Journal of Applied Psychology
2000, VoL 85, No. 2, 273-283
Copyright 2000 by the American Psychological Association, Inc.
0021-9010/00/$5.00 DOI: 10.I037t/0021-9010.85.2.273
The Influence of Shared Mental Models on Team Process and Performance
John E. Mathieu
Tonia S. Heffner
Pennsylvania State University
University of Tennessee at Chattanooga
Gerald F. Goodwin
Eduardo Salas and Janis A. Cannon-Bowers
Naval Air Warfare Center
Pennsylvania State University
The influence of teammates' shared mental models on team processes and performance was tested
using 56 undergraduate dyads who "flew" a series of missions on a personal-computer-based flightcombat simulation. The authors both conceptually and empirically distinguished between teammates'
task- and team-based mental models and indexed their convergence or "sharedness" using individually
completed paired-comparisons matrices analyzed using a network-based algorithm. The results illustrated
that both shared-team- and task-based mental models related positively to subsequent team process and
performance. Furthermore, team processes fully mediated the relationship between mental model
convergence and team effectiveness. Results are discussed in terms of the role of shared cognitions in
team effectiveness and the applicability of different interventions designed to achieve such convergence.
Increased technology has contributed to the complexity of many
tasks performed in the workplace, making it difficult for employees to complete their work independently. In response to the
technological advances, many organizations have adopted a team
approach to work. Teams are viewed as being more suitable for
complex tasks because they allow members to share the workload,
monitor the work behaviors of other members, and develop and
contribute expertise on subtasks. An abundance of research has
been conducted on the factors that contribute to high team performance (for reviews, see Gist, Locke, & Taylor, 1987; Salas,
Dickinson, Converse, & Tannenbaum, 1992). One variable that
has recently received much theoretical attention concerns the influence of team members' mental models on team-related processes and behaviors (Klimoski & Mohammed, 1994; Kraiger &
Wenzel, 1997; Rentsch, Heffner, & Duffy, 1994; Stout, Salas, &
Kraiger, 1996). The present research was designed to empirically
examine the impact that teammates' mental models have on team
process and performance in a dynamic and exciting laboratory
flight simulation.
A team can be defined as "a distinguishable set of two or more
people who interact, dynamically, interdependently, and adaptively toward a common and valued goal/objective/mission, who
have each been assigned specific roles or functions to perform, and
who have a limited life-span of membership" (Salas et al., 1992, p.
4). Research has identified numerous factors that affect teams and
has offered several models of team functioning (Guzzo & Dickson,
1996). Although these different models vary in details, they all
share an input-process-outcome (I-P-O) framework. Inputs to
such models are conditions that exist prior to a performance
episode and may include member, team, and organizational characteristics. Performance episodes are defined as distinguishable
periods of time over which performance accrues and feedback is
available. Processes describe how team inputs are transformed into
outputs, Outcomes are results and by-products of team activity that
are valued by one or more constituencies. Hackman (1990) identified three primary types of outcomes: (a) performance-including
quality and quantity, (b) team longevity, and (c) members' affecfive reactions. Although we recognize the importance of all three
types, for our purposes we will concentrate on performance
outcomes.
Empirical examinations of I-P-O models have demonstrated
their utility (e.g., Campion, Medsker, & Higgs, 1993; Gladstein,
1984; Guzzo & Dickson, 1996). However, the large number of
factors that influence outcomes has prohibited a comprehensive
examination of the model. Many variables that have been proposed
to influence team processes and thereby team performance have
yet to receive much empirical examination. Included among these
are members' knowledge and its organizational structure. This
oversight has occurred despite acknowledgement of the importance of knowledge organization for individual and team performance (Cannon-Bowers & Salas, 1990; Cannon-Bowers, Salas, &
John E. Mathieu and Gerald F. Goodwin, Department of Psychology,
Pennsylvania State University, University Park; Tonia S. Heffner, Department of Psychology, University of Tennessee at Chattanooga; Eduardo
Salas and Janis A. Cannon-Bowers, Training Systems Division, Naval Air
Warfare Center, Orlando, Florida.
Tonia S. Heffner is now at the Arm), Research Institute, Washington,
DC. Gerald F. Goodwin is now at the American Institutes for Research,
Washington, DC. Eduardo Salas is now at the Department of Psychology,
University of Cenlral Florida, Orlando, Florida.
The views, opinions, and findings contained in this article are those of
the authors and should not be construed as an official position, policy, or
decision of their respective affiliations. We thank Dana Born, Gwen Fisher,
Kirsten Hobson, Kara Ivory, Mike Trip, and Nikki Windefelder, who were
instrumental in the running of the teams and data coding portions of this
work.
Correspondence concerning this article should be addressed to John E.
Mathieu, who is now at the School of Business Administration, Department of Management, University of Connecticut, 368 Fairfield Road,
6-41MG, Storrs, Connecticut 06269-2041. Electronic mail may be sent to
jmathieu @sba.uconn.edu.
273
274
MATHIEU ET AL.
Converse, 1993; Klimoskl & Mohammed, 1994; Kralger & Wenzel, 1997; Rentsch & Hall, 1994).
Shared Mental M o d e l T h e o r y
The notion of shared mental models in teams has been used to
help explain team functioning for several years. According to
Cannon-Bowers, Salas, and Converse (1993), shared mental models help explain how teams are able to cope with difficult and
changing task conditions. In fact, researchers have suggested that
the ability to adapt is an important skill in high-performance teams
(Cannon-Bowers, Tannenbaum, Salas, & Volpe, 1995; Mclntyre
& Salas, 1995). The shared mental model theory offers an explanation of what the mechanisms of adaptability might be--that is,
how teams can quickly and efficiently adjust their strategy "on the
fly." The following sections provide more details regarding shared
mental model theory and its relationship to effective teamwork.
Mental Models
The term mental model has been used as an explanatory mechanism in a variety of disciplines over the years (see Wilson &
Rutherford, 1989). Essentially, mental models are organized
knowledge structures that allow individuals to interact with their
environment. Specifically, mental models allow people to predict
and explain the behavior of the world around them, to recognize
and remember relationships among components of the environment, and to construct expectations for what is likely to occur next
(see Rouse & Morris, 1986). Furthermore, mental models allow
people to draw inferences, make predictions, understand phenomena, decide which actions to take, and experience events vicariously (Johnson-Laird, 1983). For the purpose of this article, we
defined a mental model, in keeping with Rouse and Morris (1986),
as a "mechanism whereby humans generate descriptions of system
purpose and form, explanations of system functioning and observed system states, and predictions of future system states"
(p. 360). Hence, mental models serve three crucial purposes: They
help people to describe, explain, and predict events in their
environment.
Shared Mental Models
As we noted above, Cannon-Bowers et al. (1993) suggested that
teams that must adapt quickly to changing task demands might be
drawing on shared or common mental models. The rationale behind Cannon-Bowers et al.'s assertion was that in order to adapt
effectively, team members must predict what their teammates are
going to do and what they are going to need in order to do it.
Hence, the function of shared mental models is to allow team
members to draw on their own well-structured knowledge as a
basis for selecting actions that are consistent and coordinated with
those of their teammates.
To refine the concept of shared mental models even further,
Stout, Cannon-Bowers, and Salas (1996) suggested that the manner in which shared mental models operate is related to task
demands. Specifically, these authors argued that under conditions
that allow team members to freely communicate with one anothe r - t o strategize--shared mental models will not be very important. This is because the team can discuss its next moves and does
not need to rely on preexisting knowledge. However, under conditions in which communication is difficult--because of excessive
workload, time pressure, or some other environmental feature-teams are not able to engage in necessary strategizing. In this case,
shared mental models become crucial to team functioning because
they allow members to predict the information and resource requirements of their teammates. Hence, members are able to act on
the basis of their understanding of the task demands and how these
will affect their team's response. It is this ability to adapt quickly
that enables teams in dynamic environments to be successful.
Types of Shared Mental Models
Cannon-Bowers et al. (1993) and others have argued that there
is probably not a single mental model that must be shared among
team members. In fact, Klimoski and Mohammed (1994) contended that "there can be (and probably would be) multiple mental
models co-existing among team members at a given.point in time.
These would include models of task/technology, of response routines, of team work, etc." (p. 432). Rentsch and Hall (1994)
advanced similar notions and argued that team members' schema
similarity (a concept quite similar to mental models) could be
described in terms of both team work and task work. According to
Cannon-Bowers et al. (1993), complex tasks probably dictate that
multiple mental models be shared among members. Table 1 describes several of these mental models.
First, team members must understand the technology or equipment with which they are interacting. The dynamics and control of
the technology and how it interacts with the input of other team
members is particularly crucial for team functioning. Second, team
members must hold shared job or task models. Such models
describe and organize knowledge about how the task is accomplished in terms of procedures, task strategies, likely contingencies
or problems, and environmental conditions. Third, team members
must hold shared conceptions of how the team interacts. These
models describe the roles and responsibilities of team members,
interaction patterns, information flow and communication channels, role interdependencies, and information sources. The final
model that team members must share is the team member model.
This model contains information that is specific to the member's
teammates--their knowledge, skills, attitudes, preferences,
strengths, weaknesses, tendencies, and so forth. According to
Cannon-Bowers et al. (1995), such knowledge is crucial for team
effectiveness because it allows team members to tailor their behavior in accordance with what they expect their teammates to do.
The more knowledge team members have about one another (and
the more accurate that information is), the more efficient and
automatic this process can be.
Although the detailed breakdown of mental-model types depicted in Table 1 helps to clarify the nature of the underlying
knowledge structures, operationalizing four different types of mental models becomes unwieldy in a single study. The four types of
model described above can be viewed as reflecting two major
content domains: (a) task-related features of the situation (e.g., the
technology/equipment and job/task models) and (b) team-related
aspects of the situation (e.g., the team interaction and team models). This division is also consistent with the idea that teams
develop two tracks of behavior--a teamwork track and a taskwork track (see Mclntyre & Salas, 1995; Morgan, Glickman,
SHARED MENTAL MODELS AND TEAM EFFECTIVENESS
275
Table 1
Types of Shared Mental Models in Teams
Type of model
Knowledge content
Comment
Technology/equipment
Equipment functioning
Operating procedures
System limitations
Likely failures
Likely to be the most stable model in terms of content.
Probably requires less to be shared across team
members.
Job/task
Task procedures
Likely contingencies
Likely scenarios
Task strategies
Environmental
constraints
Task component
relationships
In highly proceduralized tasks, members will have
shared task models. When tasks are more
unpredictable, the value of shared task knowledge
becomes more crucial.
Team interaction
Roles/responsibilities
Information sources
Interaction patterns
Communication channels
Role interdependencies
Information flow
Shared knowledge about team interactions drives how
team members behave by creating expectations.
Adaptable teams are those who understand well and
can predict the nature of team interactions.
Team
Teammates'
Teammates'
Teammates'
Teammates'
Teammates'
Team-specific knowledge of teammates helps members
to better tailor their behavior to what they expect
from teammates.
knowledge
skills
attitudes
preferences
tendencies
Note. Adapted from "Shared mental models in expert team decision making," by J. A. Cannon-Bowers, E.
Salas, and S. A. Converse, 1993, in Individual and group decision making, by N. J. Castellan, Jr. (Ed.), Hillsdale,
NJ: Erlhaum. Copyright 1993 by Erlbaum. Adapted with permission.
Woodard, Blaiwes, & Salas, 1986). These researchers argued that
in order to be successful, team members not only need to perform
task-related functions well but also must work well together as a
team. Accordingly, we distinguished b e t w e e n - - a n d measured separately the extent to which teammates s h a r e - - t w o different types
of mental models, task and team, and then related them to indices
of team processes and performance. To our knowledge, empirical
research has yet to relate multiple types of shared mental models
to team effectiveness.
I-P-O Framework
The nature of teams and teamwork as represented by the I-P-O
model suggests that the impact of mental models on team effectiveness may be indirect and mediated by team processes. For
example, Kralger and Wenzel (1997) suggested that team processes are likely to be influenced by shared mental models. They
argued that "the areas in which the greatest impact is expected are
decision making and communication" (p. 79). Widely different
mental models suggests that teammates work toward different
objectives and predict different future system states, and therefore
have difficulty coordinating their efforts. Alternatively, highly
similar mental models would suggest that teammates work toward
common objectives and have a shared vision of how their team
will function. Thus, teammates with shared mental models will
easily coordinate their actions and be "in sync," whereas differences in team mental models would likely result in greater process
loss and ineffective team processes. Therefore, we hypothesized
that the convergence of both teammates' team and task mental
models would relate positively to all three team processes described more fully below. Furthermore, we hypothesized that
teammates would exhibit greater sharedness over time as they
gained experience both with the task and with each other. For
example, Rentsch et al. (1994) found that more experienced team
members conceptualized teamwork more concisely and coherently
than did less experienced members. Dyer (1984) advanced similar
conclusions and added that with increased experience teams are
able to better coordinate their efforts and therefore exhibit better
team processes and performance.
Many research studies have focused on the relationship between
team process and outcomes (e.g., Campion et al., 1993; Gladstein,
1984). This body of research has identified a myriad of variables
that can fall under the broad title of process variables. Although
some recent work has advanced different taxonomies of team
processes (cf. Fleishman & Zaccaro, 1992; Prince, Brannick,
Prince, & Salas, 1992), Klimoski and Mohammed (1994) have
argued that at least the following are important for linking shared
mental models with team performance: communication processes,
strategy and coordinated use of resources, and interpersonal relations or cooperation.
Roberts and O'Reilly (1976) found that communication frequency was related to increased performance across a variety of
tasks in fighter aircraft crews. Similarly, Foushee and Manos
(1981) found that air crews who communicated more frequently
performed better than crews who communicated less frequently in
a simulator environment. Kanki and Foushee (1989) found that air
276
MATHIEU ET AL.
crews who had previously flown together had significantly fewer
errors than air crews who had not, and the difference was attributable to the more seasoned air crews having better communications. Therefore, we hypothesized that team communication would
relate positively to team performance. In effect, then, we hypothesized that team communication serves as a mediator of the influence of shared mental models on team performance.
Kleinman, Luh, Pattipati, and Serfaty (1992) defined strategy as
"the mechanisms by which the team attains its performance"
(p. 184). Several researchers have noted the importance of strategy
formation for effective group performance (Gist et al., 1987;
Gladstein, 1984), particularly for complex tasks (Hackman, Brousseau, & Weiss, 1976). For example, Hackman et al. (1976) studied
the impact of strategy discussions on group performance and found
them to relate positively to performance on new tasks, yet negatively to performance on established tasks. They also found that
groups without instructions were unlikely to discuss performance
strategies, even for new tasks. Accordingly, given the emergent
nature of the task used and the fact that we encouraged teams to
strategize in the current study, we hypothesized that team strategy
would be positively influenced by shared mental models and, in
turn, relate positively to team performance.
The commonsense notion that members must get along and
cooperate with each other if a team is to be successful has proven
to be difficult to validate. For example, reviews of the relationship
between interpersonally based team cohesion and performance
consistently conclude that the empirical findings are mixed or
equivocal (cf. Bettenhausen, 1991; Salas, Bowers, & CannonBowers, 1995; Tannenbaum, Beard, & Salas, 1992). At issue here
is whether teammates' interpersonal relationships distract from or
contribute to team performance. Some interpersonal conflicts are
inevitable in teams, particularly ones that are placed in complex
challenging situations. Thus, while overly enhancing team members' interpersonal relationships may not enhance their task capabilities, the teams certainly will not be successful if members
refuse to work with one another. This is not to say that cooperative
or happy teams will always be successful. Rather, we submit that
the most successful teams will likely be those that combine a true
concern for one another with a collective commitment toward the
team task (Bettenhausen, 1991). Thus, we hypothesized that team
cooperation would serve as an important mediating role linking
shared mental models with team performance.
In summary, we anticipated that team and task mental models
could be assessed and distinguished. Furthermore, we hypothesized that the convergence or sharedness of teammates' mental
models, both task and team, would relate significantly to the
subsequent team processes they exhibited over a series of performance episodes. Moreover, we hypothesized that team processes
would relate significantly to team performance, and thereby mediate the influence of shared mental models on team outcomes.
Although we had hypothesized that team processes would fully
mediate the influence of shared mental models on team performance, we also tested whether there were any direct effects.
Finally, we hypothesized that with increased experience with each
other and the task, teammates would exhibit greater convergence
among their mental models as well as better team processes and
performance.
Method
Participants
Participants were 52 male and 60 female undergraduate students enrolled in psychology classes at Pennsylvania State University. Their ages
ranged from 18 to 29 years (M = 20.96, SD = 2.02). They participated in
the experiment in exchange for extra course credit and were randomly
assigned to 56 two-person teams. There were no significant effects of the
gender composition of teams on the relationships reported here, so we have
pooled all teams into a single sample.
Task Apparatus
Weaver, Bowers, Salas, and Cannon-Bowers (1995) recently noted that
"researchers interested in investigating team performance have begun to
employ low-fidelity networked simulations to gain an increased understanding of the various factors that might impact team performance"
(p. 13). They concluded, "Low fidelity simulation provides a particularly
effective method for furthering empirical knowledge of theory related to
team performance and processes" (Weaver et ai., 1995, p. 22). Brannick,
Roach, and Salas (1993) and Stout, Cannon-Bowers, Salas, and Milanovich
(in press) have successfully used such tasks to investigate a variety of
inputs and processes related to team performance. Furthermore, Baker,
Prince, Shrestha, Oser, and Salas (1993) reported results indicating that real
military aviators believed that dynamic cockpit-crew processes could be
well emulated using personal-computer-based simulations. Accordingly,
we used a low-fidelity personal-computer-based flight simulator task (Falcon 3.0; see Spectrum Holobyte, 1991) as our research platform.
The software is a computer-generated simulation program of the F-16
fighter fixed-wing aircraft. The program is extremely flexible and enables
one to modify several parameters of the simulation. For example, various
scenarios can be scripted and controlled in terms of flight plans, aircraft
weapons loadouts, the number of enemy and allied planes present, and
even the ability and temperament of computer-guided pilots. On one hand,
as compared with actual flight or with military and commercial airline
simulators, this task is an extremely low grade simulation. On the other
hand, as compared with common video and computer games, this task is
extremely detailed and complicated. The hardware included an IBMcompatible personal computer equipped with a joystick and keyboard and
video and audio recording equipment. A divider separated players, and
they communicated via microphone-equipped headsets.
Procedure
The experimental sessions ran between 2.5 and 3 hr and were conducted
in three phases. First, participants received an overview of the task and
watched an automated demonstration of the simulation. Second, they were
guided through a structured hands-on training program, conducted by a
highly skilled experimenter, that lasted approximately 45 min. One player
was trained to operate a joystick and had the responsibility of flying and
positioning the plane. The second teammate used a standard computer
keyboard and had the responsibilities of setting airspeed, calling up different weapon systems, and gathering information. Both players were able
to fire weapons.
The training program taught both individual task responsibilities and
basic team processes (e.g., coordination of activities), following the guidelines for effective team training outlined by Swezey and Salas (1992). A
three-step sequence was followed: Step 1 focused on how to fly the plane,
reach waypoints and destinations, and interpret information from a
heads-up display; Step 2 concentrated on training how to call up and fire
different weapon systems, how to read and to interpret radar information,
and how to use basic flight maneuvering tactics and strategies; and Step 3
had teams practice the skills they developed during Steps 1 and 2 and then
had them attempt a practice version of the missions they would complete.
277
SHARED MENTAL MODELS AND TEAM EFFECTIVENESS
The third phase of the experiment started at the conclusion of training.
Players completed an initial survey and then flew Missions 1 and 2, after
which they received a detailed summary of their performance in terms of
waypoints reached and enemy aircraft shot down. They then completed a
second survey that repeated the earlier administered measures, and the
sequence was then repeated for Missions 3 and 4 and for Missions 5 and 6.
Measures
Performance. After training, teams completed three blocks of two
missions each. Each mission lasted 10 min, or less time if the team was
shot down or crashed. During each mission, the team had three objectives:
(a) to survive--worth three points; (b) to fly a preset route that had four
waypoints, or destination markers--worth two points each); and (c) to
shoot down enemy planes--worth one point each. This scoring scheme
presented a strategic dilemma for teams, as the three objectives were
incompatible. For example, flying directly toward all waypoints placed the
team in great risk and left little time for fighting enemy planes. Alternatively, actively engaging enemy planes left little time for reaching waypoints. Finally, maximizing survivability by staying away from enemy
planes minimized the likelihood that they would reach waypoints, because
that was where the enemy planes were typically circling.
The series of six missions was programmed in blocks of two such that
although the specific aspects of each mission varied, the three blocks were
equally difficult (as pilot testing confu'med). The total distance between the
four waypoints was held constant across missions (although the distances
from one to the next varied both within and across missions), and on
average four enemy planes were present in each mission (specifically: 3,
5, 4, 4, 5, 3 for Missions 1-6, respectively). Furthermore, the types and
relative hostility of the enemy planes were balanced across the two missions within blocks. In short, although each of the six missions represented
a unique, complicated, and unpredictable sequence of events, all aspects of
the task were balanced over time. Team performance for each block was
measured simply by the points earned across two missions.
Team process. Team processes were rated independently by two observers who watched videotapes of the experimental missions. Each rater
watched two missions (one performance episode), after which he or she
completed the process ratings. We took care not to show the coders the
detailed mission feedback so as to minimize potential performance-cue
effects (see Martell, Guzzo, & Willis, 1995). We developed 21 items
intended to measure three dimensions: (a) strategy formation and coordination (6 items, e.g., "To what extend did the team plan together and
coordinate its efforts?"); (b) cooperation (4 items, e.g., "To what extent did
they cooperate well during the missions?"); and (c) communication ( l l
items, e.g., "To what extent was information about important events and
situations shared within the team?"). Each item was rated using a 5-point
scale that ranged from 1 (not at all) to 5 (to a very great extenO.
The intercoder reliabilities (i.e., correlations) for these scales were all
significant (p < .001) and ranged from .50 to .86 (M = .67). The
aggregate-scale internal consistencies were quite acceptable for both coordination (as = .85, .83, and .86 for Times 1-3, respectively) and
communication (t~s = .86, .82, and .81 for Times 1-3, respectively), but
not for cooperation (as = .55, .40, and .59 for Times 1-3, respectively).
Furthermore, although the three process dimensions are conceptually distinguishable, empirically they were highly correlated (Ms = .59, .71, and
.59, ps < .001, for Times 1-3, respectively). This degree of correlation
limits the extent to which unique effects can be observed in regression
analyses. In fact, we attempted to use the three scales separately in a
repeated measures multiple regression analysis (described more fully later)
and encountered signs of multicollinearity problems (e.g., several high
tolerances). Moreover, unrestricted maximum likelihood analyses of the
process variables, conducted separately at each time, consistently revealed
the existence of a single underlying dimension (p < .05). Consequently,
we combined all 21 ratings in to an omnibus team-process variable at each
time. The intercoder reliabilities were all significant and high (rs = .76,
.76, and .59, ps < .001, for Times 1-3, respectively) as were the aggregate
alphas (as = .91, .90, and .89 for Times 1-3, respectively).
Mental models. Task and team mental models were assessed using
teammates' individual ratings of the relationships between various attributes. For the task mental model, we first conducted a detailed task
analysis (cf. Baker, Salas, & Cannon-Bowers, 1998; Tesluk, Zaccaro,
Marks, & Mathieu, 1997) of the simulation using subject-matter experts
and technical documentation (e.g., Powell & Basham, 1992; Spectrum
Holobyte, 1991). This effort identified eight critical attributes for the task
mental model: (a) diving versus climbing; (b) banking or turning; (c)
choosing airspeed; (d) selecting and shooting weapons; (e) reading and
interpreting radar; (f) intercepting the enemy; (g) escaping the enemy; and
(h) dispensing chaff and flares. Although each of these dimensions can be
distinguished from the others, many are highly related. For example, during
training the participants were instructed to radically alter altitude and
direction as well as to dispense chaff and flares to avoid incoming enemy
missiles. To engage an enemy, participants were instructed to adjust
airspeed and modify direction and altitude in order to position their plane
behind the enemy, and also to select the appropriate type of weapon system
to use given their flight situation. Both of these examples would also
require the use of the radar to determine the existence of and distance from
an enemy plane.
The seven attributes for the team mental models were selected based on
a literature review of previous taxonomic research on teamwork dimensions (e.g., Fleishman & Zaccaro, 1992; Stout, Cannon-Bowers, Salas, &
Morgan, 1990) with emphasis on the current simulation context. Specifically, these attributes were: (a) amount of information, (b) quality of
information, (c) coordination of action, (d) roles, (e) liking, (f) team spirit,
and (g) cooperation. These particular attributes were selected because they
represent concrete indicators of the three targeted process dimensions.
Namely, the amount and quality of information were designed to tap the
communication dimension, coordination of action and roles were directed
at the strategy component, and the remaining three facets were directed at
team cooperation.
Participants were provided with two matrices (one for each mental
model) that listed the attributes along the top and side of the grid. The
definitions for each attribute were present along the side, immediately
following the attribute label. Respondents rated each attribute of the mental
model in relation to all other attributes for that model using a 9-point scale
that ranged from - 4 (negatively related, a high degree of one requires a
low degree of the other) to 4 (positively related, a high degree of one
requires a high degree of the other). The ratings were completed before
each of the three blocks of performance.
Mental model centrality and convergence. Task and team mental
models were analyzed individually using a network-analysis program
(UCINET; Borgatti, Everett, & Freeman, 1992). Among a variety of
available indices, UCINET provides an index of convergence (QAP correlation) between two matrices. This correlation was calculated between
the teammates' task mental models at each time period. Similarly, a QAP
correlation was calculated between the teammates' team mental models at
each time period. Thus, two mental-model convergence indices, task and
team, were produced for each team at each time. QAP correlations are
essentially zero-order correlations between the identical elements of two
mental-model matrices and therefore range from - 1 (complete disagreement or counter-sharedness) to 1 (complete sharedness). Moreover, like
any correlation, their magnitudes and the stability of the estimates are
complex products of the number of attributes rated and the diversity of
underlying relationships (i.e., whether the true cell entries would be homogeneous or variable).
Results
The results are presented in two sections. The first section
details the hypothesized mean changes over time for each variable.
278
MATHIEU ET AL.
Table 2
Variable Correlations and Descriptive Statistics Per Time
Variable
Time t
1. Task mental model
2. Team mental model
3. Team process
4. Performance
Time 2
5. Task mental model
6. Team mental model
7. Team process
8. Performance
Time 3
9. Task mental model
10. Team mental model
1I. Team process
12. Performance
M
SD
1
2
3
4
-.15
.25
.21
-.27
.09
-.38
--
.53
.01
.23
.26
.28
.38
.18
.05
.27
.23
.89
.19
.05
.24
.42
.54
.40
.24
.28
.38
0.30
0.24
5
6
7
--.08
.20
-.09
-.29
.30
-.31
.10
.05
.09
.42
.28
.26 -.04
.30
.I6
.83
.35
.20
.17
.16
.30 -.02
0.24 2.30 2 . 5 5 0.35
0 . 2 7 0.50 1 . 2 4 0.26
.01
.40
.19
.21
0.20
0.30
8
9
.11
.10
.tl
.04
.24
.28
.14
.93
.15
.52
.12
2.37 3.05 0.32
0.50 1.52 0.26
-
10
11
12
-
-.20 - .16
.20 - 0 . 2 3 2 . 4 2 3.03
0.30 0.49 1.36
Note. N = 56 teams. If[r[ /> .26, p < . 0 5 ; i f [ r I ~>.35, p <.01.
The second section reviews the observed correlations between all
variables, both within and across times, as well as the tests of the
mediational framework. Descriptive statistics and intercorrelations
for all variables over time are presented in Table 2.
Change O v e r Time
We hypothesized that each of the study variables should increase over time as a result of team experience. Repeated measures
analyses of variance (ANOVAs) and one-tailed dependent t tests
were used to test such differences. The ANOVA results indicated
whether, in general, there were significant differences across time
for each variable. Should any differences be evident, inspecting the
mean contrasts across times would reveal the nature of such
effects. Because we advanced a priori hypotheses concerning the
nature of such effects, we were justified in conducting a priori
mean contrasts even if the omnibus ANOVAs were not significant
(see Winer, 1971).
Performance. K o l m o g o r o v - S m i r n o v tests revealed that the
observed performance distributions differed significantly ( p <
.01) from a normal distribution at each time. Therefore, we
applied a square-root transformation to the scores to better
approximate a normal distribution. The transformed means
and standard deviations are presented in Table 2. The performance A N O V A did not indicate a significant change over
time, F(2, 110) = 2.34, ns. However, the a priori t tests
revealed a significant change from Block 1 to Block 2,
tBlock l--mock 2(55) = 2.23, p < .05, yet no significant differences between Blocks 1 and 3, tBlo~k l--mock 3(55) = 1.40, ns,
or between Blocks 2 and 3, tmock 2--mook 3(55) = .61 ns. Thus,
the performance hypothesis is only partially supported, because
it rose significantly from Time 1 to Time 2 but then tapered off
to a level that did not differ from either of the first two times.
Process. The process ANOVA revealed a significant change
over time, F(2, 1 I0) = 9.28, p < .001. The a priori contrasts
revealed significant differences between all three time periods,
tmock l-m,,~k 2(55) = 2.18, p < .05; tmo~k l--BIo~k 3(55) = 3.72,
p < .001; and tmo~k 2--Block 3 ( 5 5 ) = 2.58, p < .01. As we
hypothesized, these analyses demonstrated that process increased
significantly from each time period to the next.
Task and team mental model convergences. The task mental
model convergence ANOVA demonstrated no significant difference over time, F(2, 110) = .90, ns, and none of the a priori
contrasts were significant. Similarly, team mental model convergence failed to show any significant differences over time, F(2,
110) = .38, ns, and none of the specific contrasts were significant.
Therefore, contrary to our hypotheses, neither type of mental
model exhibited greater convergence over time.
The Mediational M o d e l
Within-episode correlations. An examination of the correlations presented in Table 2 illustrates several interesting relationships. First, the two mental-model-convergence indices failed to
correlate significantly ( p > .05) at any time (rs = .15, - . 0 8 , and
.04 for Blocks 1-3, respectively). This provides evidence that the
two types of mental models are tapping qualitatively different
processes. Second, the correlations between the same mentalmodel-convergence indices over time were significant and moderately high for both task (r = .53, p < .001; r = .40, p < .01; and
r = .42, p < .01, for Blocks 1-2, 1-3, and 2-3, respectively), and
team (r = .38, p < .01; r = .28, p < .05; and r = .40, p < .01,
for Times 1-2, 1-3, and 2-3, respectively). These results, in
combination with the mean contrasts detailed above, suggest that
teams that were more in sync initially continued to be so and that
few mental-model-convergence changes were evident over time.
A third pattern was evident in that both team process levels
(Time 1-Time 2 r = .89, p < .001; Time 2-Time 3 r = .93, p <
.001) and team performance levels (Time 1-Time 2 r = .54, p <
.01; Time 2-Time 3 r = .52, p < .01) exhibited high significant
correlations over time. These findings, in combination with the
results reviewed above, indicate that although team processes
generally improved over time, team rank orders remained fairly
consistent. Similarly, although teams showed some change in
SHARED MENTAL MODELS AND TEAM EFFECTIVENESS
performance over time, generally those that did better initially
continued to be superior.
In terms of more substantive findings, the task-mental-model
convergence failed to exhibit any significant correlations with
either team process (r = .25, ns; r = .20, ns; and r = .14, ns, for
Times 1-3, respectively) or team performance (r = .21, ns; r =
- . 0 9 , ns; and r = . 12, ns, for Times 1-3, respectively). In contrast,
team mental model convergence exhibited significant correlations
with team process at Times 1 and 2 but not at Time 3 (r = .27, p <
.05; r = .29, p < .05; and r = .20, ns, respectively), and only with
team performance at Time 2 (r = .09, ns; r = .30,p < .05; and r =
.16, ns, for Times 1-3, respectively). Finally, team processes
exhibited significant correlations with team performance during
the first two episodes, although not during the third (r = .38, p <
.05; r = .31, p < .05; and r = .20, ns, for Times 1-3, respectively).
Repeated measures framework.
Although the within-episode
correlational results reviewed above are encouraging, they are
based on the findings from only 56 teams per time. A more
powerful way to analyze these data and to test the mediational
model is to apply a repeated measures multiple regression
(RMMR; see Cohen & Cohen, 1983). In brief, an RMMR regression analysis essentially "stacks" observations from each team
over time to create a unified database, analyzes the relationships
among variables using this stacked data and traditional multiple
regression techniques, and then adjusts the results for the fact that
the observations from the same teams over time are not independent. This analytic strategy provides a far more powerful alternative than a series of within-time analyses, and also allows one to
partition total variance into that which exists across teams and that
which occurs within teams across time.
We used the RMMR analyses in a path-analysis framework to
test the mediational model. We compared the hypothesized model
against a saturated model (that includes nonhypothesized paths),
which yields two indices of fit: Q and chi-square (Pedhazur, 1982).
Q is a descriptive index that reveals how much of the total
potential variance in a matrix is captured by the hypothesized
model. It ranges from 0 to 1, with values greater than or equal to
.90 generally viewed as acceptable. The chi-square tests the difference between the hypothesized and saturated models. It should
be noted, however, that with a reasonable sample size (such as that
yielded by our RMMR design), the chi-square test is extremely
powerful and can reject even acceptable models. Therefore, Pedhazur (1982) suggested that researchers should rely more heavily
on the Q index. Notably, because the overall fit indices in path
analysis reflect whether there are any significant omitted paths in
the model, a lack of fit here would suggest that one or both of the
shared-mental-model indices had a significant direct impact on
team performance. Thus, by examining both the overall-model fit
indices and the significance of the hypothesized paths, we can
ascertain whether team processes mediate the influence of shared
mental models on team performance.
The results for the hypothesized model are depicted in Figure 1.
The direct effects of both mental-model-convergence indices on
team process were significant, R 2 = .10, F(4, 108) = 3.30, p <
.05;/3te~,~ = .26, p < .01; and/3t~ k = .31, p < .01. Furthermore,
the direct effect of process on performance was significant, R E =
.09, F(2, 110) = 18.70, p < .01;/3p . . . . = .49, p < .0 l, yet adding
the two mental model convergence indices to the equation failed to
account for significant additional variance, and overall the model
279
Team Mental
Model
Convergence
/
I
Task Mental
Model
Convergence
Figure
1.
observations.
Team
Process
.90
.49 *
.91
/
Teanq
Performance
.31 *
Results of hypothesized model. N = 56 teams, 168
* p < .05.
yielded impressive fit indices, Q = .98 and )(2(166, N = 168
observations) = 2.20, ns. T h e significance of the chi-square was
determined using the degrees of freedom from the RMMR because
it is a more conservative estimate than the degrees of freedom
associated with the sample size.
Finally, to more fully explore the nature of the underlying
relationships between shared mental models and team performance, we regressed performance onto the two convergence indices without including the team process variable. To conclude that
team processes mediated the influence of shared mental models on
performance, one must demonstrate that mental models indeed
exhibited significant relationships with performance when taken
by themselves (see Baron & Kenny, 1986). This final analysis
revealed that although the direct effect of team-mental-model
convergence was significant (/3t~m = .87, p < .01), the task
convergence index was not (/3t~k = .39, ns). Thus, in summary,
we found substantial support for the hypothesized model. Furthermore, the impact of members' team mental convergence on team
performance was fully mediated by team processes. The taskmental-model convergence effect, however, was less clear.
Whereas task mental models did not show any significant relationship with team performance, they did have a positive influence on
team processes, which in turn related significantly to performance.
Technically, these results do not fulfill the requirements for mediation (again, see Baron & Kenny, 1986), but they are indicative
of an indirect effect.
Discussion
We had a number of objectives for this study. First, we sought
to distinguish, both conceptually and empirically, between task
and team mental models. Second, we sought to examine the
influence of convergence, or sharedness, of team members' mental
models as related to team processes and performance. And third,
we were interested in whether sharedness, as well as team processes and performance, developed over time. Our results supported the distinction between task and team mental models and
illustrated that they each had unique effects on subsequent team
processes. In turn, team processes were related significantly to
team performance. More detailed analyses revealed that teammental-model sharedness related significantly to team performance, but that the relationship was fully mediated by team
280
MATHIEU ET AL.
processes. In contrast, task-mental-model sharedness did not correlate significantly with team performance but did have an indirect
effect through its impact on team processes. These results offer
empirical confirmation for the often inferred but yet to be empirically demonstrated impact of shared mental models on team
effectiveness (Cannon-Bowers & Salas, 1990; Klimoski & Mohammed, 1994; Kraiger & Wenzel, 1997; Rentsch & Hall, 1994).
In terms of the hypothesized changes over time, neither mentalmodel convergence increased significantly. In contrast, team processes did improve significantly from each performance episode to
the next. Team performance, however, rose significantly from
Episode 1 to Episode 2, but then faded during Episode 3 to a level
that did not differ significantly from either of the first two. One
potential reason for the lack of increase in mental-model sharedhess may lie in the nature of the feedback we provided. Specifically, although we provided outcome feedback to all teams following each performance episode, we did not offer any
developmental feedback in terms of pointing out mistakes, making
process suggestions, and so forth. In real-world settings the military conducts what are known as "after action reviews" (AARs) in
which seasoned officers provide detailed analysis of and feedback
on how the teams performed and discusses with them how they
might improve. We intentionally avoided providing AAR-type
feedback in order to maintain experimental control and to test
whether experience alone would act to align members' mental
models. Apparently, practice alone was sufficient to enhance how
well the teams could coordinate their efforts, strategize, and cooperate, but it did not crystallize in terms of greater mental model
sharedness. Thus, our results echo the old training adage that time
on task alone is not enough--teams need guided experiences and
developmental feedback if we expect them to learn.
Significance of Shared Mental Models in Teams
The findings of this study are particularly important because
they contribute evidence that supports the construct validity of the
shared-mental-model construct. Specifically, we have shown that
the similarity of knowledge structures between two team members
can predict the quality of team processes and performance. This
evidence adds to similar findings that indicate that shared mental
models among team members can be measured and that they are
related to important team outcomes (see Stout et al., in press). The
significance of these findings to team performance research is
worth considering in more detail.
To begin with, all of the arguments we presented earlier regarding the mental model construct share one important assumption-that it is the organization or structure of a person's knowledge that
contributes to his or her performance (Goldsmith & Kraiger, 1997;
Kraiger, Ford, & Salas, 1993). In fact, several studies have found
that measuring an individual's knowledge structures can be effective in predicting performance in various domains (e.g., see
Kralger, Salas, & Cannon-Bowers, 1995; Lesgold et al., 1989;
McKeithen, Reitman, Rueter, & Hirtle, 1981) above and beyond
more traditional measures. One can draw two rather obvious
conclusions from this work. The first is that that if one knows
something about how a person's knowledge is organized, one can
predict performance outcomes. Second, and perhaps more important to the current discussion, the predictive ability of knowledgeorganization measures is superior to traditional knowledge tests, at
least for some tasks. Hence, invoking the mental-model construct
at the individual level is justified because it seems to describe
something important about knowledge that is not captured in
traditional knowledge measures (i.e., the organization of that
knowledge).
Extending this line of thinking to the team level also has
important implications. Specifically, the shared-mental-model
construct suggests that it is not only the overlap of knowledge
among team members that is predictive of team outcomes but also
the synergy of the knowledge organizations. Hence, when it is
shown (as we have here) that the sharedness of teammates' mental
models predicts subsequent team-processes performance, the implication is that knowledge organization--and the relationship
among the ways various team members organize their own task
knowledge--is a crucial concept. Once again, we believe that this
fact justifies the use of the shared mental construct as something
above and beyond simple shared task knowledge (see also Stout et
al., in press).
Generalizability Issues
This study examined newly formed teams of relative novices
performing a complex low-fidelity simulation with no expectation
of later interactions. Naturally this raises some questions about the
generalizability of our findings. Notably, questions of generalizability, or external validity, hinge on making inferences about the
applicability of the results of a given study to some other target
population and setting. Important considerations in this regard
include questions about the comparability of the tasks or situational demands, the sample populations, and time-related factors
(see Cook & Campbell, 1979; Pedhazur & Schmelkin, 1991).
As for the comparability of our low-fidelity simulation with
other task environments, the flight simulation task that we used
was dynamic, interactive, interdependent, and allowed for objective measures of team performance. It was not too simplistic and
mundane or too complex for this sample, and proved to be a very
engaging "middle-ground task." We are not suggesting that our
results would generalize directly to flight crews and combat situations. Obviously, physical factors such as gravitational forces and
life-threatening elements, along with a myriad of other characteristics, make the latter settings quite different than our situation.
The point is that we were interested in testing how teammates'
mental models and team processes and performance develop in
general and sought a research platform that would enable us to
simulate, yet control, task elements while maintaining critical team
features (e.g., member interdependence, dynamic interaction, designated roles and duties, shared and valued objectives). In effect,
we argue that while the use of low-fidelity simulations limits the
point-to-point correspondence that is critical for direct generalizability to real-world settings, they do provide for a controlled
examination of the critical factors influencing team processes and
performance in dynamic environments. In this sense we are interested in studying the construct interrelationships and their applicability to teams in other situations. We believe that our simulation
provides a useful vehicle fbr such tests (for similar views, see
Baker et al., 1993; Driskell & Salas, 1992; Weaver et al., 1995).
Perhaps a more salient issue regarding generalizability relates to
the participant sample in this study as compared with most realworld teams. Because we formed teams strictly for the basis of the
SHARED MENTAL MODELS AND TEAM EFFECTIVENESS
experiment and they had no expectation of future interactions, our
results would probably best apply to real-world crews. Airline,
military, and other real-world crews are often formed on a moment's notice, perform their activities, and then disband (see
Ginnett, 1990). Therefore, there are no (or limited) influences of
previous experience working with one another, role relationships
need to be established, and members have relatively little concern
for the long-term consequences of their activities with one another.
This may well alter the underlying relationships among variables
in the I-P-O framework. Although we recognize that the crew
versus team distinction may be an important boundary condition
for this and other studies that examine newly formed teams, we are
not aware of any systematic research that has sought to examine
this issue. We believe that exploration of this issue represents an
important avenue for future research.
A final issue regarding generalizability pertains to the relative
skill levels of our participants. For example, U.S. military teams
(or crews) flying multimillion-dollar jets, operating complex command and control networks, and functioning in complex and demanding environments with life and death literally hanging in the
balance are among the most highly skilled and trained teams in
existence. They know what they are supposed to do and have
demonstrated that they have the capabilities to be successful in
countless hours of simulated and real-life engagements. Nevertheless, history has shown that breakdowns occur and even these
teams make mistakes. Can much be applied to these teams from
the study of novices? At issue here is whether the underlying
nature of the construct relationships that we examined would be
isomorphic across team-competency levels. On one hand, it is
possible that the fundamental nature of how novice teams perform
may differ significantly from how expert teams perform. On the
other hand, the differences between the two may be more a matter
of degree than nature. While we submit that the construct relationships that we examined are likely to be fairly robust, here too we
are unaware of systematic research that has addressed this issue.
This offers yet another fruitful direction for future research.
Suggestions f o r Future R e s e a r c h and Application
The caveats regarding generalizability discussed above notwithstanding, our results highlight several important areas for future
research and application. We found that greater mental-model
convergence relates significantly to better team process and
thereby performance. This suggests that efforts to increase teammates' sharedness might lead to greater team effectiveness. One
strategy for doing so might be to investigate common underlying
cognitive abilities or experiences that give rise to certain knowledge structures. In other words, if individual differences can be
tied consistently with the development and use of particular mental
models, then teams might be composed so as to enhance members'
sharedness. Traditional human resources efforts such as selection,
staffing, and placement could be used to achieve such matches.
Alternatively, or perhaps in addition, there are a variety of intervention strategies that could help to develop shared mental models.
Naturally, training applications immediately come to mind, but
there may be other strategies, such as job rotations and feedback
programs (e.g., AARs), that might also prove valuable. Finally,
perhaps certain mental models naturally emerge from exposure to
events, whether those events be merely working with a teammate
281
(for longer periods than were examined here) or having to confront
a given environmental challenge. In any case, there appear to be a
number of different avenues that can be pursued to help team
members develop shared mental models.
A second finding from this research is that different types of
mental models can be identified and assessed and that they have
unique influences on team processes. This suggests that work on
"the" team mental model may be short-sighted at best, and confounded at worst. Our findings suggest that researchers and practitioners should conduct thorough team task analyses to identify
the most critical knowledge requirements for a given situation and
which of those knowledges must be shared. In this way, the human
resources programs mentioned above could be best aligned to
target the most critical knowledge levers for team effectiveness.
One final direction for future research warrants mentioning.
High mental-model convergence, as operationalized here, does not
imply that the models formed by the team members are appropriate. In other words, convergence does not equal quality--and
teammates may share a common vision of their situation yet be
wrong about the circumstances that they are confronting. One
approach to indexing the quality of members' mental models is to
compare them against experts' models. This would help to identify
deficiencies in members' models as compared with a standard
referent structure. A potential problem with this approach, however, is that it assumes that there is a single expert model. For
complex constructs such as task work and team work, multiple
"expert" mental models may exist. Thus, while we suggest that
future research should also consider the quality of teammates'
models, we believe that such efforts should go beyond the use of
a single "expert" model as the gauge.
References
Baker, D., Prince, C., Shrestha, L., Oser, R., & Salas, E. (1993). Aviation
computer games for crew resource management training. International
Journal of Aviation Psychology, 3, 143-156.
Baker, D. P., Salas, E., & Cannon-Bowers, J. (1998). Team task analysis:
Lost but hopefully not forgotten. Industrial-Organizational Psychologist, 35, 79-83.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator distinction
in social psychological research: Conceptual, strategic, and statistical
considerations. Journal of Personality and Social Psychology, 51, 11731182.
Benenhausen, K. L. (1991). Five years of groups research: What we have
learned and what needs to be addressed. Journal of Management, 17,
345-381.
Borgatti, S., Everett, M. G., & Freeman, L. C. (1992). UCINET IV network
analysis software. Columbia, SC: Analytic Technologies.
Brannick, M. T., Roach, R. M., & Salas, E. (1993). Understanding team
performance: A multimethod study. Human Performance, 6, 287-308.
Campion, M. A., Medsker, G. J., & Higgs, A. C. (1993). Relations between
work group characteristics and effectiveness: Implications for designing
effective work groups. Personnel Psychology, 46, 823-850.
Cannon-Bowers, J. A., & Salas, E. (1990, August). Cognitive psychology
and team training: Shared mental models in complex systems. Paper
presented at the Annual Meeting of the Society of Industrial and Organizational Psychology, Miami, FL.
Cannon-Bowers, J. A., Salas, E., & Converse, S. A. (1993). Shared mental
models in expert team decision making. In N. J. Castellan, Jr. (Ed.),
Current issues in individual and group decision making (pp. 221-246).
Hillsdale, NJ: Erlbaum.
282
MATHIEU ET AL.
Cannon-Bowers, J. A., Tannenbaum, S. I., Salas, E., & Volpe, C. E. (1995).
Defining team competencies and establishing team training requirements. In R. Guzzo & E. Salas (Eds.), Team effectiveness and decision
making in organizations (pp. 333-380). San Francisco: Jossey-Bass.
Cohen, J., & Cohen, P. (1983). Applied multiple regression~correlation for
the behavioral sciences. Hillsdale, NJ: Erlbaum.
Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design
and analysis issues for field settings. Boston: Houghton Mifflin.
Driskell, J. E., & Salas, E. (1992). Can you study real teams in contrived
settings? The value of small group research to understanding teams. In
R. W. Swezey & E. Salas (Eds.), Teams: Their training and performance
(pp. 101-126). Norwood, NJ: Ablex.
Dyer, J. L. (1984). Team research and team training: A state-of-the-art
review. In F. A. Muckler (Ed.), Human factors review (pp. 285-323).
Santa Monica, CA: Human Factors Society.
Fleishman, E. A., & Zaccaro, S. J, (1992). Toward a taxonomy of team
performance functions. In R. W. Swezey & E. Salas (Eds.), Teams:
Their training and performance (pp. 31-56). Norwood, NJ: Ablex.
Foushee, H. C., & Manos, K. L. (1981). Information transfer within the
cockpit: Problems in intracockpit communications. In C. E. Billings &
E. S. Cheaney (Eds.), Information transfer problems in the aviation
system (NASA Report No. TP-1875). Moffett Field, CA: NASA-Ames
Research Center. (NTIS No. N81-31162)
Ginnett, R. C. (1990). Airline cockpit crews. In J. H. Hackman (Ed.),
Groups that work (and those that don't) (pp. 427-448). San Francisco:
Jossey-Bass.
Gist, M. E., Locke, E. A., & Taylor, M. S. (1987). Organizational behavior:
Group structure, process, and effectiveness. Journal of Management, 13,
237-257.
Gladstein, D. L. (1984). Groups in context: A model of task group effectiveness. Administrative Science Quarterly, 29, 499-517.
Goldsmith, T. E., & Kraiger, K. (1997). Applications of structural knowledge assessment to training evaluation. In J. K. Ford, S. W. S. Kozlowski, K. Kraiger, E. Salas, & M. Teachout (Eds.), Improving training
effectiveness in work organizations (pp. 73-96). Mahwah, NJ: Erlbaum.
Guzzo, R. A., & Dickson, M. W. (1996). Teams in organizations: Recent
research on performance and effectiveness. Annual Review of Psychology, 47, 307-338.
Hackman, J. R. (1990). Work teams in organizations: An orienting framework. In J. R. Hackman (Ed.), Groups that work (and those that don't)
(pp. 1-14). San Francisco, CA: Jossey-Bass.
Hackman, J. R., Brousseau, K. R., & Weiss, J. A. (1976). The interaction
of task design and group performance strategies in determining group
effectiveness. Organizational Behavior and Human Performance, 16,
350-365.
Johnson-Laird, P. (1983). Mental models. Cambridge, MA: Harvard University Press.
Kanki, M. A., & Foushee, H. C. (1989). Communication as group performance mediator of aircrew performance. Aviation, Space, and Environmental Medicine, 60, 402-410.
Kleinman, D. L., Luh, P.B., Pattipati, K. R., & Serfaty, D. (1992). Mathematical models of team performance: A distributed decision-making
approach. In R. W. Swezey & E. Salas (Eds.), Teams: Their training and
performance (pp. 177-217). Norwood, NJ: Ablex.
Klimoski, R., & Mohammed, S. (1994). Team mental model: Construct or
metaphor? Journal of Management, 20, 403-437.
Kraiger, K., Ford, J. K., & Salas, E. (1993). Application of cognitive,
skill-based, and affective theories of learning outcomes to new methods
of training evaluation. Journal of Applied Psychology, 78, 311-328.
Kraiger, K., Salas, E., & Cannon-Bowers, J. A. (1995). Measuring knowledge organization as a method for assessing learning during training.
Human Factors, 37, 804-816.
Kraiger, K., & Wenzel, L. H. (1997). A framework for understanding and
measuring shared mental models of team performance and team effec-
tiveness. In E. Salas, M. T. Brarmick, & C. Prince (Eds.), Team performance assessment and measurement: Theory, methods, and applications
(pp. 63-84). HiUsdale, NJ: Erlbaum.
Lesgold, A., Rubinson, H., Feltovich, P., Glaser, R., Klopfer, D., & Wang,
Y. (1989). Expertise in a complex skill: Diagnosing x-ray pictures. In
M. T. H. Chi & R. Glaser (Eds.), The nature of expertise (pp. 311-342).
Hillsdale, NJ: Erlbaum.
Martell, R. E., Guzzo, R. A., & Willis, C. E. (1995). A methodological and
substantive note on the performance-cue effect in ratings of work-group
behavior. Journal of Applied Psychology, 80, 191-195.
Mcintyre, R. M., & Salas, E. (1995), Measuring and managing for team
performance: Emerging principles from complex environments. In R.
Guzzo & E. Salas (Eds.), Team effectiveness and decision making in
organizations (pp. 149-203). San Francisco: Jossey-Bass.
McKeithen, K. B., Reitman, J. S., Rueter, H. H., & Hirtle, S. C. (1981).
Knowledge organization and skill differences in computer progranuners.
Cognitive Psychology, 13, 307-325.
Morgan, B. B., Jr., Glickman, A. S., Woodard, E. A., Blaiwes, A. S., &
Salas, E. (1986). Measurement of team behaviors in a Navy environment
(NTSC Tech. Rep. No. 86-014). Orlando, FL: Naval Training Systems
Center.
Pedhazur, E. J. (1982). Multiple regression in behavioral research. New
York: Holt, Rinehart & Winston.
Pedhazur, E. J., & Schmelkin, L. P. (1991). Measurement, design, and
analysis: An integrated approach. Hillsdale, NJ: Erlbaum.
Powell, J., & Basham, T. (1992). Falcon 3: The complete handbook. Corte
Madera, CA: The Waite Group.
Prince, A., Brannick, M. T., Prince, C., & Salas, E. (1992). Team process
measurement and the implications for training. Proceedings of the Human Factors Society 36th annual meeting (pp. 1351-1355). Santa
Monica, CA: Human Factors Society.
Rentsch, J. R., & Hall, R. J. (1994). Members of great teams think alike:
A model of the effectiveness and schema similarity among team membets. Advances in Interdisciplinary Studies of Work Teams, 1, 223-261.
Rentsch, L R., Heffner, T. S., & Dully, L. T. (1994). What you know is
what you get from experience: Team experience related to teamwork
schemas. Group and Organization Management, 19, 485-509.
Roberts, K. H., & O'Reilly, C. A. (1976). Interpersonal, work group, and
organizational communication: A systematic approach to understanding
organizations (Technical report, Contract No. N00014-69-A-02001054). Berkeley, CA: Institute of Industrial Relations.
Rouse, W. B., & Morris, N. M. (1986). On looking into the black box:
Prospects and limits in the search for mental models. Psychological
Bulletin, 100, 349-363.
Salas, E., Bowers, C. A., & Cannon-Bowers, J. A. (1995). Military team
research: 10 years of progress. Military Psychology, 7, 55-75.
Salas, E., Dickinson, T. L., Converse, S. A., & Tannenbaum, S. I. (1992).
Toward an understanding of team performance and training. In R. W.
Swezey & E. Salas (Eds.), Teams: Their training and performance (pp.
3-29). Norwood, NJ: Ablex.
Spectrum Holobyte. (1991). Falcon 3.0 flight manual. Alameda, CA:
Author.
Stout, R. J., Cannon-Bowers, J. A., & Salas, E. (1996). The role of shared
mental models in developing team situational awareness: Implications
for training. Training Research Journal, 2, 85-116.
Stout, R. J., Cannon-Bowets, J. A., Salas, E., & Milanovich, D. M. (in
press). Planning, shared mental models, and coordinated performance:
An empirical link is established. Human Factors.
Stout, R. J., Cannon-Bowers, J. A., Salas, E., & Morgan, B. B. (1990).
Does crew coordination behavior impact performance? Proceedings of
the Human Factors Society 34th annual meeting (pp. 1382-1386). Santa
Monica, CA: Human Factors Society.
Stout, R. J., Salas, E., & Kraiger, K. (1996). The role of trainee knowledge
SHARED MENTAL MODELS AND TEAM EFFECTIVENESS
structures in aviation team environments. International Journal of Aviation Psychology, 7, 235-250.
Swezey, R. W., & Salas, E. (1992). Guidelines for use in team-training
development. In R. W. Swezey & E. Salas (Eds.), Teams: Their training
and performance (pp. 219-246). Norwood, NJ: Ablex.
Tannenbaum, S. I., Beard, R. L., & Salas, E. (1992). Team building and its
influence on team effectiveness: An examination of conceptual and
empirical developments. In K. Kelley (Ed.), Issues, theory, and research
in industrial~organizational psychology (pp. 117-153). New York:
Elsevier Science.
Tesluk, P., Zaccaro, S. J., Marks, M. A., & Mathieu, J. E. (1997). Task and
aggregation issues in analysis and assessment of team performance. In
M. T. Brannick, E. Salas, & C. Prince (Eds.), Team performance
assessment and measurement: Theory, methods, and applications (pp.
197-224). Mahwah, NJ: Erlbaum.
283
Weaver, J. L., Bowers, C. A., Salas, E., & Cannon-Bowers, J. A.
(1995). Networked simulations: New paradigms for team performance research. Behavior Research Methods, Instruments and Computers, 27, 12-24.
Wilson, J. R., & Rutherford, A. (1989). Mental models: Theory and
application in human factors. Human Factors, 31, 617-634.
Wirier, B. J. (1971). Statistical principles in experimental design (2nd ed.).
New York: McGraw-Hill.
Received May 13, 1998
Revision received June 4, 1999
Accepted June 9, 1999 •