Transactive Memory Errors
SLIPPAGE IN THE SYSTEM: EXAMINING THE EFFECTS OF ERRORS IN
TRANSACTIVE MEMORY BEHAVIOR ON TEAM EFFECTIVENESS
MATTHEW J. PEARSALL
The University of Arizona
The Eller College of Management
McClelland Hall, 405
Tucson, AZ, 85721-0108
Tel: (520) 621-7461
Fax: (520) 621-4171
e-mail: mpearsal@email.arizona.edu
ALEKSANDER P.J. ELLIS
The University of Arizona
The Eller College of Management
McClelland Hall, 405T
Tucson, AZ, 85721-0108
Tel: (520) 621-7461
Fax: (520) 621-4171
e-mail: ellis@eller.arizona.edu
BRADFORD S. BELL
Cornell University
School of Industrial and Labor Relations
386 Ives Hall
Ithaca, NY 14853
Phone: (607) 254-8054
Fax: (607) 255-1836
e-mail: bb92@cornell.edu
Author note: This research was supported in part by contract F49620-03-1-0377
from the AFOSR/MURI to the University of Arizona. Although support for this work is
gratefully acknowledged, the ideas expressed herein are those of the authors and not necessarily
endorsed by the funding agency.
1
Transactive Memory Errors
2
SLIPPAGE IN THE SYSTEM: EXAMINING THE EFFECTS OF ERRORS IN
TRANSACTIVE MEMORY BEHAVIOR ON TEAM EFFECTIVENESS
ABSTRACT
In an effort to extend theory and research regarding transactive memory, we examined
the effects of errors in transactive memory behavior on team performance. We also examined
whether the relationship between errors in transactive memory behavior and team performance is
the result of specific cognitive processes in the form of team-interaction mental models and the
emergent cognitive manifestations of transactive memory. Results from 69 teams working on a
command and control simulation indicated that errors in transactive memory behavior negatively
predicted team performance. Results also indicated that the relationship between errors in
transactive memory behavior and team performance was mediated by team-interaction mental
models and the emergent cognitive manifestations of transactive memory.
Transactive Memory Errors
3
Fast paced and complex strategic environments have prompted organizations to structure
work around teams rather than individuals (e.g., Cannon-Bowers, Oser, & Flanagan, 1992;
Colquitt, 2004; Ilgen, 1994; Stewart & Barrick, 2000). By 1999, nearly half (48%) of U.S.
organizations were using teams (Devine, Clayton, Philips, Dunford, & Melner, 1999). Work
teams, composed of two or more interdependent members who share common goals, are
considered highly adaptable and allow organizations to take advantage of the distributed
expertise of team members (Cohen & Bailey, 1997; Keller, 2001). Teams are frequently
assigned to complex tasks because they allow members to share work, monitor each other's
behavior, and contribute their expertise (Mathieu, Heffner, Goodwin, Salas, & Cannon-Bowers,
2000; Smith-Jentsch, Mathieu, & Kraiger, 2005).
The increased dependence on team based work structures has led to a substantial amount
of research aimed at identifying factors that predict team effectiveness (see Ilgen, Hollenbeck,
Johnson, & Jundt, 2005; Kozlowski & Bell, 2003). One factor that has recently received
attention within the organizational literature is transactive memory, which has been defined as a
cooperative division of labor for learning, remembering, and communicating relevant team
knowledge (e.g., Hollingshead, 2001; Lewis, 2003; Wegner, 1987). Researchers have
consistently shown that transactive memory positively affects team performance (e.g., Austin,
2003; Cannon-Bowers & Salas, 2001; Lewis, 2003; Moreland & Myakovsky, 2000).
However, because transactive memory systems represent implicit structures for assigning
responsibility for new information within the team based on a shared conception of one another's
expertise (Brandon & Hollingshead, 2004), they rely heavily on team members’ ability to
distribute and retrieve information (Gibson, 2001; Hinsz et al, 1997). Communication is an
essential component of transactive memory that drives the development and operation of the
system (e.g., Hollingshead, 1998a; 1998b). As a result, transactive memory systems are not
Transactive Memory Errors
4
infallible and opportunities for error abound. For example, Hollingshead (1998a, p. 427) notes
that “information may be transferred or explicitly delegated to the “wrong” individual in the
system, e.g., one who does not have responsibility for that type of information or is unlikely to
remember it due to a lack of expertise.” While researchers recognize the likelihood of such
errors, little has been done to examine their potential impact on team effectiveness.
Therefore, the purpose of this study is to identify errors in transactive memory behavior
and examine their implications for team performance. Specifically, we expect that errors in
transactive memory behavior will undermine efficiency and effectiveness, negatively predicting
team performance. We also attempt to uncover the cognitive mechanisms through which errors
in transactive memory behavior influence team performance. Specifically, we examine teaminteraction mental models and the emergent cognitive manifestations of transactive memory as
mediators of the relationship between errors in transactive memory behavior and team
performance. To develop our hypotheses, we first discuss the types of behaviors involved in the
development and operation of transactive memory, identify where errors may enter the system,
and discuss their effects on team performance. We then introduce team-interaction mental
models and the emergent cognitive manifestations of transactive memory and discuss their role
in explaining the relationship between these errors and team performance.
Errors in Transactive Memory Behavior
Transactive memory was originally conceived by Wegner (e.g., 1987) to describe how
individuals in close relationships divide up their combined workload. Rather than cluttering their
memory, each partner would take responsibility for specific duties. Any pertinent information
would then be sent to or retrieved from that individual. Wegner labeled this system “transactive
memory,” as it allowed couples to efficiently and effectively process information (e.g., Wegner,
1987; Wegner, Erber, & Raymond, 1991; Wegner, Giuliano, & Hertel, 1985). To illustrate the
Transactive Memory Errors
5
inner workings of the system, Lewis (2003) described a typical couple, John and Jane, who need
to remember friends’ and family members’ birthdays. John cannot remember any of the dates,
but since Jane is responsible for remembering birthdays, he can access her memory and retrieve
the information whenever needed. In addition, when new friends or family are added to the list,
John can send birthday information to Jane without encoding any of it into memory.
Because team members work interdependently with one another, they also tend to
develop close relationships. As a result, organizational researchers have translated the original
conceptualization of the construct to describe how team members synergistically combine their
individual memory capacities (e.g., Hinsz et al., 1997; Moreland, 1999). Like individuals in close
relationships, “transactive memory allows different members of the group to process
information, so they remember the information that is directly related to their area of expertise”
(Hinsz et al., 1997, p. 48). Utilizing their transactive memory system, team members can rely on
their teammates’ expertise, enabling the team to access a larger pool of task-relevant information
and avoid wasting cognitive effort (Hollinshead, 1998b).
Transactive memory develops and operates in teams through the communication of
expertise specific information between team members. Consequently, communication plays a
vital role in the team’s transactive memory system (Hollingshead, 1998a; 1998b; Lewis, 2003),
allowing team members to delegate or assign responsibility for learning and storing new
information to certain individuals (Wegner, 1987). For example, Hollinshead (1998b) notes that
“communication serves many beneficial functions in the encoding and storage of new
information in transactive memory systems” (Hollingshead, 1998b, p. 427).
Communication in transactive memory systems can be broken down into three specific
behaviors: directory updating, information allocation, and retrieval coordination (e.g., Ellis, in
press; Hollingshead, 1998a, 1998b). Through directory updating, team members learn about each
Transactive Memory Errors
6
other’s areas of expertise by sharing or requesting information about “who knows what.”
Through information allocation, specific information is communicated to the team member that
possesses the relevant area of expertise to apply it. Through retrieval coordination, team
members use their “directory of directories” to request information from the teammate with the
proper area of expertise.
Given that transactive memory relies on team members’ use of directory updating,
information allocation, and retrieval coordination, we suggest that there are numerous
opportunities for human error to enter the system. Psychologists generally define human error as
planned action that fails to achieve the desired outcome (e.g., Reason, 1990). In the context of
transactive memory, team members may commit errors when engaging in directory updating,
information allocation, and retrieval coordination. Directory updating errors occur when team
members misstate information about their own area of expertise, giving other members a false
view of the breadth of their knowledge. Information allocation errors occur when a team member
shares information with a teammate who lacks the proper expertise to apply it. Retrieval
coordination errors occur when a team member requests information from a teammate based on a
mistaken understanding of his or her area of expertise.
To illustrate these transactive memory behavior errors, consider the example of a crossfunctional product team in the pharmaceutical industry. The team combines experts from R&D,
sales, marketing, operations, and other divisions to develop and sell a drug. In this example, a
directory updating error may occur when the sales expert accidentally leads the other members of
the team to believe that he or she has expertise regarding federal regulations surrounding drug
warnings, when in fact this expertise is possessed by the individual from marketing. An example
of an information allocation error would be if the expert from R&D provided the drug’s formula
to the sales expert rather than the production manager overseeing the manufacture of the drug.
Transactive Memory Errors
7
Finally, an example of a retrieval coordination error would be if the production manager
requested a projection of next month’s production volume from the marketing expert rather than
the sales expert.
While a number of studies have shown that directory updating, information allocation,
and retrieval coordination behavior positively affects team performance (e.g., Ellis, in press;
Hollingshead, 1998a), researchers suggest that errors occurring within the system will negatively
affect team efficiency and effectiveness (e.g., Kozlowski & Bell, 2003; Wegner, 1986). When
critical information is shared with or requested from the wrong person within the team, team
members need to identify incorrect information, resolve conflicting claims about each team
member’s area of expertise, and identify the individual with whom the information should be
shared with or requested from. For example, when a team member commits a directory updating
error, his or her teammates need to realize he or she is incorrect, figure out who actually
possesses the expertise, and ensure that the whole team is aware of it. Otherwise team members
will continue to look to the individual for expertise that he or she does not have. Likewise, when
a team member commits an information allocation error, the individual who receives the
information needs to call attention to the mistake, the team must then search for the correct
individual, and the information must be sent again. If they do not, the individual will continue to
retain information that they cannot use, while the team member who actually needs it remains
unaware of the situation and works on other tasks. A similar process occurs when a team
member commits a retrieval coordination error. The team member who receives the request must
realize that he or she does not have the relevant information and the team must then search for
the individual with the correct area of expertise. Otherwise the individual making the request will
never retrieve the information that they need. In each instance, team members allocate attention
Transactive Memory Errors
8
and time to fixing their mistakes, which detracts from team performance (e.g., Ellis, in press;
Hinsz et al., 1997).
Fixing problems stemming from errors in transactive memory behavior is made more
difficult by the fact that team members are often unwilling to correct their behavior in such
situations and persist with a losing course of action (e.g., Ross & Staw, 1993; Staw, 1976). Even
if they are willing to correct the problem, team members may not always be able to follow
through with the necessary steps if they are unaware that the information shared or requested is
inaccurate or meant for someone else (Janis, 1982; Stewart & Stasser, 1995). This passive
acceptance of specialized knowledge within the team promotes the collective use of incorrect
information, impairing team performance (Faucheux & Mackenzie, 1966; Gouran, 1986; Janis,
1982; Sundtrom, Busby, & Bobrow, 1997). As a result, we hypothesize the following:
H1: Errors in transactive memory behavior (i.e., directory updating, information
allocation, and retrieval coordination) will negatively predict team performance
Although we expect that errors in transactive memory behavior will negatively predict
team performance, we are also interested in why such problems occur. According to Ellis (in
press), errors in transactive memory behavior can be considered a particular manifestation of
what Marks et al. (2001) label as action processes, which include activities such as systems
monitoring and coordination behavior that lead directly to the accomplishment of specific team
goals. Researchers argue that these types of team processes influence emergent cognitive states,
which include constructs that “characterize properties of the team that are typically dynamic in
nature and vary as a function of team context, inputs, processes, and outcomes” (Marks et al.,
2001, p. 357). In the current research we focus on two emergent states, team-interaction mental
models and the emergent cognitive manifestations of transactive memory.
Team-interaction mental models.
Transactive Memory Errors
9
A team mental model is an organized knowledge structure, or psychological map, that
depicts how the characteristics, duties, and needs of teammates fit with one another (e.g.,
Mohammed, Klimoski, & Rentsch, 2000). They provide team members with a heuristic that can
help them interpret information in a similar manner (Hinsz et al., 1997) and allow them to select
actions that are consistent and coordinated with those of their teammates (Mathieu, et al., 2000).
Each map differs in its level of accuracy and similarity across team members. Team members
need to hold accurate representations of the team performance environment in order to be
effective and similar representations in order to be efficient (e.g., Smith-Jentsch, Campbell,
Milanovich, & Reynolds, 2001). Researchers have shown that team-interaction mental model
similarity and accuracy both exhibit significant and unique positive relationships with team
performance (e.g., Ellis, in press; Marks et al, 2002; Marks, Zaccaro, and Mathieu, 2000;
Mathieu et al, 2000)
Cannon-Bowers, Salas, and Converse (1993) identified four types of mental models that
team members can hold: technology/equipment mental models, job/task mental models, team
mental models, and team-interaction mental models. While other types of mental models focus
on factors such as equipment functioning and environmental constraints, team-interaction mental
models focus on team members’ roles, responsibilities, and interaction patterns within the team.
Team-interaction mental models contain procedural knowledge about how team members' roles
and responsibilities should be combined to accomplish interdependent tasks (Marks, et al., 2002)
and may be particularly susceptible to errors in transactive memory behavior.
When team members erroneously identify their own expertise or mistake someone else’s
roles and responsibilities, team-interaction mental models will suffer as team members will
develop inaccurate and dissimilar conceptions of each other’s expertise and responsibilities. For
example, when directory updating errors occur, both the team member sharing an incorrect
Transactive Memory Errors 10
expertise and any team members receiving the information will generate inaccurate maps of team
roles and interactions. These maps will also be dissimilar to those of other team members who
did not note the error. Likewise, when team members make an information allocation or
retrieval coordination error, if the team member that receives the information or request does not
correct them, they will continue to harbor mistaken conceptions of other team member’s roles
and expertise that differ from those of their teammates, harming both team-interaction mental
model accuracy and similarity. If unresolved, these differences in team-interaction mental
models may lead to disagreements, tension and conflict (Hinsz et al, 1997). So pervasive are
team communication errors that researchers have identified their identification and correction as
critical steps in the successful development of team-interaction mental models (Blickensderfer,
Cannon-Bowers, & Salas, 1997; Kozlowski, Gully, McHugh, Salas, & Cannon-Bowers, 1996;
Kozlowski, Gully, Salas, & Cannon-Bowers, 1996). Therefore, we expect errors in directory
updating, information allocation, and retrieval coordination to hinder the development of teaminteraction mental models, leading to the following hypothesis:
H2: Errors in transactive memory behavior will negatively affect team-interaction
mental model similarity and accuracy.
Emergent Cognitive Manifestations of Transactive Memory
The emergent cognitive manifestations of transactive memory focus on the
distinctiveness of team members’ knowledge and their evaluation of how that knowledge is
distributed and shared within the team (Ilgen et al., 2005). Organizational researchers focusing
on transactive memory as an emergent cognitive state have defined it as “a combination of
knowledge possessed by each individual and a collective awareness of who knows what”
(Austin, 2003: 866). Three cognitive components reflect the distributed, cooperative nature of
transactive memory: specialization, coordination, and credibility (e.g., Liang et al, 1995;
Transactive Memory Errors 11
Moreland & Myaskovsky, 2000). Specialization refers to the level of memory differentiation
within the team, credibility refers to team members’ beliefs about the reliability of other
members’ knowledge, and coordination refers to the ability of the team members to work
together efficiently (e.g., Ellis & Bell, 2005; Lewis, 2003; Moreland & Myaskovsky, 2000).
Researchers have shown that specialization, credibility, and coordination exhibit significant
positive relationships with team performance (e.g., Austin, 2003; Lewis, 2003; Liang et al, 1995;
Moreland & Myaskovsky, 2000).
Specialization, credibility, and coordination are highly susceptible to errors in transactive
memory behavior. Specialization requires team members to understand both the breadth and
interrelated nature of their teammates’ expertise. When directory updating, information
allocation, and retrieval coordination errors occur, team members' perceptions of how knowledge
is distributed and shared within the team become uncertain and inaccurate. Furthermore, when
team members hear conflicting accounts of who knows what, or receive information or requests
for information unrelated to their own specialty, they can no longer rely on the expertise of their
teammates and credibility is inhibited. Finally, when team members receive faulty information
regarding who knows what within the team, and give and request information from team
members with the wrong expertise, members will be unable to understand how to coordinate
their efforts and utilize each other's expertise. In support of these assertions, researchers have
found that accurate communication regarding each other's expertise is critical to the successful
emergence of specialization, credibility, and coordination within teams (Lewis, 2004). Therefore,
we hypothesize that:
H3: Errors in transactive memory behavior will negatively affect specialization,
credibility, and coordination.
Transactive Memory Errors 12
Based on our first three hypotheses, we expect that errors in transactive memory behavior
will negatively affect team performance, team-interaction mental model similarity and accuracy,
and the emergent cognitive manifestations of transactive memory. Given that team-interaction
mental model similarity and accuracy and the emergent cognitive manifestations of transactive
memory have also been linked to team performance (e.g., Ellis, in press; Lewis, 2003; Marks et
al., 2002), we suggest that these cognitive states represent the theoretical mechanisms through
which errors in transactive memory behavior affect team performance. This is consistent with the
idea that emergent cognitive states intervene and transmit the influence of inputs such as errors
in transactive memory behavior to outcomes such as team performance (Ilgen et al., 2005; Marks
et al., 2001). Therefore, we hypothesize that:
H4: The effects of errors in transactive memory behavior on team performance will be
mediated by team-interaction mental model similarity and accuracy and the emergent
cognitive manifestations of transactive memory (i.e., specialization, credibility, and
coordination).
METHODS
Research Participants
Participants included 276 students from introductory management courses at a large
Southwestern University who were arrayed into 69 four-person teams. Out of the 276 students,
146 (53%) were male and 179 (65%) were white, with an average age of 21.6 years. In
exchange for their participation, each earned class extra credit and all were eligible for cash
prizes (up to $120 per team) based upon the team’s performance.1
Task
Participants engaged in a modified version of Distributed Dynamic Decision-making
(DDD) Simulation (see Miller, Young, Kleinman, & Serfaty, 1998). The DDD is a
Transactive Memory Errors 13
computerized, dynamic command and control simulation requiring team members to monitor a
geographic region and defend it against invasion from unfriendly tracks, which are radar
representations of enemy forces moving through the region (see Figure 1). The objective of the
task is to maximize the number of team points, which can be accomplished by identifying tracks,
determining whether they are friendly or unfriendly, and, if unfriendly, keeping them out of the
restricted zones by engaging them. However, to maximize their score, team members also had to
make sure that they were not disabling a friendly track or disabling an unfriendly outside the
restricted zones.
---------------------------------Insert Figure 1 about here
---------------------------------Geographic region. As shown in Figure 1, the geographic region is partitioned into four
quadrants of equal size (one per team member). In the center of the screen is a 4 x 4 square
designated as the "highly restricted zone" which is nested within a larger 12 x 12 square called
the "restricted zone." Outside the restricted zone is neutral space.
Bases and vehicles. In terms of monitoring the geographic region, each team member
has a home base of operations located in their quadrant with a detection ring that allows them to
detect the presence or absence of tracks within its radius. To detect tracks outside of their base's
detection ring, team members must rely on teammates or the vehicles located at their base.
Each team member was assigned four vehicles that were used to defend the space (i.e., keep
unfriendly tracks out of restricted areas). There are four different types of vehicles; (a) AWACS
(surveillance planes), (b) tanks, (c) helicopters, and (d) jets. Assets vary on five capabilities:
range of vision, speed, duration of operability, identification capacity, and power (see Table 1).
Capabilities are distributed among the assets so that each has both strengths and
Transactive Memory Errors 14
weaknesses. For example, the AWACS has the greatest range of vision but no power to engage
unfriendly tracks. Tanks, on the other hand, have the highest level of power but their range of
vision is small and their speed is slow. While all vehicles can detect tracks, only the AWACS
planes can identify tracks as friendly or enemy and share the information with the rest of the
team.
---------------------------------Insert Table 1 about here
---------------------------------Tracks. When tracks enter a detection ring, they show up as unidentified. Once the track
is identified using an AWACS plane, a team member can engage it with a tank, helicopter or jet,
depending on the power level of the track and the vehicle engaging it. If the vehicle has the
correct level of power, the target can be disabled. In this study, teams faced four different types
of tracks: A, B, C, D. Each track had either a power of 0 (friendly), 1, 3, or 5 depending on
whether it appeared in the training or experimental task (see Table 1).
Team member specialties. During DDD training, team members did not have specific
areas of expertise. Each team member controlled one AWACS plane, one tank, one helicopter,
and one jet, and knew that A, B, C, and D tracks corresponded to power 0, 1, 3, and 5
respectively. During the actual experimental task, team members did have specific areas of
expertise. Areas of expertise were created by splitting up knowledge regarding the targets, and
possession of four different vehicles. Each team member knew the power level of one track and
was responsible for one type of vehicle. DM1 knew that track D had a power of 0 (friendly) and
had all four AWACS planes, DM2 knew that track B had a power of 5 and had all four tanks,
DM3 knew that track C had a power of 3 and had all four helicopters, and DM4 knew that track
A had a power of 1 and had all four jets.
Transactive Memory Errors 15
Procedure
Immediately after entering the laboratory, participants were randomly assigned to one of
four computer stations (e.g., DM1, DM2, DM3, or DM4) within a four-person team. Participants
were trained on the declarative and procedural knowledge necessary for successful task
completion for approximately 30 minutes. This training described various aspects of the task,
including the location of the bases, the scoring, the functions of the various rings, the different
vehicles, and the tracks. Participants then played a 30-minute practice game, where they learned
how to launch and move vehicles, identify tracks, and attack targets without specific areas of
expertise.
After training, each team performed a 30-minute experimental task with specific areas of
expertise. Each team member was given a sheet that illustrated their own specific role, which
they were able to keep during the experimental task. Each team faced a total of 100 targets
during the task and could only communicate verbally with one another.
Measures
Errors in transactive memory behavior. To measure errors, this study employed direct
measures of verbal behavior based on the team members’ areas of expertise (see Hollingshead
1998a; 1998b), utilizing an additive index (i.e., sum) to represent errors in directory updating,
information allocation, and retrieval coordination at the team level (Chan, 1998). Directory
updating errors occurred when team members incorrectly shared expertise with their teammates
(e.g., “I’m DM2 and I have the helicopters” or “I'm DM2 and B targets are power 1").
Information allocation errors occurred when team members sent information to the person with
the incorrect correct track or vehicle specialty (e.g., “DM3, I have several B tracks in my
restricted zone”). Retrieval coordination errors occurred when team members requested specific
Transactive Memory Errors 16
information from someone with the wrong track or vehicle specialty (e.g., “DM3, what is the A
track again?”).
Two experimenters were in charge of coding. To ensure that the coding was accurate and
consistent, both experimenters participated in a two-hour training session, which included a
review of the construct definitions for each dimension as well as the coding of two practice
teams. The experimenters then coded 10 (14%) of the teams together. Cohen’s (1960) 
provided an index of interrater agreement. In this study,  = .78 for directory updating errors,  =
.76 for information allocation errors, and due to the relatively small number of cases,  = 1.00
for retrieval coordination errors, which indicated acceptable levels of agreement (see Landis &
Koch, 1977). The remaining 59 teams were divided between the two experimenters.
Team-interaction mental models. According to Kozlowski and Klein (2000), teaminteraction mental models are configural in nature, capturing an array of divergent contributions
to the whole rather than convergent perceptions among the members of a unit. While several
techniques have been developed to measure team-interaction mental models (see Mohammed et
al., 2000), this study adapted Ellis’ (in press) concept mapping technique. Team members were
given a task scenario accompanied by eight blank spaces (two per team member) that needed to
be filled with one of ten concepts that represented different aspects of the task domain. Team
members completed the map by placing concepts that best represented the actions of each team
member on the map, allowing for more than one correct response.
The concept map was constructed specifically for this study and based on team members’
specialties during the experimental task. It was designed to assess whether team members had an
understanding of their teammates’ roles and responsibilities (e.g., when the highly restricted area
contains one of each type of track, DM2 should (1) launch a tank and (2) disable the B track).
Transactive Memory Errors 17
To calculate the similarity between team members’ concept maps, one point was given
when two team members shared two linked concepts (A-B), three points were given when three
team members shared two linked concepts, and nine points were given when all four team
members shared two linked concepts. Accuracy scores were assessed by a judge who was a
subject matter expert in the DDD command and control simulation; a technique adapted from
Marks et al. (2000) and Ellis (in press). Each team member’s concept map was rated from 1
(inaccurate) to 7 (highly accurate). The judge paid particular attention to (1) the critical DDD
functions, (2) the appropriate role assignments for each team member, and (3) a reasonable
sequence of actions for successful completion of the task. Team accuracy scores were formed by
taking the sum of the four team members’ scores.
Emergent cognitive manifestations of transactive memory. The emergent cognitive
manifestations of transactive memory represent shared properties and capture the convergent
perceptions among the members of a unit. As a result, specialization, credibility, and
coordination were measured using the scale developed by Lewis (2003). The scale contains 15
items (5 items per dimension) designed to assess specialization (e.g., “Different team members
are responsible for expertise in different areas”), credibility (e.g., “I trusted that other members’
knowledge about the project was credible”), and coordination (e.g., “Our team worked together
in a well-coordinated fashion”). Each item was scored on a 5-point Likert-type scale ranging
from 1 (strongly disagree) to 5 (strongly agree). Coefficient alpha for this study was .73 for
specialization, .71 for credibility, and .73 for coordination. Aggregation was supported though
the use of intraclass correlation coefficients (see Klein et al. 2000). For specialization, ICC(1) =
.29 and ICC(2) = .62. For credibility, ICC(1) = .31 and ICC(2) = .65. For coordination, ICC(1) =
.40 and ICC(2) = .73.
Transactive Memory Errors 18
Team performance. The measure of team performance in this study was adapted from
Ellis, Hollenbeck, Ilgen, Porter, West, and Moon (2003) and focused on the team’s main
objective, which was to maximize the number of points represented by offensive and defensive
scores. The offensive score started at 1,000 points and went up by 5 points every time an enemy
track was disabled within one of the restricted zones. In an enemy track was disabled in the
neutral space or a friendly track disabled, the offensive score dropped by 25 points. The
defensive score started at 50,000 points and decreased 1 point for every second an enemy resided
within the restricted zone and 2 points for every second an enemy resided within the highly
restricted zone. Team performance was measured by standardizing and combining the offensive
and defensive scores.
RESULTS
Means, standard deviations, and intercorrelations among all the variables included in the
hypothesis tests are included in Table 2.
---------------------------------Insert Table 2 about here
---------------------------------Hypothesis 1 proposed that errors in directory updating, information allocation, and
retrieval coordination would negatively predict team performance. As shown in Table 2,
directory updating errors were negatively related to team performance (r = -.37, p<.01),
explaining a significant 14% of the variance. Information allocation errors (r = -.05, n.s.) and
retrieval coordination errors (r = .14, n.s.), on the other hand, failed to exhibit a significant
relationship with team performance. Therefore, Hypothesis 1 was partially supported.
Transactive Memory Errors 19
The second hypothesis proposed that directory updating, information allocation, and
retrieval coordination errors would negatively predict team-interaction mental model similarity
and accuracy. Table 2 indicates that directory updating errors negatively predicted teaminteraction mental model similarity (r = -.31, p < .01) and accuracy (r = -.43, p < .01), explaining
a significant 10% and 19% of the variance respectively. Information allocation errors failed to
significantly relate to team-interaction mental model similarity (r = -.02, n.s.) or accuracy (r =
.01, n.s.) and retrieval coordination errors exhibited non-significant relationships with both teaminteraction mental model similarity (r = -.03, n.s.) and accuracy (r = -.01, n.s.). These results
provide partial support for Hypothesis 2.
The third hypothesis proposed that directory updating, information allocation, and
retrieval coordination errors would negatively predict specialization, credibility, and
coordination. As shown in Table 2, directory updating errors negatively predicted specialization
(r = -.23, p < .05), credibility (r = -.21, p < .05), and coordination (r = -.27, p < .05), explaining a
significant 5%, 4%, and 7% of the variance respectively. However, information allocation errors
(specialization, r = -.01, n.s.; credibility, r = -.13, n.s.; coordination r = -.01, n.s.) and retrieval
coordination errors (specialization, r = -.07, n.s.; credibility, r = .13, n.s.; coordination, r = .02,
n.s.) failed to exhibit any significant relationships with the emergent cognitive manifestations of
transactive memory. Thus, Hypothesis 3 was partially supported.
Hypothesis 4 proposed that the relationship of errors in transactive memory behavior with
team performance is mediated by team-interaction mental model similarity and accuracy and
specialization, credibility, and coordination. As described in Baron and Kenny (1986), several
steps are necessary to demonstrate mediation. First, the independent variable needs to
significantly predict the dependent variable. Since directory updating errors were negatively
related to team performance, but information allocation and retrieval coordination errors were
Transactive Memory Errors 20
not, only directory updating errors satisfied the first mediational requirement. In the second step,
the independent variable needs to significantly predict the mediators, which was supported in
testing the second and third hypotheses. Third, the direct relationship of the independent variable
with the dependent variable needs to be reduced with the inclusion of the mediators. Given the
intercorrelation between the mediators in this study, researchers have suggested the best option is
to use path analysis (Humphrey, Ellis, Conlon, & Tinsley, 2004). Because path analysis allows
for simultaneous analysis, the total indirect path between the independent variable and the
dependent variable captures the mediation effect of each variable and sums them into a
coefficient that can be tested for significance. If the t-test of the indirect effect is significant,
mediation can be inferred.
Accordingly, we performed a path analysis using EQS version 5.7b (Bentler, 1995).
Team-interaction model similarity and accuracy and the emergent cognitive manifestations of
transactive memory (i.e., specialization, credibility, and coordination) were entered
simultaneously as mediators of the directory updating errors-team performance relationship.
Table 3 presents the direct, indirect, and total effects for the mediated structural model, along
with a t-test for each coefficient. The last three columns present full-model fit statistics. Given
the size of the sample, we followed researchers’ recommendations (see Hoyle and Panter, 1995)
and evaluated model fit using the standardized root-mean-square residual (SRMR), incremental
fit index (IFI), and the comparative fit index (CFI). Other indices (e.g., chi-square, goodness-offit index, adjusted goodness-of-fit index, normed fit index, Akaike Information Criterion) were
not used because they tend to behave erratically and are less robust when sample sizes are small
(see Hu & Bentler, 1995).
The path analytic results for the mediated structural model indicated that the indirect path
was significant with all indices falling within acceptable ranges. In total, the amount of variance
Transactive Memory Errors 21
in team performance accounted for by directory updating errors dropped from 14% to 1% when
team-interaction mental model similarity and accuracy and specialization, credibility, and
coordination were entered as mediators. This provides support for Hypothesis 4.
---------------------------------Insert Table 3 about here
---------------------------------DISCUSSION
This study examined the effects of errors in transactive memory behavior on team
efficiency and effectiveness. Results generally supported our hypotheses. We found that errors in
transactive memory behavior negatively predicted team performance. We also found that the
relationship between errors in transactive memory behavior and team performance was mediated
by team-interaction mental models and the emergent cognitive manifestations of transactive
memory. More specifically, directory updating errors negatively predicted team performance,
and this relationship was fully mediated by the negative effects of directory updating errors on
team-interaction mental model similarity and accuracy, as well as specialization, credibility and
coordination. These results have a number of theoretical and practical implications, which we
discuss below.
Theoretical Implications
The current study builds on recent research that has identified transactive memory as an
important predictor of team effectiveness (e.g., Austin, 2003; Cannon-Bowers & Salas, 2001;
Lewis, 2003; Moreland & Myakovsky, 2000). To date, this research has concentrated on the
positive effects of transactive memory on team performance. However, researchers have noted
that even when operating smoothly, transactive memory can result in confusion, lost information,
Transactive Memory Errors 22
and delays in information sharing (Hollingshead, 1998a; Kozlowski & Bell, 2003; Wegner,
1986). Further, the development and operation of transactive memory relies on interactions
among team members, and, as Pavitt (2003, p. 593) notes, “… group interaction is far from
perfectly reliable.” Yet, very little work has been done to examine the consequences associated
with the improper functioning of the transactive memory system. In the current study, we found
evidence of errors in all three areas of transactive memory behavior - directory updating,
information allocation, retrieval coordination. These results not only demonstrate the fallibility
of the transactive memory system, but show that errors can enter the system through any of the
three primary transactive memory behaviors.
Although previous research has acknowledged the likelihood of these errors, the
presumption has been that they can be corrected through sufficient discussion and their effects on
performance subsequently minimized. However, the results of the current study suggest that this
may be true for certain types of errors in transactive memory behavior but not others. In
particular, information allocation errors and retrieval coordination errors did not have a
significant influence on the team’s efficiency or effectiveness. Directory updating errors, on the
other hand, were associated with a number of negative consequences. These errors inhibited the
development of team-interaction models and negatively influenced specialization, coordination,
and credibility, which ultimately led to lower levels of team performance. Team members may
be able to ignore or quickly correct information allocated incorrectly or mistaken retrieval
requests, but cannot or will not challenge the validity of claims of expertise that do not impinge
on their own domain of expertise (Stasser & Stewart, 1995).
This study also extends our understanding of the relationship between action processes,
emergent states, and team outcomes. The path analytic mediation results provided support for
Transactive Memory Errors 23
the notion that emergent cognitive states (e.g., team-interaction models, cognitive manifestations
of transactive memory) intervene and transmit the influence of team action processes (e.g., errors
in transactive memory behaviors) to outcomes such as team performance (Ilgen et al., 2005;
Marks et al., 2001). Given that the current research employed a cross-section research design,
one direction for future research is to examine the dynamic interplay of these action processes
and emergent cognitive states over time.
Practical Implications
An important practical question raised by this research is how to leverage transactive
memory for team success while also minimizing or eliminating errors in transactive memory
behavior that negatively impact team efficiency and effectiveness. The challenge is that many of
the behaviors that are crucial for development and operation of the transactive memory system,
such as communication and information sharing, also appear to create the potential for errors.
One way to deal with this dilemma may be through training. Teams can be provided training on
communication skills to help them understand information exchange networks and how to utilize
these networks to enhance information sharing (Stevens & Campion, 1994; Ellis, Bell, Ployhart,
Hollenbeck, & Ilgen, in press). The goal would be to not only promote frequent communication
but also influence the quality or nature of that communication. For instance, the communication
training could teach team members how to use group discussion to disclose and demonstrate
their expertise and to learn about the domains of expertise of other group members (Hollingshead
& Brandon, 2003; Wegner, 1987).
Our results also suggest that teams that evidence underdeveloped team-interaction models
or low levels of specialization, credibility, or coordination may be suffering from directory
updating errors. Thus, by preventing these errors organizations may be able to enhance their
Transactive Memory Errors 24
team’s efficiency. One option is to use organizational policies or procedures to prevent directory
updating errors. For example, job descriptions and written manuals can be used to more clearly
delineate what different individuals know within a group (Hollingshead & Brandon, 2003), or
leaders can be tasked with ensuring that new “entries” in the team’s directory are accurate.
Alternatively, technology can be used to help structure the directory and ensure the accuracy of
new entries. For example, a knowledge management system could be created that provides
computerized “yellow pages” of members’ expertise (Alavi & Tiwana, 2002). A verification
process could be used to ensure the accuracy of all new entries to the system.
Limitations and Future Directions
Several limitations of this study can be identified to help guide future research. First, the
current study focused on a specific type of team – action teams. While it is important to
understand the implications of errors in transactive memory behavior within action teams, the
characteristics of these teams (e.g., high differentiation, high integration) are not shared by all
teams (Sundstrom, De Meuse, Futrell, 1990). It will be important, therefore, for future research
to examine the extent to which the current findings extend to other team settings. For example,
directory updating errors may be less prevalent when team members share areas of expertise
(e.g., low differentiation) and they may be less detrimental to team performance when the
workflow within the team is characterized by low levels of interdependence. Future research
that explores such issues will help identify the boundary conditions for the current findings.
A second issue that should be examined in future research concerns the impact of a
team’s life-cycle on errors in transactive memory behaviors. In the current study, teams were in
initial stages of development and members were inexperienced with one another and their own
role. One could argue that errors in the transactive memory system are more prevalent or potent
early in a team’s life-cycle. Wegner (1986, p. 194), for example, argues that “Faulty location
Transactive Memory Errors 25
knowledge on either member’s part dooms the system, allowing items to pass through the group
without being stored. Early in the relationship, this is to be expected; later on, it signals the
improper construction of the transactive memory system and can result in chronic memory
failure.” However, even in more established teams, change is constant as teams face dynamic
tasks and turnover in membership. Future research is needed to examine the prevalence and
importance of errors in transactive memory behaviors at different stages of teams’ life-cycles, as
well to identify the primary sources of the errors at different stages.
Third, future research is needed to understand how teams manage errors in transactive
memory behaviors. In the current study, information allocation errors and retrieval coordination
errors did not significantly predict the emergent team cognitive states or team performance. One
potential explanation for these findings is that teams were able to correct for these errors before
they could exert a detrimental influence. While this issue is beyond the scope of the current
investigation, future research on the error correction process may help inform interventions
designed to help teams prevent or manage errors in transactive memory behaviors.
Finally, because this study was conducted in a laboratory context, future research needs
to examine the external validity of these results. The laboratory allowed us to closely observe
team members’ behaviors and precisely map those behaviors onto the team’s cognitive states and
performance, thereby serving as a meaningful venue for testing our hypotheses. We are simply
asking the “can it happen” question, which according to Ilgen (1986), is exactly the type of
question that bears investigation in this type of a laboratory setting. Our results suggest that
errors in transactive memory behaviors can have important implications for teams, but whether
these results will generalize beyond the current context is an open question. We hope that future
research will extend our findings by further examining the effects of errors in transactive
memory behaviors across a variety of team and task contexts.
Transactive Memory Errors 26
Endnotes
The data utilized in this research is part of a larger study designed to examine the utility
of two team training interventions. The training manipulation was not of interest in the current
study. Analyses revealed that controlling for the manipulation did not significantly change the
results nor did it alter the substantive conclusions of the current research.
Transactive Memory Errors 27
References
Alavi, M., & Tiwana, A. (2002). Knowledge integration in virtual teams: The potential role of
KMS. Journal of the American Society for Information Science and Technology, 53, 10291037.
Hollingshead, A. B., & Brandon, D. P. (2003). Potential benefits of communication in
transactive memory systems. Human Communication Research, 29, 607-615.
Pavitt, C. (2003). Colloquy: Do interacting groups perform better than aggregates of
individuals? Why we have to be reductionists about group memory. Human Communication
Research, 29, 592-599.
Transactive Memory Errors 28
TABLE 1
Summary of Vehicles and Tracks
Vehicles
Tracks
Duration
(in min.)
Speed
Vision
Power?
Identify
ring?
Tank
8:00
slow
yes (5)
no
Helicopter
Jet
AWACs
4:00
2:00
6:00
medium
very fast
fast
very
limited
limited
far
very
far
yes (3)
yes (1)
no
no
no
yes
Speed
Power
(Exp.
Tasks)
Power
(Training)
Nature
Slow
Slow
Slow
Slow
low (1)
high (5)
med. (3)
none
none
low (1)
med. (3)
high (5)
Enemy
Enemy
Enemy
Friendly
Vehicles
Tracks
A
B
C
D
Note: For vehicles: duration = amount of time a vehicle may stay away from the base before
needing to refuel, speed = how fast the vehicle travels across the task screen, vision = refers to
the range of vision the vehicle has to both see and identify tracks, power = the ability of the
vehicle to engage enemy tracks, identify ring = the ability of the vehicle to identify tracks. For
tracks: speed = how fast the track travels across the task screen, power = the level of power
needed to successfully engage the tracks, nature = whether the track is an enemy or friend.
Transactive Memory Errors
29
TABLE 2
Means, Standard Deviations, and Intercorrelations Among Variables of Interest
Variable
Mean
SD
.59
1.39
--
.30
.99
-.11
--
.01
.12
-.05
-.04
--
18.23
8.85
-.31**
-.02
-.03
--
5.79
.93
-.43**
.01
-.01
.86**
--
6. Specialization
3.64
.41
-.23*
-.01
-.07
.24*
.15
.--
7. Credibility
4.05
.33
-.21*
.13
-.13
.30**
.28**
.29**
--
8. Coordination
9. Team
Performance
3.27
.45
-.27*
-.01
.02
.24*
.27*
.05
.51**
--
.00
1.68
-.37**
-.05
.14
.39**
.52**
.10
.33**
.47**
1. Directory
Updating Errors
2. Information
Allocation Errors
3. Retrieval
Coordination Errors
4. Team-Interaction
Mental Model Similarity
5. Team-Interaction
Mental Model Accuracy
Note: N=69. *p<.05, ** p<.01. One-tailed tests.
1
2
3
4
5
6
7
8
9
--
Transactive Memory Errors
30
TABLE 3
Mediation Effects Decomposition
Model Statistics
Independent
Variable
Directory
Updating
Errors
Direct Effect
(Unmediated
effect)
t
(Direct)
Indirect Effect
(Mediated effect)
t
(Indirect)
Total Effect
IFI
CFI
SRMR
-.09
-.81
-.29
-3.39**
-.38
.99
.99
.08
Note: N=69. Model examines team-interaction model similarity and accuracy and specialization, credibility, and coordination as
mediators of the directory updating errors-team performance relationship. IFI = incremental fit index; CFI = comparative fit index;
SRMR = standardized root-mean-square residual. *p<.05, **p < .01.
Transactive Memory Errors 31
FIGURE 1
The DDD Grid, Including Bases, Vehicles, and Tracks