Content uploaded by Sirkka L. Jarvenpaa
Author content
All content in this area was uploaded by Sirkka L. Jarvenpaa
Content may be subject to copyright.
Organization Science
Vol. 19, No. 2, March–April 2008, pp. 260–276
issn 1047-7039 eissn 1526-5455 08 1902 0260
informs®
doi 10.1287/orsc.1070.0315
© 2008 INFORMS
Knowledge Collaboration Among Professionals
Protecting National Security: Role of Transactive
Memories in Ego-Centered Knowledge Networks
Sirkka L. Jarvenpaa
Center for Business, Technology and Law, McCombs School of Business, University of Texas at Austin, Austin, Texas 78712,
sirkka.jarvenpaa@mccombs.utexas.edu
Ann Majchrzak
Information and Operations Management, Marshall School of Business, University of Southern California,
Los Angeles, California 90089, majchrza@usc.edu
Current social cognition models of knowledge coordination based on transactive memory systems (TMS) theory have
not generally considered conditions in which goals among partners are incongruent, and that those with specialized
knowledge will not necessarily act to share their knowledge. As expected from previous literature, when facing a problem
requiring inputs from others, an individual will draw on her personal or ego-centered network using the knowledge of her
network’s TMS; however, we theorize that the mixed motives within her network will cause the individual to also take into
account her perception of the level of distrust within the network when combining the received knowledge from others in
the network. Moreover, an individual’s view of her network’s TMS will be shaped not by specific policies or enforcement
mechanisms, but by semistructures for how knowledge is disseminated, owned, and discussed. Our theory is supported
based on a survey of security professionals responding to national security threats. The findings encourage a reexamina-
tion of certain assumptions of TMS theory, as well as extending theories of ego-centered networks and social-cognitive
information processing to include how individuals manage the knowledge-sharing/protection tension in interorganizational
collaborations.
Key words: interorganizational collaboration; transactive memories; security management; ego-centered network
History: Published online in Articles in Advance January 25, 2008.
Introduction
Professionals are known to seek knowledge from their
own personal networks—ego-centered networks—which
extend beyond the formal organizational structures (Burt
1980, 1992; Cross and Cummings 2004; Cross and
Sproull 2004) in response to the need for rapid ad hoc
knowledge collaboration. The ego-centered networks are
comprised of ties with whom the professionals (the egos)
have had some prior professional contact (Wellman
1999). Sometimes, the network will consist of mem-
bers of public and private organizations with multiple
and often conflicting interests (Gal-Or and Ghose 2005).
Early AIDs researchers were found not to share their
knowledge with other institutions so as to hoard reputa-
tion rewards (Kramer 1999b). Similarly, studies of pro-
fessionals protecting national security have identified the
conflicting interests that create barriers when interorga-
nizational collaboration is needed as threats dynamically
evolve with short time horizons, even when profession-
als reach out to individuals in their ego-centered net-
works (GAO 2006, Goodman and Wilson 2000). How
can knowledge of others be integrated in these short time
horizon, ad hoc, mixed-motive collaborations that cross
organizational boundaries? This is a research area of
national strategic importance but also has broader impli-
cations for theory development on knowledge networks.
Individuals may generally trust members of their
interorganizational ego-centered networks in terms of
member competence, but distrust them in terms of their
own interests. They may be watchful of others’ interests
and vigilance fearing that others may expropriate val-
ued knowledge (Heiman and Nickerson 2004, Norman
2002), leak a trade secret (Hannah 2005), or otherwise
harm the long-term reputation of the individual or his
organization (Scott and Walsham 2005).
One way that individuals may manage this concern
about mixed motives is through the use of a cogni-
tive structure—transactive memory systems (TMS)—for
who knows what and who knows who knows what
(Wegner et al. 1991). A TMS facilitates task perfor-
mance and learning through coordination efficiencies
(Faraj and Sproull 2000, Lewis et al. 2005, Wegner
1986). It may help an individual to distinguish between
helpful and potentially harmful combinations of pieces
of information gathered from different individuals during
different encounters in different contexts with different
intentions.
260
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
Organization Science 19(2), pp. 260–276, © 2008 INFORMS 261
Existing TMS literature has assumed first that mem-
bers share the same goals, and second that those with
certain expertise and assigned tasks will accept the
responsibility to act based on that expertise (Brandon
and Hollingshead 2004). This assumption of congruence
in knowledge and action cannot be made in a mixed-
motive situation with dynamically changing problems.
Others may know something, but not have the necessary
rights to share it for the specific context or problem, or
may not share it because it may unduly harm their own
organizations. What are the necessary factors for devel-
oping transactive memory of professionals when congru-
ent interests cannot be assumed? Although others have
studied TMS in large and volatile groups (Ren et al.
2006) and interdisciplinary teams (Akgun et al. 2006,
Faraj and Xiao 2006), we know of no empirical research
prior to ours that has studied the development and use
of a professional’s TMS in mixed-motive interorganiza-
tional collaborations.
In this paper we integrate research on knowledge net-
works, trust, and distributive cognition to delineate the
semistructures that build awareness of others’ expertise
and improve interorganizational collaborations involving
security professionals. These semistructures encompass
practices and procedures that affect how other parties
interpret shared knowledge, whether other parties own
the knowledge, and whether other parties agree to pri-
vacy, dissemination, and sensitivity protocols. Although
we develop our theory in the context of security col-
laboration, it may be applicable to other mixed-motive
interorganizational knowledge collaborations.
Theoretical Model and Hypotheses
Research on ad hoc knowledge collaboration has found
that individuals often develop and rely on their own ego-
centered networks in deciding with whom to collaborate
and how to collaborate (Reagans and McEvily 2003).
Individuals’ ego-centered networks may have a large
membership and diverse expertise, particularly among
professionals with a long tenure and varied problem sets
(Borgatti and Cross 2003, Burt 1992, Cross and Cum-
mings 2004). When an individual must quickly draw on
the knowledge of members in an ego-centered network
with mixed motives, benevolence cannot be assumed. In
fact, benevolent action in a mixed-motive situation may
be interpreted as having harmful intentions (Korsgaard
et al. 2002). The lack of benevolence complicates build-
ing the transactive memory and rendering it actionable
in a knowledge collaboration.
Definition and Perspective of TMS
A TMS is traditionally defined as “(a) an organized
store of knowledge contained entirely in the individual
systems of group members, and (b) a set of knowledge-
relevant transactive encoding, storage, and retrieval pro-
cesses that occur among group members” (Hollingshead
2001, p. 1080). A TMS has two components: (1) internal
memory, or what the individual members know person-
ally, and (2) external memory, or what the individuals
know about what is known by other team members or
can be located and retrieved from various storage devices
(Wegner 1986). When a TMS has been developed for a
network, individuals can specialize in different knowl-
edge domains, yet locate and integrate the specialized
expertise of others, leading to increased coordination effi-
ciencies (Brandon and Hollingshead 2004, Lewis et al.
2005, Liang et al. 1995, Wegner 1986).
TMS research has identified three indicators of the
level of development of a TMS (Lewis 2003, Liang et al.
1995, Moreland and Argote 2003): (1) expertise spe-
cialization, (2) competence-based trust, and (3) exper-
tise coordination. The more developed the TMS, (a) the
greater the tendency for groups to delegate responsibility
and specialize in different knowledge domains, (b) the
higher the beliefs about the competence or the validity
of a member’s expertise, and (c) the higher the abil-
ity of team members to coordinate their work efficiently
based on the knowledge of who knows what. Even
though TMS theory was originally developed for dyads
in close relationships (Wegner 1986) and small, well-
defined interacting groups (Liang et al. 1995), Anand
et al. (1998) extended TMS theory to settings where
knowledge is distributed among people who belong to
groups both inside and outside organizational bound-
aries. In such settings, a TMS exists at individual, group,
and organizational levels (Moreland and Argote 2003).
We suggest that TMS may also be used to describe indi-
viduals’ mental models of their ego-entered networks.
In ad hoc knowledge collaborations, Moreland and
Argote (2003) suggest that TMS need to be developed
not at the level of the problem-specific collaboration, but
rather at a higher organizational level to provide more
stability for understanding the expertise of future collab-
orators. For ad hoc collaborations, this higher-order level
may be the ego-centered network. Because members
of the network are only known to the focal individual
(Macdonald and Piekkari 2005), it is the responsibility
of the individual to learn who knows what and to use
that knowledge to decide what knowledge can or cannot
be shared with others in the network (Scott and Walsham
2004). This suggests that the TMS of an individual’s
ego-centered network may affect her ability to coordi-
nate with others in her network.
The Effect of TMS on an Individual’s Ability to
Combine Others’ Knowledge from the
Mixed-Motive Network
Knowledge collaboration requires what Kogut and
Zander (1992) call combinative capabilities. Such capa-
bilities include both the know-how about the process
by which problems that involve combining different
sources of expertise are solved, and the know-why of
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
262 Organization Science 19(2), pp. 260–276, © 2008 INFORMS
cause-effect relationships that explain what expertise is
needed. One antecedent of combinative capabilities men-
tioned by several authors has been TMS among those
from which knowledge is obtained (Borgatti and Cross
2003, Kogut and Zander 1992).
In conditions of mixed motives, combinative capabil-
ities are likely to be affected by perceived risks of ex-
changing knowledge with others. To reduce these risks,
network research has focused on the role of trust at
the individual and organizational levels (Carley 1991,
McEvily et al. 2003, Tsai 2001, Zaheer et al. 1998).
Trust is a multidimensional concept with different com-
ponents facilitating different behaviors (Mayer et al.
1995, Zaheer et al. 1998). Whereas benevolence-based
trust is fundamental for collaborative behavior (Blau
1964), in a mixed-motive situation, benevolence cannot
be assumed. Although a well-developed TMS embeds
trust in the source’s competence (competence-based
trust), it does little to reduce uncertainty over the lack
of benevolence or the conflicting interests to which the
knowledge may be applied. The lack of benevolence
necessitates wariness, skepticism, vigilance, and watch-
fulness over other’s interest (Lewicki et al. 1998), and
has implications for combinative capabilities.
In situations lacking benevolence, Lewicki et al. (1998)
argue that people develop a situationally specific cog-
nitive assessment, and label this assessment “distrust,”
defined as “confident negative expectations regarding
another’s conduct” (p. 439). Focusing on benevolence-
based distrust, we define it as confident negative expec-
tations about others’ interests that may harm or damage
one’s own interest. Benevolence-based distrust keeps
the professional alert for possible alternative motives,
aware of the possibility of suspicious situational cues,
and engaged in more sophisticated analyses of suspi-
cious situations (Fein 1996). Kramer (1999a) argues that
distrust can promote more active and mindful process-
ing of information. At moderate levels, benevolence-
based distrust can help an individual manage hazards in
mixed-motive interorganizational knowledge collabora-
tion. Therefore, we hypothesize:
Hypothesis 1 (H1). In mixed-motive, interorganiza-
tional collaborations, an individual’s combinative capa-
bility will be higher when the individual has a more
developed sense of the TMS of his ego-centered network
and there is greater benevolence-based distrust of others
in the network.
Antecedents of TMS Development with
Mixed Motives
Existing TMS theories suggest that the best way to con-
struct, evaluate, and use a TMS is through shared face-
to-face experiences such as joint training (Liang et al.
1995, Moreland and Levine 2000). When team mem-
bers are trained together, rather than apart, they are able
to better locate, integrate, and use each other’s skills
and knowledge. However, with ad hoc problem-specific
collaborations, there is little time for joint training, and
the collaborators may not have previously shared expe-
riences (Moreland and Argote 2003).
Although techniques other than training have been
suggested (Hollingshead 2001, Moreland and Argote
2003), these techniques rest on the assumption that
knowing what others know indicates how they will act.
TMS literature has largely assumed that members have
an interest in acting on the knowledge that the group
believes they have. However, such congruence of knowl-
edge and action cannot be assumed in mixed-motive
networks where members often have no interest in par-
ticipating in certain sharing activities. To protect them-
selves, they may act in ways that optimize a motive not
known to others in the network. Therefore, in mixed-
motive collaborations, antecedents for the development
of a TMS must extend beyond joint training.
According to Okhuysen and Eisenhardt (2002), pre-
dictability about others’ behaviors comes from “semi-
structures.” Semistructures are simple or minimalist
rules that help members of a group organize their knowl-
edge integration processes, yet remain flexible enough to
adapt to an evolving situation. Problem-solving methods
are an example of a semistructure. Such semistructures
may be generated in ego-centered networks or imposed
by powerful parties. We argue that three semistruc-
tures—dialogic practices, clarity of knowledge owner-
ship, and knowledge dissemination protocols—may help
reduce situational ambiguity and increase predictabil-
ity about how others will act given what they know
and hence facilitate development of a TMS in a mixed-
motive network.
Dialogic Practices. Intraorganizational teams struggle
with ambiguity and differences in interpreting the same
knowledge (Carlile and Rebentisch 2003, Dougherty
1992, Te’eni 2001). Professionals in mixed-motive inter-
organizational collaborations face these challenges to a
greater extent because of different communication norms,
thought worlds, practice sets, and domain principles in
different organizations. The result can be unsurfaced dif-
ferences in the interpretation of incoming information
and unexpected differences in actions. Unless these dif-
ferences surface, actions cannot be linked to expertise
and the TMS updated.
One semistructure for surfacing differences in inter-
pretation may be the use of dialogic practices (Boland
et al. 1994, Faraj and Xiao 2006, Te’eni 2001). Dialogic
practices are semistructures that describe rules of con-
versation. Based on principles of hermeneutic inquiry,
Boland et al. (1994) offer several “rules” of dialogic
practices that include discussing sources of knowledge,
encouraging knowledge emergence, comparing multi-
ple perspectives, keeping knowledge indeterminant to be
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
Organization Science 19(2), pp. 260–276, © 2008 INFORMS 263
repeatedly revised in response to new information, and
structuring discussions to move between summary-level
knowledge and detailed analysis. Research has shown
that the use of dialogic practices increases the amount
of knowledge shared (Majchrzak et al. 2005), heedful
interrelating (Weick and Roberts 1993) and joint sense
making (Faraj and Xiao 2006) by causing participants to
think about and reflect on the expertise of others and to
critically assess motives for sharing unique knowledge.
Therefore, the use of dialogic practices may increase
predictability of others’ behaviors by helping individuals
to better link others’ knowledge to their actions. In net-
works using such practices, the TMS may be perceived
as more fully developed.
Hypothesis 2A (H2A). In mixed-motive, interorga-
nizational collaborations, the more that individuals use
dialogic practices, the more the TMS of their ego-cen-
tered network will be perceived as developed.
Clarity of Knowledge Ownership. The traditional view
of TMS assumes that the location of knowledge (i.e., who
knows what) implies expert ownership of that knowledge.
When problems such as security threats are confronted
by drawing on the knowledge of members in an interor-
ganizational mixed-motive network, the knowledge that
is shared is likely to evolve over time as new informa-
tion about the problem is discovered and multiple inter-
pretations of the incoming information are shared. This
creates ambiguity over who has what responsibilities
and rights to that knowledge (Ulmer and Sellnow 2000),
such that an individual may know something but may not
have the right to use or share that knowledge. However,
what knowledge is considered “owned” and what owner-
ship means varies across situations and different member
intentions (Jarvenpaa and Tanriverdi 2006). Ideas gener-
ated exclusively within a company might be perceived
as “owned” by the company, whereas ideas generated
in collaboration with members of other companies may
be perceived as sharable with those members (Jarvenpaa
and Staples 2001). Making judgments about owner-
ship is challenging, even in nonemergent situations.
In dynamically evolving situations, such judgments are
likely to be ambiguous which in turn reduces the ability
to link knowledge and action.
Policies that clarify ownership issues increase pre-
dictability by reducing divergence in views and in-
creasing accountability (Ulmer and Sellnow 2000). In
mixed-motive collaborations, clarity over ownership of
knowledge should increase congruence between knowl-
edge and action. When it is clear that a party owns
knowledge, and what that ownership means, an infer-
ence is made that the party has the right and responsi-
bility to act on that knowledge. Clarity over ownership
may reduce fears of accidental or intentional infringe-
ment on another’s property when distributed knowledge
is coordinated and integrated. Therefore, by increasing
congruence between knowledge and action, ownership
clarity may foster a more developed TMS.
Hypothesis 2B (H2B). In mixed-motive, interorga-
nizational collaborations, the more individuals per-
ceive that knowledge ownership is clear, the more the
TMS of their ego-centered network will be perceived as
developed.
Knowledge Dissemination Protocols. Finally, TMS
theory assumes that when people own knowledge, they
should share it (Lewis 2003). However, in mixed-motive
settings, not all knowledge should be shared with every
member of the network or in a collaboration. With-
holding some knowledge may protect members of the
network; conversely, withholding the wrong knowledge,
or sharing sensitive knowledge too broadly, may hurt
the network and individual firms within the network
(GAO 2006, Kramer 1999a). Therefore, semistructures
that clarify how knowledge should be disseminated
within the network may be needed in such settings to
reduce ambiguity and allow members to quickly iden-
tify discrepancies between knowledge and action. These
semistructures would need to be fluid enough to adjust
to an evolving situation, but rigid enough to increase
predictability about how others will behave with their
knowledge, thus helping to develop a TMS.
Hypothesis 2C (H2C). In mixed-motive, interorgani-
zational collaborations, the more individuals perceive
that knowledge-dissemination protocols are adequate,
the more the TMS of their ego-centered network will be
perceived as developed.
Because ego-centered networks are not formal organi-
zations, these semistructures are unlikely to be imposed
by external governing bodies, or to be contractually bind-
ing. They are not even going to take the shape of for-
mal administrative coordination procedures (Faraj and
Sproull 2000). Rather, the informal nature of the per-
sonal networks calls for flexible protocols that help to
create the understanding of responsibilities, priorities,
and risks to build the confidence to act. Dougherty
(2007) describes such rules as enabling professionals to
“go on” in their work life, even in unforeseen situations,
and make the right calls. Thus, as suggested by Brown
and Eisenhardt (1997) and Okhuysen and Eisenhardt
(2002), these semistructures may be effective because
they flexibly allow emergent approaches to task execu-
tion by providing simple general rules focused on knowl-
edge integration rather than on the task. Such protocols
would be adequate when others know the protocols and
use them appropriately.
Antecedents to Dialogic Practices
Unlike the semistructures of knowledge ownership poli-
cies and dissemination protocols, dialogic practices re-
quire intensive effort and time to engage members within
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
264 Organization Science 19(2), pp. 260–276, © 2008 INFORMS
the network—time that takes members away from the
responsibilities of their own organizations. Therefore,
motivation and ease of communication must exist to
encourage members to engage in dialogic practices.
Motivation is critical in reducing ambiguity between
knowledge and action within the firm (Argote 1999,
Szulanski 1996). The prospect of acquiring new knowl-
edge has been shown to motivate people to share their
own knowledge on a corporate intranet (Kalman et al.
2002), in communications (Te’eni 2001), and in interfirm
strategic alliances (Norman 2002). Professionals asso-
ciated with organizations with a culture that embraces
learning from outside sources may be more likely to
engage in dialogic practices because they see such dia-
logue as not only potentially beneficial to them, but also
to the organization. This learning intent may be man-
ifested in organizational human resource practices that
support permeable boundaries to learning from outside
(Swart and Kinnie 2003) or from cultural or manage-
ment practices that encourage learning.
Hypothesis 3A (H3A). In mixed-motive, interorga-
nizational collaborations, the greater the learning intent
of the organization to which the ego party belongs, the
more the individual will engage in dialogic practices.
Dialogic practices may also be facilitated by ease of
communicating with others in the network. In complex
conditions in which diverse information is being con-
veyed and where a variety of media will be needed to
transfer the knowledge (Boland et al. 2001), there is
unlikely to be any one particular communication media
that outperforms others in the collaboration (Massey
and Montoya-Weiss 2006). Thus, ease of communication
may be facilitated by the presence and use of multiple
communication channels, including phone calls, e-mails,
exchanges of documents, and face-to-face meetings.
Hypothesis 3B (H3B). In mixed-motive, interorgani-
zational collaborations, the greater the use of multiple
channels of communication, the more the individual will
engage in dialogic practices.
Control Variables
In addition to the antecedents identified in this paper
that specifically focus on a mixed-motive condition, we
include three controls. Two of these—task interdepen-
dence (Brandon and Hollingshead 2004, Hollingshead
2001) and network size (Moreland and Levine 2000,
Ren et al. 2006)—are likely to affect the development of
TMS. Length of time the individual has been associated
with a network is included as a control on combinative
capability because the longer the tenure in the network,
the more the accumulated practice and expertise needed
for combining knowledge of others from the network
(von Hippel 1988).
Method
Study Context
We spent approximately 18 months building relation-
ships with security personnel in the private and public
sectors, and identified the U.S. Federal Bureau of Inves-
tigation (FBI) InfraGard program as a possible source of
respondents for our survey. InfraGard encourages ad hoc
private and public collaborations among security profes-
sionals. InfraGard consists of an e-mail distribution list
of individuals cleared to receive high-security informa-
tion about security-related threats or potential threats, as
well as discussion threads and regional chapter meetings
to encourage individuals from various organizations to
collaborate and share knowledge in response to specific
security threat information. The governing bodies of two
InfraGard chapters agreed to allow us to conduct the
study within their chapters.
We initially interviewed the sponsor at each chap-
ter and 10 other individuals involved in the security
domain familiar with InfraGard to understand who they
collaborated with, how they shared knowledge, and the
consequences of sharing. We learned that even though
InfraGard was initially set up to be a virtual institutional
network for collaboration, members felt that it was just
an e-mail distribution list of security alerts. InfraGard
did not serve as an important collaborative forum for
ad hoc security collaboration for any of the 12 peo-
ple interviewed. One interviewee notes, “If there were
a threat, the last thing I would do is inform anyone
on InfraGard.” Interviewees relied on extensive personal
networks across many organizations and individuals and
included some members of InfraGard, but also other for-
mal and informal sources.
Interviewees felt the tension between knowledge shar-
ing and protection such as “knowing what information
to share and ensuring that others receiving the infor-
mation know how to use it.” Professionals also feared
knowledge leaks. One informant shared a bitter experi-
ence: “Sharing the wrong information is disastrous. At
my company, blabbermouths published their work in the
company newsletter, which got included in a proposal
from another company that beat us out.” Another respon-
dent reported how he had saved his own company $90
million. “What we learned from listening to them was
that when improving a process, the incremental value
they had gained from documenting their process in detail
was marginal. So we took $10 M to do what they took
$100 M to do. They gave away too much.”
Several interviewees mentioned the important role
of transactive memory in their networks. For exam-
ple, one said, “I still prefer my informal network of
experts. They may float from company to company, but
I know them, they know me, and we can trust each
other to keep secrets and know what to do in case of an
event.” The need for structures that organize how pro-
fessionals share and protect knowledge was echoed in
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
Organization Science 19(2), pp. 260–276, © 2008 INFORMS 265
our interviews. One noted, “Policies and procedures are
needed that assure reliable vetting make sure neces-
sary agreements are in place MOA, [memo of under-
standing], nondisclosure agreements, corporate indemni-
fication agreements, privacy policies, etc.” Another said,
“For me to share, there needs to be assurance that infor-
mation can be shared in a way that protects all par-
ties from embarrassment, humiliation, or brand damage.”
Others noted inadequate policies or procedures as rea-
sons for not sharing. With regard to the use of multiple
communication channels, interviewees mentioned using
a variety of channels to communicate in an effort to
overcome the shortcomings of any single channel.
Survey Design
The InfraGard sponsor for each chapter sent an e-mail on
the chapter’s list server to the complete membership: 500
people in one region and 120 in another, requesting them
to complete our survey, which was administered through
a university, independent of the FBI. Respondents were
informed that the survey would provide feedback to
InfraGard on ways to improve knowledge sharing among
public and private organizations, as well as within their
own networks. The survey questions directed the respon-
dents to describe their ego-centered network of security
professionals. We defined this network as: “an informal
personal network or circle of professionals interested in
security generally, or interested in a specific aspect of
security, with whom you have collaborated in the past to
better understand new security information or confront
a security risk.” Our pretests suggested that this defini-
tion provided a consistent frame of reference to personal
networks.
The survey collected data from the professionals about
their perceptions of their ego-centered networks. To
increase the reliability of responses, we asked the pro-
fessionals to first think about collaborations they had
engaged in when responding to security threats, then to
think of their own “informal personal network or cir-
cle of professionals interested in security” that constitute
their list of potential collaboration partners. We fur-
ther grounded responses by asking them to think about
how many people were in this network, and background
information about the network (e.g., the percentage who
were also members of InfraGard, the percentage new to
the network this year, etc.). In this way, we felt we were
able to encourage the respondents to think about their
entire egocentric network, rather than specific names
of individuals they could enumerate, as had been done
in previous studies (e.g., Laumann 1966, Burt 1984,
Campbell et al. 1991). Moreover, because we were not
interested in the network’s structure or the influence or
communication role of the individual vis-à-vis his net-
work, we did not use the name generator technique to
elicit the network.
One hundred four security professionals completed the
web-based survey. It is impossible to assess the represen-
tativeness of this sample. Owing to the sensitive nature
of the e-mail distribution list, we had no access to sur-
vey respondents’ names, e-mail addresses, or other iden-
tifying information. We asked our InfraGard sponsors
to ask key informants (i.e., those security professionals
who collaborated with others and who had an interest in
providing feedback to InfraGard) whether they had com-
pleted the survey. In addition, the sponsors conducted
follow-up interviews with individuals who did not com-
plete the survey. Reasons for not responding included
that: they were not a security professional who acted on
security information (i.e., only used the e-mail distribu-
tion list to stay informed), they did not collaborate with
other security professionals outside their organization,
or they did not feel any affiliation or identification with
InfraGard. Therefore, we believe that our sample can be
characterized as consisting of active security profession-
als engaged in security-related interorganizational col-
laborations, with some interest in developing further col-
laborations through InfraGard. However, as our objec-
tive is not to assess the prevalence of the conditions that
foster TMS, but rather to examine the theoretical links
between the key constructs, the representativeness of the
sample should be less of an issue.
Survey Measures
Survey questions (see the appendix) directed the respon-
dents to describe their ego-centered network of security
professionals. The 104 respondents reported networks
with sizes ranging from 4 to 500, with a mean of 103.
On the average, 16%–30% of the members in the net-
works were new each year. These averages are com-
parable to other studies of personal networks (Hill and
Dunbar 2003, Killworth et al. 1990, McCarty et al.
2001). The respondents identified, on average, fewer
than 15% of the members of these networks as Infra-
Gard members. Individuals reported collaborating in the
past with an average of only 31%–45% of their network
members, where collaboration was defined as working
with another member as partners in a joint problem-
solving process. On average, 46%–60% of the mem-
bers in their network were geographically local (within
driving distance). Despite the geographical closeness,
the most common media for interaction among network
members were one-on-one e-mail and group e-mail lists,
followed by phone calling. Face-to-face meetings were
the least-used mode of interaction, occurring on average
“a few times per year.”
Respondents had been members of their networks for
10 years, on average, with a substantial range (from 0.5
to 42 years). They came from a variety of organizations:
61% private (versus 39% public), 50% for-profit (ver-
sus 50% nonprofit), and 31% with security as the main
line of business. They had held their security jobs, on
average, for 15 years (range of 1 to 42 years).
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
266 Organization Science 19(2), pp. 260–276, © 2008 INFORMS
Combinative Capabilities. To develop our measure
of combinative capabilities, we followed the procedure
used by Cummings (2004). We first conducted inter-
views with the 12 security professionals to identify the
types of know-how and know-why knowledge needed
in ad hoc collaborative security responses, ensuring that
the definitions of know-how and know-why were fol-
lowed. We supplemented our interviews with a review
of research to identify knowledge shared during secu-
rity threat responses. Finally, extensive piloting yielded
more refinements to our list of three know-how questions
and three know-why questions. We asked respondents
to report how frequently they had received each type
of knowledge from their network using a one-to-seven
scale.
Transactive Memory Development. We measured
transactive memory using the 10-item scale developed
by Lewis (2003). We asked individuals to rate their net-
work’s TMS based on their own personal interactions
with others outside their employing organizations.
Frequency of Use of Dialogic Practices. We adapted
the 10-item scale measuring the use of dialogic prac-
tices during interactions among network members from
Majchrzak et al. (2005) based on the five rules of
dialogic practices identified by Boland et al. (1994)—
source, easy travel, multiple perspectives, emergence,
and indeterminance—measuring each element by two
items. Because the original scale focused on information
technology support for these elements, we changed the
stem to have respondents focus on the frequency of ele-
ments among network members regardless of the media
used.
Organizational Learning Intent. We used Norman’s
(2002) five-item measure of organizational learning
objectives from strategic alliances.
Knowledge-Dissemination Protocols. We developed a
three-item scale of the adequacy of the policies indicat-
ing what knowledge to share. We measured the extent to
which the respondents found such protocols to be ade-
quate to protect sources, knowledge sharing, and infor-
mation sensitivity.
Clarity of Knowledge Ownership. We developed a
three-item scale measuring the clarity about who owns
the knowledge that is shared in the network. The items
measured the clarity of the basic rights and responsibil-
ities assumed with legal ownership (Barzel 1989).
Benevolence-Based Distrust in Network Members.
From the literature on interorganizational collaboration,
we adapted a four-item scale measuring the perceived
threat of other collaborators’ opportunism (Heiman and
Nickerson 2004). A 0.01 correlation between this scale
and the three items in the TMS instrument measuring
competence-based trust indicated the divergent validity
of the two forms of trust.
Use of Multiple Communication Channels. We asked
respondents to indicate the degree to which they used
six different communication channels to interact with the
people in their networks. These communication channels
included: (1) meeting one on one or in small-group face-
to-face meetings, (2) phone calls, (3) attending local
chapter meetings and conferences, (4) using e-mail one
on one, (5) using group e-mail lists or list servers,
(6) logging in to a portal. Channels used at least once a
month were included in the count.
Controls. Task interdependence was measured using
a standardized four-item scale (Goodhue and Thompson
1995). Years in the network was measured as a single-
item question. We measured personal network size via a
single global measure, which is considered an efficient,
reliable, and valid way to differentiate respondents, and
has been found to correspond well with more com-
plex network measures (Bernard et al. 1990, Fu 2005,
Marsden 1990, McCarty et al. 2001, Marin and Hampton
2007).
Analysis Strategy
Preliminary analyses of the model variables across the
two InfraGard chapters showed no significant differ-
ences. Accordingly, we pooled data across the two
chapters and used partial least squares (PLS), a latent
structural equation modeling technique that uses a cor-
relational, principle component-based approach to esti-
mation (Chin 1998). Each multi-item construct was
modeled as reflective (rather than formative) of the latent
variable because we expected the items measuring each
construct to covary. For example, the items correspond-
ing to organizational learning objectives measured the
underlying construct of learning. Our model exceeded
Chin’s (1998) sample-size recommendation of five to ten
times the largest number of structural paths to any one
construct. To estimate the significance of the path coef-
ficients, we used bootstrapping with a sample size of
200, as recommended by Chin (1998). We included the
three controls: the person’s tenure in the network, task
interdependence, and network size.
Results
The results are divided into a discussion of the mea-
surement model to confirm convergent and divergent
validity of constructs, and a discussion of the structural
model to test the hypothesized relationships between the
constructs.
Measurement Model
Results of the PLS component-based analysis, correla-
tions among the constructs, alpha coefficients, reliabil-
ity tests, PLS-computed variability for each construct,
and interconstruct correlations are presented in Tables 1
and 2. Of the 10 items in the TMS instrument, 4 were
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
Organization Science 19(2), pp. 260–276, © 2008 INFORMS 267
Table 1 Item Discriminant Analysis (Cross-Loadings)
Knowledge- Organizational Benevolence-
Combinative Dialogic dissemination Knowledge learning based
Items capabilities TMS practices protocol ownership intent distrust Interdependence
CAPEA 077 023 030 018 −009 028 017 020
CAPEB 084 034 024 020 −001 026 022 017
CAPEC 081 046 039 024 009 030 018 024
CAPED 086 039 037 038 013 026 015 012
CAPEE 087 034 037 030 009 028 028 018
CAPEF 085 030 025 023 013 023 024 007
TMA 021 059 037 027 008 021 001 025
TMB 032 078 054 033 012 031 007 020
TMC 035 085 042 037 022 025 001 010
TMD 032 087 042 042 030 026 −006 008
TME 028 071 039 041 041 014 −006 021
TMF 040 077 037 045 045 029 0 016
DIALA 030 049 076 026 012 027 007 025
DIALB 037 045 069 032 004 027 022 028
DIALC 026 042 085 021 004 044 015 029
DIALD 018 044 086 020 009 034 012 026
DIALE 046 055 086 037 018 033 018 025
DIALF 034 042 084 031 015 033 014 020
DIALG 034 042 082 025 013 041 009 027
DIALH 032 048 082 034 026 031 011 027
DIALI 026 039 081 025 009 027 016 026
DIALJ 031 036 079 020 011 041 013 021
DISSEMA 022 046 019 083 025 0 −010 021
DISSEMB 026 040 030 092 041 009 −004 016
DISSEMC 033 044 037 088 032 021 −004 017
OWNA 0 023 009 021 086 −008 −031 −017
OWNB 004 035 016 033 090 −004 −024 −015
OWNC 019 025 012 039 060 006 −013 007
LEARNA 024 024 035 003 −007 089 −003 010
LEARNB 030 025 039 004 −007 086 010 003
LEARNC 023 028 033 014 0 091 007 011
LEARND 028 030 039 016 −003 090 008 022
LEARNE 036 034 040 015 0 086 −003 026
BEN.-BASED 021 006 010 003 −012 −004 078 020
DISTRUST A
BEN.-BASED 014 −011 017 −018 −037 002 081 037
DISTRUST B
BEN.-BASED 017 −007 009 −011 −030 001 083 033
DISTRUST C
BEN.-BASED 029 008 019 003 −020 014 088 017
DISTRUST D
DEPENDA 010 017 026 014 −016 019 028 091
DEPENDB 017 020 027 021 −004 009 028 091
DEPENDC 024 022 030 019 −012 019 027 094
DEPENDD 017 013 027 019 −013 010 032 075
Notes. Boldface numbers are loadings (correlations) of indicators to their own construct; other numbers are cross-loadings. To calculate
cross-loadings, a factor score for each construct was calculated based on the weighted sum, provided by PLS-Graph, of that factor’s
standardized and normalized indicators. Factor scores were correlated with individual items to calculate cross-loadings. Boldface item
loadings should be greater than cross-loadings. See the appendix for actual item wording in surveys.
dropped for low loadings on the TMS construct, fol-
lowing suggestions for trimming by Gray and Meister
(2004).
Table 1 provides the correlations of each item to its
intended construct (i.e., loadings) and to all other per-
ceptual constructs (i.e., cross-loadings). Although there
is some cross loading, all items load more highly on
their own construct than on other constructs, and all
constructs share more variance with their own mea-
sures than with other constructs. Table 2 shows that the
alpha coefficients for the items within each construct
are sufficiently high, as are the more accurate compos-
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
268 Organization Science 19(2), pp. 260–276, © 2008 INFORMS
Table 2 Interconstruct Correlations: Consistency and Reliability
Benevolence-
Composite Combinative Knowledge- Learning based Multi- Network
Construct Alpha reliabilities AVE capabilities TMS Dialogic dissemination Ownership intent distrust Dependence channels size
Combinative 0.91 0.93 0.70 0.83
capabilities
TMS 0.84 0.89 0.59 0.43 077
Dialogic 0.94 0.95 0.66 0.39 055 081
practices
Knowledge- 0.85 0.91 0.77 0.31 050 033 088
dissemination
Ownership 0.70 0.84 0.64 0.09 038 016 039 080
Learning intent 0.93 0.95 0.78 0.33 032 042 012 −003 088
Ben.-based 0.85 0.89 0.68 0.27 002 017 −005 −026 006 083
distrust
Depend 0.90 0.93 0.78 0.20 022 031 021 −011 017 029 0.88
Multichannels NA 0.18 015 054 007 −009 020 026 0.38 NA
Network size NA 0.03 −011 −011 −013 −004 0 010 0.03 0 NA
Network years NA 0.01 010 010 007 003 008 016 0.13 001 −010
Notes. Boldface numbers on diagonal are the square root of the average variance shared between the constructs and their measures. Off-diagonal elements are correlations among
constructs. For single-item constructs, only correlations are presented. For discriminant validity, diagonal elements should be larger than off-diagonal elements.
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
Organization Science 19(2), pp. 260–276, © 2008 INFORMS 269
Figure 1 Results
0.27**
– 0.08
0.47**
Use of multiple
channels for
communication
Organizational
learning intent
Dialogic
practices
Knowledge
dissemination
protocols
Clarity of
knowledge
ownership
Distrust in
network
members
Level of
network’s TMS
development
R2 = 0.50
Control: Years in
network
Control: Task
interdependence
Control:
Network size
Combinative
capabilties
R2 = 0.25
0.33**
0.27**
0.43**
– 0.23**
0.07
0.18*
0.45**
ite reliabilities. Table 2 also presents average variance
extracted as well as correlations between constructs,
including the control variables. Comparing the square
root of the average variance extracted (AVE) (i.e., the
diagonals in Table 2 representing the average associa-
tion of each construct to its measures) with the corre-
lations among constructs (i.e., the off-diagonal elements
in Table 2 representing the overlap association among
constructs) indicates that each construct is more closely
related to its own measures than to those of other con-
structs. Moreover, all AVEs are above the 0.50 recom-
mended level (Chin 1998). In sum, these results support the
convergent and discriminant validity of our constructs.
Structural Model
Figure 1 is a graphical depiction of the PLS results, and
Table 3 contains the outermodel loadings of the items
on each construct. The hypothesized paths of predictors
of the use of dialogic practices are significant, account-
ing for 39% of the variance. The hypothesized paths
between the three TMS antecedents and TMS develop-
ment are also significant, accounting for 50% of the vari-
ance in TMS development. Finally, distrust and TMS
development accounted for 25% of the variance in com-
binative capability. Tenure in the network was not a sig-
nificant control variable for combinative capability, nor
was task interdependence a significant control for TMS
development. Finally, network size was a significant con-
trol, negatively related to TMS development. Because
hypothesized, combinative capability is explained in part
by TMS development coupled with distrust, TMS devel-
opment is explained in part by three semistructures; the
use of dialogic practices is explained by the communi-
cation channels and the learning intent of the ego-party’s
organization.
To rule out alternative plausible explanations, we ex-
amined the direct paths between TMS antecedents and
combinative capability. These are not significant. We
also examined paths between the use of multiple chan-
nels of communication and TMS and found all paths to
be insignificant. We checked whether task interdepen-
dence or size moderated the relationship between TMS
development and combinative capabilities and found no
significant relationships. We eliminated the 12 cases
with networks greater than 150 (a network size that Hill
and Dunbar 2003 argue is the cognitive limit for human
information processing), and the 10 cases with networks
smaller than 10 (the size of small teams), and obtained
results similar to those of the full sample.
Finally, we examined the potential for our results
to be explained by common method variance. We fol-
lowed suggestions by Podsakoff et al. (2003) to reduce
this potential in the design of the survey by ensuring
anonymity, which reduces the likelihood of bias from
social desirability and respondent acquiescence. We
also separated the predictor and criterion variables psy-
chologically, proximally, and through varying scales.
Respondents were not psychologically primed to con-
nect our predictors because the survey was introduced
as focusing on collaborations in general, not specifically
on combinative capability (Podsakoff et al. 2003). In
addition, we tested for the effect of common method
bias in three ways. First, we used the partial correla-
tion approach, a method recommended by Lindell and
Whitney (2001) and used in a variety of studies assess-
ing common method variance (cf. Pavlou and El Sawy
2006). In the partial correlation approach, the researcher
partials out for the effect of a method variance that
would be equivalent to the lowest correlation among the
variables of interest. Disattentuation did not affect the
significance of the relationships among the variables,
which suggests that the results cannot be accounted for
by common method variance. Second, we performed
Harman’s one-factor test by entering all the principal
constructs into a principal components factor analysis
(Podsakoff and Organ 1986). Eight factors resulted. The
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
270 Organization Science 19(2), pp. 260–276, © 2008 INFORMS
Table 3 Outer Model Loadings
Original
sample Mean of Standard
estimate subsamples error t-stat
Combinative capabilities
CAPEA 077 077 007 132
CAPEB 083 083 006 174
CAPEC 081 080 006 180
CAPED 086 087 003 245
CAPEE 087 089 002 426
CAPEF 085 085 003 223
TM development
TMA 060 061 010 64
TMB 078 078 005 153
TMC 085 085 005 218
TMD 087 086 004 213
TME 071 071 005 122
TMF 076 077 004 144
Dialogic practices
DIALA 076 076 005 126
DIALB 069 069 008 79
DIALC 085 086 003 309
DIALD 086 086 002 334
DIALE 086 085 003 295
DIALF 084 084 003 269
DIALG 082 082 004 203
DIALH 082 082 004 254
DIALI 081 080 005 181
DIALJ 079 078 004 190
Knowledge dissemination protocols
DISSEMA 083 082 007 126
DISSEMB 092 091 003 375
DISSEMC 088 089 003 279
Knowledge ownership clarity
OWNA 086 085 013 77
OWNB 090 088 012 90
OWNC 059 059 018 37
Organizational learning intent
LEARNA 089 088 004 200
LEARNB 086 086 004 219
LEARNC 091 090 005 159
LEARND 090 089 005 151
LEARNE 086 085 005 180
Benevolence-based distrust
DISTRUSTA 088 087 005 60
DISTRUSTB 078 076 011 51
DISTRUSTC 081 078 011 59
DISTRUSTD 083 082 009 106
Task interdependence
DEPENDA 091 088 011 79
DEPENDB 090 088 011 88
DEPENDC 094 091 010 96
DEPENDD 075 074 014 42
first accounted for 30% of the variance. The other seven
(with eigenvalues greater than one) contributed to the
remaining 40% of the variance, each accounting for
3%–11%. This suggests that while there is likely to
be some common method variance, the effect is small.
Finally, following Podsakoff et al. (2003), we performed
a single-method factor approach in PLS by having indi-
cators measure both their theoretical constructs and a
common method latent construct, and rerunning the
structural model. The results did not change.
Discussion and Implications
Our approach focuses on professionals who, faced with
a difficult problem, must be able to engage others in
ad hoc collaborations and quickly combine knowledge
from these sources to solve the problem. Profession-
als in such situations often draw on their interorganiza-
tional ego-centered networks of personal contacts. These
networks often involve members with different inter-
ests and motives. By integrating research and theory
on knowledge networks, trust, and distributive cogni-
tion, we find support for the role of TMS development
and benevolence-based distrust in explaining an individ-
ual’s ability to combine knowledge from others drawn
from these mixed-motive networks. In mixed-motive sit-
uations, TMS achieves its coordination benefits by indi-
cating not only what should be shared (because others
do not know what you might know) and what need not
be shared (because others already know it), but also
what should not be shared (since others may act in a
harmful way with that knowledge). In a context where
knowing what others know is not equivalent to action,
the hypothesized semistructures help match action and
knowledge. Finally, the results suggest that the time-
intensive activity of dialogic practices can be encour-
aged by an organization’s perceived learning intent, as
well as by providing multiple channels for communi-
cation. This study makes a contribution by identifying
ways to increase combinative capabilities of profession-
als who must draw on interorganizational ego-centered
networks where mixed motives may prevail, and where
solutions must be arrived at too quickly for classic trust-
building and expertise-building sessions. Although our
hypotheses have been tested exclusively with profession-
als protecting our national security, there are broader
implications, which are discussed later.
Limitations
Our study suffers from several limitations. We used
our own measure of ego-centered networks. Our data
were perceptual and retrospective, captured from a single
source. Our measure of ego-centered network data is sub-
ject to recall and estimation errors as well as ambiguity
of the network boundary and of the subpopulation in the
eyes of the respondents. These are not uncommon prob-
lems in ego-centered network research (McCarty et al.
2001, Marin and Hampton 2007). Although common
method variance did not appear to be a major contribu-
tor to the results, these limitations suggest that this study
should be considered exploratory. In addition, causal
direction is problematic to infer from cross-sectional sur-
vey designs. For example, it is possible to revert the
direction of the arrows in Figure 1. That is, combina-
tive capabilities may facilitate the development of TMS,
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
Organization Science 19(2), pp. 260–276, © 2008 INFORMS 271
and the more developed the TMS of the network, the
more likely that professionals will use dialogic practices,
find knowledge ownership clear, and perceive knowl-
edge dissemination protocols as adequate. Lewis et al.
(2005) research on group TMS development suggests
that causal directions are likely to be reciprocal over
time. Thus, TMS development would support and be
supported by both combinative capabilities and the sim-
ple semistructures. Although a longitudinal field study
of security professionals may help to sort out the causal
relationships, the difficulty of accessing this population
and securing adequate research participation may make
such a study unlikely.
Another limitation to consider is the sample that may
render findings unique. Professionals protecting national
security may be classified as “collectively paranoid”
(Kramer 1999a), characterized as hypervigilant in pro-
cessing information, dwelling on negative interpretations
of events, overattributing hostile intentions to others,
and exaggerating conspiracy theories. Nevertheless, it is
possible that the sample may not be as unique as one
might initially believe. Lewicki et al. (1998) argue, for
example, that the high-trust/high-distrust condition of
the ego-centered networks of security professionals, is
“the most prevalent form of multiplex working relation-
ships in modern organizations” (p. 447). Other examples
of ad hoc collaborations with mixed motives across orga-
nization include emergency response teams (Majchrzak
et al. 2007), software developers who temporarily work
together on open-source code (Grand et al. 2004), some
forms of cross-firm, temporary new product develop-
ment groups (Engwall and Svensson 2001), and vir-
tual network organizations that are quickly assembled to
respond to a client’s need for manufacturing a product
(Hagel and Brown 2005). Therefore, although security
professionals may be unique in their level of paranoia,
the temporary, interorganizational, mixed-motive nature
of ad hoc collaborations can be observed in a range of
other settings.
Knowledge Networks in Mixed-Motive,
Interorganizational Collaborations
In the interorganizational knowledge network literature,
there has been little discussion of structures except those
of a formal nature. Formal contractual governance poli-
cies between organizations may be delineating mech-
anisms that constrain the boundaries or periphery of
these networks, but fail to account for the richness of
behavioral and cognitive choices made by participants.
Policies that formally protect the intellectual property
between two companies in a joint venture may ignore
the reality that participants from those two companies
may be drawing on external sources to solve a difficult
problem. In addition, corporate policies limiting release
of sensitive information may be largely irrelevant in
an ad hoc brainstorming session when external sources
are called upon to solve an urgent problem (Majchrzak
and Jarvenpaa 2005). Moreover, and perhaps most dis-
turbing, organizational policies discouraging interorga-
nizational collaboration to protect leaks also discourage
knowledge sharing in a way that hampers interorganiza-
tional coordination and national security (GAO 2006).
Thus, by creating constraints on boundaries, existing
research not only ignores the behavioral realities of
interorganizational collaborations (such as their ad hoc
nature), but perpetuates their shortcomings.
Our findings suggest that although ego-centered net-
works are not governed by formal contracts, profes-
sionals do structure their social context to match action
and knowledge. In promoting combinative capabilities,
benevolence-based distrust helps professionals decide
what not to share and what can be shared for task
completion. Dialogic practices, adequate knowledge-dis-
semination protocols, and knowledge ownership clarity
provide structure for professionals to develop awareness
of others’ expertise in a mixed-motive situation. These
semistructures help individuals identify discrepancies in
expectations in interpreting incoming information (by
dialogic practices), in knowledge dissemination (with
protocols), and in how knowledge will be used (with
clarity over ownership). Together, these semistructures
increase expectations about what others in the network
know and do not know, and how they will act given their
knowledge.
Conceptualization of TMS in Interorganizational,
Mixed-Motive, Ego-Centered Networks
Suggesting that a network (rather than a group or organi-
zation) has a TMS exposes researchers to a richer array
of expertise that individuals may draw on when the need
for collaboration arises (Moreland and Argote 2003).
Although the negative correlation between network size
and TMS development suggests that TMS may be harder
to develop in larger networks, the positive effects of the
three semistructures suggest ways of increasing the like-
lihood of achieving the intended value from these net-
works despite the negative effects of size.
Conceptualizing networks as having TMS may have
an additional implication beyond ad hoc collaborations.
TMS theory has relied on the interdependence assump-
tion that TMS develops only when “one person relies
on another to know different information necessary for
completing a joint task” (Lewis 2003, p. 600). How-
ever, in our study, fewer than half of the network mem-
bers, on average, had been engaged in any joint task
with the responding professionals. Yet the professionals
were able to develop a TMS for the network (with help
from the semistructures). Moreover, task interdepen-
dence was not significantly related to TMS development.
We believe the respondents developed a TMS in the
absence of joint tasks because they based their depen-
dence on future potential opportunities rather than cur-
rent task needs. Moreland and Argote’s (2003) research
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
272 Organization Science 19(2), pp. 260–276, © 2008 INFORMS
demonstrating that information about an imminent per-
sonnel turnover negatively affects a TMS, and Lewis’
(2003) study showing that a TMS grows stronger as time
passes, could be interpreted as support for the suggestion
that expectations about future dependencies may impact
TMS development. We therefore suggest that limiting
TMS applicability to current dependencies may be too
constraining.
Temporary membership in dynamic organizations re-
duces opportunities for shared experiences. This is a
problem for existing TMS theory because shared expe-
riences are a key precursor of TMS development (Lewis
et al. 2005, Moreland and Levine 2000). Moreland and
Argote (2003) suggest that, in the absence of shared
experiences among all organizational members, “infor-
mation of the type that shared experience provides”
(p. 139) should be circulated, including publicity about
what other workers do. Our research extends this the-
orizing about antecedents to TMS when shared experi-
ences are not possible. We suggest that semistructures
that clarify expectations may help to provide a basis for
TMS development in the absence of shared experiences.
Dialogic practices, for instance, reveal tacit assumptions
that help each member to understand how others inter-
pret incoming information. We argue that by helping to
understand others’ interpretations, dialogic practices not
only complement TMS, as argued by Faraj and Xiao
(2006), but also foster it because they make clear who
knows what.
Applying TMS to ego-centered networks also ques-
tions the assumption commonly made in TMS research
that “who knows what” translates to “who acts on what”
(Brandon and Hollingshead 2004). In interorganizational
contexts, people may know something, but not have
the ownership rights and responsibilities to act on that
knowledge. Alternatively, they may know something that
leads them to take strategic actions that confuse competi-
tors (or the enemy), or to incite reactions from others to
see what they know (Brown and Eisenhardt 1997). Thus,
the assumption of a tightly linked relationship between
knowledge and action has meant that TMS could not
be applied to mixed-motive, interorganizational relation-
ships. Our research suggests that this assumption may
not be needed if semistructures are in place for assessing
the degree to which actions and knowledge converge. An
extension is needed to TMS theory, then, that explicitly
includes the degree of congruence between knowledge
and action.
Role of Trust in Mixed-Motive Knowledge
Collaborations
Trust has long been considered fundamental for coop-
eration in risky situations (Mayer et al. 1995) includ-
ing interorganizational settings (Zaheer et al. 1998).
Trust is multifaceted (Mayer et al. 1995), including
dimensions such as competence and benevolence. TMS
incorporates competence-based trust via the subindica-
tor of credibility (beliefs about the reliability of mem-
bers’ expertise). Levin and Cross (2004) argue that
benevolence-based trust is necessary in all knowledge
collaborations because concerns over harm would make
one reluctant to learn from the knowledge source. Our
results challenge their argument and suggest that profes-
sionals can still learn from others as long as they are
aware of the lack of benevolence and manage accord-
ingly. Our findings show that professionals’ combinative
capabilities are improved with the level of benevolence-
based distrust in the network. Although some of the trust
literature portrays distrust as the inverse of trust, convey-
ing an avoidable negative evaluation of a social relation-
ship (Kramer 1999b), other literature portrays distrust
as a necessary dimension of any mixed-motive situa-
tion, conveying not a negative evaluation of the relation-
ship, but rather a level of certainty about others’ actions
(Lewicki et al. 1998). With a greater certainty level,
professionals can act to protect themselves while suc-
cessfully collaborating with others. Future work needs
to validate our findings on benevolence-based distrust
and examine the dynamics that lead to it in ego-centered
networks.
Conclusion
Our research tentatively suggests that a paradigm shift
is needed to develop an information environment for
national security collaborations. The paradigm used
today is centered on policy and process design with
numerous audit functions to ensure congruence with
those policies and processes (GAO 2006). The paradigm
recognizes the need for localization of such policies to
the formal organizational level, but fails to recognize
the critical role of the individual. Our findings suggest
that an alternative to this hierarchical and rigid paradigm
may be the use of semistructures that serve as organiz-
ing principles in informal peer-based structures. Such
semistructures form a critical role in developing the nec-
essary TMS, which in turn facilitates security collabo-
ration. Because security collaborations increasingly rely
on a professional’s personal network, which lies outside
the control of the organization, formal policies and pro-
cesses, even if localized, are an unrealistic prescription.
Acknowledgments
The authors are grateful to security professionals who gave
their time for the conduct of this study. Clearly, they are not
responsible for the contents of this manuscript. The authors
also appreciate the insightful and constructive comments of
the Organization Science Editor-in-Chief and the reviewers.
Encouraging and useful feedback was also received from the
discussant and the attendees of the International Conference
of Information Systems (ICIS) in Las Vegas, NV, Decem-
ber 12, 2005.
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
Organization Science 19(2), pp. 260–276, © 2008 INFORMS 273
Appendix. Survey Items
Label Item
Combinative capabilities: How frequently have you received the following types of knowledge from the other organizations
involved [in your network]? (1 = almost never to 7 = nearly always)
CAPEA Know-how about how a threat was identified
CAPEB Know-how about steps taken to respond to a threat
CAPEC Know-how about how to prevent future similar threats
CAPED Reasons behind decisions others made in responding to the security threat
CAPEE Reasons behind involving certain people in the security response
CAPEF Reasons behind decisions made for not pursuing certain security responses
TMS development: Based on your personal interactions with the members of this network, (1 = strongly disagree to
7 = strongly agree)
TMA Each member has highly specialized knowledge of some aspect of security
TMB I am comfortable accepting security-related suggestions from the other members
TMC I trust that other members’ knowledge about security is credible
TMD I am confident relying on the information that other members bring to a discussion
TME Members in this network know each other and work together in a well-coordinated fashion
TMF Members respond to security problems smoothly and efficiently
Dialogic practices: When you think of discussions you have had with others in your network, how frequently do the
following happen? (1 = never to 7 = daily)
DIALA Develop several options for interpreting information or responding to a threat
DIALB Describe problems at both the summary level as well as the detailed level
DIALC Discuss alternative scenarios for a problem
DIALD Brainstorm about ideas or possible solutions
DIALE Describe detailed context of threat information
DIALF Understand how information changes over time
DIALG Discuss sources of ideas for handling threat
DIALH Discuss how time is affecting information
DIALI Revisit decisions or interpretations about security issues made earlier
DIALJ Discuss source of threat information
Adequacy of knowledge dissemination protocols: How adequate are the following administrative procedures used in your
network for meeting your security needs? (1 = completely inadequate to 7 = completely adequate)
DISSEMA Norms and procedures for informing others about security threat information
DISSEMB Procedures for identifying what information is sensitive
DISSEMC Safeguards to protect the privacy of the source
Clarity of knowledge ownership: Regarding your network: (1 = completely to 7 = strongly important)
OWNA It is often unclear who owns the knowledge that is shared among members (reverse)
OWNB There is a lot of ambiguity about who owns the solutions created among members (reverse)
OWNC The policy is clear about who owns what rights to knowledge, inventions, or discoveries.
Org learning intent: When participating in your community of security professionals, to what degree are these objectives
especially important to your employer? (1 = completely unimportant to 7 = strongly important)
LEARNA Learn about new technology
LEARNB Learn about new management techniques
LEARNC Learn about new ways to prevent security problems
LEARND Learn about new ways to respond to security threats
LEARNE Access to others’ skills and knowledge
Benevolence-based distrust. How frequently do you worry about people in other organizations: (1 = almost never to
7 = all the time)
DISTRUSTA Acquiring too much knowledge
DISTRUSTB Not sharing necessary knowledge with you
DISTRUSTC Aggressively protecting some of their knowledge from you
DISTRUSTD Probing you for valuable knowledge lying outside the scope of your agreement with them
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
274 Organization Science 19(2), pp. 260–276, © 2008 INFORMS
Appendix. (cont’d.)
Label Item
Multiple channels of communication: # of different communication media used at least once a month: (1) Meet one-on-one
or small group face-to-face meetings; (2) Call people on the phone; (3) Attend local chapter meetings, conferences’
(4) Use e-mail one-on-one; (5) Use Group e-mail lists or list serves; (6) Log in to a portal
Control: Task interdependence. My security responsibilities in my organization: (1 = strongly disagree to 7 = strongly agree)
DEPENDA Requires me to talk with staff from other organizations
DEPENDB Often involves me sharing information with staff at other organizations
DEPENDC Often involves using information and solutions from other organizations
DEPENDD Creates results that are dependent on the efforts of others from other organizations
Control: Network size. Approximately how many people are in your network?
Control: Years in network. For how many years have you been a member of this network?
References
Akgun, A. E., J. C. Byrne, H. Keskin, G. S. Lynn. 2006. Transactive
memory system in new product development teams. IEEE Trans.
Engrg. Management 53(1) 95–111.
Anand, V. C., C. Manz, W. H. Glick. 1998. An organizational memory
approach to information management. Acad. Management Rev.
23(4) 796–809.
Argote, L. 1999. Organizational Learning: Creating, Retaining, and
Transferring Knowledge. Kluger, Boston, MA.
Barzel, Y. 1989. Economic Analysis of Property Rights. Cambridge
Press, Cambridge, UK.
Bernard, H. R., E. C. Johnsen, P. D. Killworth, C. McCarty, G. A.
Shelley, S. Robinson. 1990. Comparing four different meth-
ods for measuring personal social networks. Soc. Networks 12
179–215.
Blau, P. M. 1964. Exchange and Power in Social Life. Wiley, New
York.
Boland, R. J., R. V. Tenkasi, D. Te’eni. 1994. Designing information
technology to support distributed cognition. Organ. Sci. 5(3)
456–475.
Boland, R. J. Jr., J. Singh, P. Salipante, J. D. Aram, S. Y. Fay,
P. Kanawattanachai. 2001. Knowledge representations and
knowledge transfer. Acad. Management J. 44(2) 393–417.
Borgatti, S. P., R. Cross. 2003. A relational view of information seek-
ing and learning in social networks. Management Sci. 49(4)
432–445.
Brandon, D. P., A. B. Hollingshead. 2004. Transactive memory sys-
tems in organizations: Matching tasks, expertise, and people.
Organ. Sci. 15(6) 633–644.
Brown, S. L., K. M. Eisenhardt. 1997. The art of continuous change:
Linking complexity theory and time-paced evolution in relent-
lessly shifting organizations. Admin Sci. Quart. 42(1) 1–34.
Burt, R. S. 1980. Innovation as a structural interest: Rethinking the
impact of network position on innovation adoption. Soc. Net-
works 2(4) 327–355.
Burt, R. S. 1984. Network items and the general social survey. Soc.
Networks 6293–339.
Burt, R. S. 1992. Structural Holes. The Social Structure of Competi-
tion. Harvard University Press, Cambridge, MA.
Campbell, K. E., B. A. Lee. 1991. Name generators in surveys of
personal networks. Soc. Networks 13(4) 203–221.
Carley, K. 1991. A theory of group stability. Amer. Sociol. Rev. 56
331–354.
Carlile, P. R., E. S. Rebentisch. 2003. Into the black box: The knowl-
edge transformation cycle. Management Sci. 49(9) 1180–1195.
Chin, W. W. 1998. The partial least squares approach for structural
equation modeling. G. A. Marcoulides, ed. Modern Methods for
Business Research. Lawrence Erlbaum Associates, Mahwah, NJ,
295–336.
Cross, R., J. N. Cummings. 2004. Tie and network correlates of indi-
vidual performance in knowledge-intensive work. Acad. Man-
agement J. 47(6) 928–937.
Cross, R., L. Sproull. 2004. More than an answer: Information rela-
tionships for actionable knowledge. Organ. Sci. 15(4) 446–462.
Cummings, J. N. 2004. Work groups, structural diversity, and knowl-
edge sharing in global organization. Management Sci. 50(3)
352–364.
Dougherty, D. 1992. Interpretive barriers to successful product inno-
vation in large firms. Organ. Sci. 3(2) 170–202.
Dougherty, D. 2007. Contingent organizing for games of innovation:
Diverse configurations of core principles for innovative organiza-
tion designs. Internat. J. Innovation Management 11(1) 115–138.
Engwall, M., C. Svensson. 2001. Cheetah teams. Harvard Bus. Rev.
79(2) 20–21.
Faraj, S., L. Sproull. 2000. Coordinating expertise in software devel-
opment teams. Management Sci. 46(12) 1554–1568.
Faraj, S., Y. Xiao. 2006. Coordination in fast response organizations.
Management Sci. 52(8) 1155–1169.
Fein, S. 1996. Effects of suspicion on attributional thinking and
the correspondence bias. J. Personality Soc. Psych. 70(6)
1164–1184.
Fu, Y.-C. 2005. Measuring personal networks with daily contacts:
A single-item survey question and the contact diary. Soc. Net-
works 27 169–186.
Gal-Or, E., A. Ghose. 2005. The economic incentives for sharing
security information. Inform. Systems Res. 16(1) 186–205.
GAO 2006. U.S. Government Accountability Office, Homeland Secu-
rity and Information Sharing, GAO-06-385, U.S. Government
Accountability Office.
Goodhue, D. L., R. L. Thompson. 1995. Task-technology fit and indi-
vidual performance. MIS Quart. 19(2) 213–236.
Goodman, P. S., J. M. Wilson. 2000. Substitutes for socialization
in exocentric teams. M. Neale, B. Mannix, T. Griffith, eds.
Research in Groups and Teams, Vol. 3. JAI Press, Greenwich,
CT, 53–77.
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
Organization Science 19(2), pp. 260–276, © 2008 INFORMS 275
Grand, S., G. von Krogh, D. Leonard, W. Swap. 2004. Resource allo-
cation beyond firm boundaries: A multi-level model for open
source innovation. Long Range Planning 37 591–610.
Gray, P. H., D. B. Meister. 2004. Knowledge sourcing effectiveness.
Management Sci. 50(6) 821–834.
Hagel, J., J. S. Brown. 2005. The Only Sustainable Edge. Harvard
Business School Publishing, Boston, MA.
Hannah, D. R. 2005. Who owns ideas? The effects of trade secret
protection procedures on employees’ obligations. Organ. Sci.
16(1) 71–84.
Heiman, B. A., J. A. Nickerson. 2004. Empirical evidence regarding
the tension between knowledge sharing and knowledge expro-
priation in collaborations. Managerial Decision Econom. 25
401–420.
Hill, R. A., R. I. M. Dunbar. 2003. Social network size in humans.
Human Nature 14(1) 53–72.
Hollingshead, A. B. 2001. Cognitive interdependence and convergent
expectations in transactive memory. J. Personality Soc. Psych.
81 1080–1089.
Jarvenpaa, S. L., S. Staples. 2001. Exploring perceptions of organiza-
tional ownership of information and expertise. J. Management
Inform. Systems 18(1) 150–183.
Jarvenpaa, S. L., H. Tanriverdi. 2006. Knowledge ownership and ter-
ritoriality: A conceptualization and scenario-based experimental
investigation. Best Paper Proc. Acad. Management Conf.
Kalman, M. E., P. Monge, J. Fulk, R. Heino. 2002. Motivations to
resolve communication dilemmas in database-mediated collabo-
ration. Comm. Res. 29(2) 125–155.
Killworth, P. D., E. Johnson, H. R. Bernard, G. A. Shelley,
C. McCarty. 1990. Estimating the size of personal networks.
Soc. Networks 12 289–312.
Kogut, B., U. Zander. 1992. Knowledge of the firm, combinative
capabilities, and the replication of technology. Organ. Sci. 3(3)
383–397.
Korsgaard, M. A., S. E. Brodt, E. M. Whitener. 2002. Trust in the
face of conflict: The role of managerial trustworthy behavior and
organizational context. J. Appl. Psych. 87(2) 312–319.
Kramer, R. 1999a. Social uncertainty and collective paranoia in
knowledge communities. L. L. Thompson, J. M. Levine, D. M.
Messick, eds. Shared Cognition in Organizations. Erlbaum,
Mahwah, NJ, 163–194.
Kramer, R. 1999b. Trust and distrust in organizations: Emerging per-
spectives, enduring questions. Annual Rev. Psych. 50 569–598.
Laumann, E. 1966. Prestige and Association in an Urban Community.
Bobbs-Merrill, New York.
Levin, D. Z., R. Cross. 2004. The strength of weak ties you can
trust: The mediating role of trust in effective knowledge transfer.
Management Sci. 50(11) 1477–1490.
Lewicki, R. J., D. J. McAllister, R. J. Bies. 1998. Trust and distrust:
New relationships and realities. Acad. Management Rev. 23(3)
438–458.
Lewis, K. 2003. Measuring transactive memory systems in the field:
Scale development and validation. J. Appl. Psych. 88(4) 587–604.
Lewis, K., D. Lange, L. Gallis. 2005. Transactive memory systems,
learning, and learning transfer. Organ. Sci. 16(6) 581–598.
Liang, D., R. Moreland, L. Argote. 1995. Group versus individual
training and group performance: The mediating role of transac-
tive memory. Personality Soc. Psych. Bull. 21(4) 384–393.
Lindell, M. K., D. J. Whitney. 2001. Accounting for common method
variance in cross-sectional research designs. J. Appl. Psych.
86(1) 114–121.
Macdonald, S., R. Piekkari. 2005. Out of control: Personal networks
in European collaboration. R&D Management 35(4) 441–453.
Majchrzak, A., S. L. Jarvenpaa. 2005. Information security in cross-
enterprise collaborative knowledge work. Emergence 6(4) 4–8.
Majchrzak, A., S. L. Jarvenpaa, A. B. Hollingshead. 2007. Coordinat-
ing expertise among emergent groups responding to disasters.
Organ. Sci. 18(1) 147–161.
Majchrzak, A., A. Malhotra, R. John. 2005. Perceived individ-
ual collaboration know-how development through information
technology-enabled contextualization: Evidence from distributed
teams. Inform. Systems Res. 16(1) 9–27.
Marin, A., K. N. Hampton. 2007. Simplying the personal network
name generator: Alternative to traditional multiple and single
name generators. Field Methods 19(2) 163–193.
Marsden, P. V. 1990. Network data and measurement. Ann. Rev. Soc.
16 435–463.
Massey, A. P., M. M. Montoya-Weiss. 2006. Unraveling the temporal
fabric of knowledge conversion: A model of media selection and
use. MIS Quart. 30(1) 94–114.
Mayer, R. C., J. H. Davis, F. D. Schoorman. 1995. An integrative
model of organizational trust. Acad. Management Rev. 20(3)
709–734.
McCarty, C., P. D. Killworth, H. R. Bernard, E. C. Johnsen,
G. A. Shelley. 2001. Comparing two methods for estimating
network size. Human Organ. 60(1) 28–39.
McEvily, B., V. Peronne, A. Zaheer. 2003. Trust as an organizing
principle. Organ. Sci. 14(1) 91–103.
Moreland, R. L., L. Argote. 2003. Transactive memory in dynamic
organizations. R. Peterson, E. Mannix, eds. Understanding
the Dynamic Organization. Lawrence Erlbaum Associates,
Mahwah, NJ, 135–162.
Moreland, R. L., J. M. Levine. 2000. Socialization in organizations
and work groups. M. Turner, ed. Groups at Work: Theory and
Research. Lawrence Erlbaum Associates, Mahwah, NJ, 69–112.
Norman, P. M. 2002. Protecting knowledge in strategic alliances
resource and relational characteristics. J. High Tech. Manage-
ment Res. 13 177–202.
Okhuysen, G. A., K. M. Eisenhardt. 2002. Integrating knowledge in
groups: How formal interventions enable flexibility. Organ. Sci.
13(4) 370–386.
Pavlou, P., O. El Sawy. 2006. The case of new product development.
Inform. Systems Res. 17(3) 198–227.
Podsakoff, P. M., D. W. Organ. 1986. Self-reports in organizational
research: Problems and prospects. J. Management 12 69–82.
Podsakoff, P. M., S. B. MacKenzie, J.-Y. Lee, N. P. Podsakoff. 2003.
Common method biases in behavioral research: A critical rev. of
the literature and recommended remedies. J. Appl. Psych. 88(5)
879–903.
Reagans, R., B. McEvily. 2003. Network structure and knowledge
transfer: The effects of cohesion and range. Admin. Sci. Quart.
48(3) 240–267.
Ren, Y., K. M. Carley, L. Argote. 2006. The contingent effects of
transactive memory: When is it more beneficial to know what
others know? Management Sci. 52(5) 671–682.
Scott, S. V., G. Walsham. 2005. Reconceptualizing and managing rep-
utation risk in the knowledge economy: Toward reputable action.
Organ. Sci. 16(3) 308–322.
Jarvenpaa and Majchrzak: Knowledge Collaboration Among Professionals Protecting National Security
276 Organization Science 19(2), pp. 260–276, © 2008 INFORMS
Swart, J., N. Kinnie. 2003. Sharing knowledge in knowledge-intensive
firms. Human Resources Management J. 13(2) 60–75.
Szulanski, G. 1996. Explaining internal stickiness: Impediments to the
transfer of best practices with the firm. Strategic Management J.
17 27–43.
Te’eni, D. 2001. Review: A cognitive-affective model of organi-
zational communication for designing IT. MIS Quart. 25(2)
251–312.
Tsai, W. 2001. Knowledge transfer in intraorganizational networks:
Effects of network position and absorptive capacity on business
unit innovation and performance. Acad. Management Rev. 44(5)
996–1004.
Ulmer, R. R., T. C. Sellnow. 2000. Ambiguity in organizational crisis
communication: Jack in the box as a case study. J. Bus. Ethics
25 143–155.
von Hippel, E. 1988. The Sources of Innovation. MIT Press,
Cambridge, MA.
Wegner, D. M. 1986. Transactive memory: A contemporary analysis
of the group mind. G. Mullen, G. Goethals, eds. Theories of
Group Behavior. Springer-Verlag, New York, 185–208.
Wegner, D. M., R. Erber, P. Raymond. 1991. Transactive memory in
close relationships. J. Personality Soc. Psych. 61 923–929.
Weick, K. E., K. H. Roberts. 1993. Collective mind in organiza-
tions: Heedful interrelating on flight decks. Admin. Sci. Quart.
38 357–381.
Wellman, B. 1999. Networks in the Global Village. Westview,
Boulder, CO.
Zaheer, A., B. McEvily, V. Perone. 1998. Does trust matter? Explor-
ing the effects of interorganizational and interpersonal trust on
performance. Acad. Management Rev. 9(2) 141–159.
