Time pressure, for example, may force a group to arrive at a decision without members sharing perceptions and beliefs on the issues under consideration. It is even possible that team mental models may not emerge until after the decision phase if the team continues to interact and discuss concerns. However, team mental models could have a major influence on the implementation of a decision (providing that the team also has to implement what it has decided on). Teams that have well developed mental models may be able to implement their decisions more quickly and with fewer problems than teams that do not have collective mental models.

COPYRIGHT 1994 JAI Press, Inc.

COPYRIGHT 2004 Gale Group