Searching for Answers: Networks of Practice among Public Administrators

The American Review of Public Administration
XX(X) 1 –24
© The Author(s) 2011
Reprints and permission: http://www.
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0275074011398956
http://arp.sagepub.com
Searching for Answers:
Networks of Practice
Among Public
Administrators
Maria Christina Binz-Scharf1,
David Lazer2, and Ines Mergel3
Abstract
How do public administrators find information about the problems they confront at work?
In particular, how and when do they reach across organizational boundaries to find answers?
There are substantial potential obstacles to such searches for answers, especially in a system of
decentralized governance such as the U.S. government. In this article, we examine the alternative
mechanisms within the public sector that compensate for this dispersion of expertise, focusing
on knowledge sharing across public DNA forensics laboratories. In particular, we propose that
the emergence of informal interpersonal networks plays an important role in providing access
to necessary expertise within a highly decentralized system. Our findings point both to the need
for further research on knowledge sharing networks within the public sector as well as practical
implications around the value of investments into facilitating the creation and maintenance of
networks of practice.
Keywords
cross-jurisdictional knowledge sharing, diffusion, innovation, network of practice, case study
Introduction
How does a lesson learned in one agency, one state government, one locale, provide insights that
others might borrow? Although innovation has been articulated as one of the major advantages
of a decentralized system, the very dispersion of government in a federal system creates enormous
obstacles to lesson sharing. We propose that in some domains there exist emergent networks of
practice, where administrators connect to each other in a search for reliable answers to questions
that arise on their job, and whose structure transcends organizational boundaries, prescribed
reporting structures, and jurisdictions.
1City College of New York, New York City
2Northeastern University, Boston MA and Harvard University, Cambridge MA
3Syracuse University, Syracuse, NY
Corresponding Author:
Maria Christina Binz-Scharf, City College of New York, 160 Convent Ave, New York 10031
Email: mbinzscharf@ccny.cuny.edu

2 The American Review of Public Administration XX(X)
This insight builds on a recent surge of research on the role that networks play in the capacity
of public organizations (Agranoff, 2006; McGuire, 2002; Milward & Provan, 2000; O’Toole &
Meier, 2004; Provan, Huang, & Milward, 2009; Weber & Khademian, 2008). Little of this research,
however, examines knowledge interdependences in informal public sector networks. This is a
striking contrast to the literature on private organizations, which has examined knowledge sharing
within and across firms (Podolny & Page, 1998; Powell, Koput, & Smith-Doerr, 1996; Uzzi,
1997). This article contributes to the existing literature on public sector networks by examining
whether and how informal interpersonal networks emerge as a result of local needs for knowledge,
and how public managers use their informal networks to overcome complex problems.
In an in-depth interpretive case study, we examine a particular population of geographically
dispersed government organizations—government crime laboratories involved in DNA analysis.
The use of DNA in the criminal justice system has grown exponentially in the last decade, and
this growth, combined with the rapidly changing technology, has created particular stresses on
decision makers within the approximately 180 government labs scattered across the country, many
with just a handful of DNA analysts. This domain thus offers an “easy case” to detect the possible
presence of networks of practice—that is, relationships formed around a set of focal practices.
We find that there is indeed a robust network of practice in this domain. However, this network
exists symbiotically with the hierarchical structures in the system, where many of the control
mechanisms of the system, such as inspections and FBI-sponsored training, catalyze peer-to-peer
knowledge sharing. Furthermore, although we find that this network is critical to the effectiveness
of particular labs, some of the emergent features of the network are traditionally associated with
hierarchies, including the centralization around particular knowledge hubs, and the development
of silos based on state boundaries.
Below, we first develop a synthesis of the varied literatures on networks in public administra-
tion and the knowledge-based view of organizations. We then discuss the particular domain of our
research—forensic DNA labs in the United States, outlining our methodology to study how public
managers actively seek knowledge in a decentralized system of government. We then provide our
findings, concluding with a discussion about the implications both for scholarship and practice.
Public Sector Networks and
Cross-Jurisdictional Knowledge Diffusion
The role of interorganizational public networks has received significant attention in recent years,
and the idea that there are networks that support interjurisdictional diffusion dates back at least
to Walker (1969). Below, we discuss these distinct literatures, and then turn to an emerging litera-
ture on the management of knowledge in public sector organizations. Our aim is to bridge these
distinctive research traditions, and extend extant literature to include the role of informal knowl-
edge sharing networks in the public sector.
Networks in the Public Sector
Although there is a growing body of literature on public sector networks (e.g., Agranoff, 2006;
Agranoff & McGuire, 2001; Bardach, 1999, 2001; McGuire, 2002; Meier & O’Toole, 2001; Milward
& Provan, 2000; O’Toole & Meier, 2004; Provan et al., 2009; Weber & Khademian, 2008), this
literature mainly focuses on process interdependence, or analyzing coordination networks among
various agencies. For example, how do agencies need to coordinate effectively to deliver public
services within a formalized, for example, contractual, network? Networks in the public sector are
generally regarded as a governance mechanism that serves to solve so-called wicked problems, or
problems of the hollow state (Rittel & Webber, 1973). Milward and Provan define the hollow state

Binz-Scharf et al. 3
as a “metaphor for the increasing list of third parties, often non-profits, to deliver social services
and generally act in the name of the state” (Milward & Provan, 2000, p. 359). Most of the empiri-
cal research in this domain has examined the joint production of public services (such as contract-
ing, or principal-agent relationships) within collaborative networks (e.g., Agranoff & McGuire,
2001; Mandell, 2001; Nelson, Bloomfield, Hales, & Libby, 2001; O’Toole, 1997).
In short, the existing public administration literature has focused mainly on collaborative prob-
lem solving in intergovernmental, intersectoral (Agranoff, 1996; Ansell & Gash, 2008), and formally
established ties (cf. Provan, Isett, & Milward, 2004) to ensure network effectiveness. In contrast
to the robust thread of research on formal networks that has emerged over the last decade, there
is relatively little research on how informal networks support task execution across agencies in
the public sector. In fact, little analytic attention has been paid to the distinction between formal
and informal networks in the existing public administration literature (Isett, Mergel, LeRoux,
Mischen, & Rethemeyer, 2011). We define formal network ties as those relationships that are
established through hierarchical positions, reporting structures, and official duties and competences
within and across organizations (Blau, 1963). In short, formal ties are role prescribed. For example,
formal ties exist between a supervisor and her subordinate, or between two government agencies
with a memorandum of understanding to temporarily collaborate on a project.
We define informal ties as interpersonal relationships that are extra role (Mergel, Lazer, &
Binz-Scharf, 2008). These relationships can cut across traditional organizational structures (such
as functions and divisions) and therefore generally are not included in formal reporting procedures
(e.g., Krackhardt & Hanson, 1993), which makes them more elusive and difficult to document.
Thus, for example, one might distinguish formal pathways to mobilize support in an organization
(call IT support for assistance with a computer) from informal ones (ask the individual in the
neighboring office if they know how to solve the problem). The management literature is replete
with examples of the essential role informal networks play within and across organizations, for
example, as a vehicle to speed up knowledge exchange, increase the efficiency of work, and foster
accidental communication about work-related topics, which adds to and interplays with the formal
communication and reporting structure (Cummings, 2004; Hansen, 1999, 2002; Knoke & Burt,
1983; Krackhardt & Stern, 1988; Tichy & Fombrun, 1979).
Cross-Jurisdictional Diffusion of Knowledge
There is a significant literature on interstate diffusion of knowledge, following from Walker (1969).
However, most of this research addresses the legislative level (e.g., Soss, Schram, Vartanian, &
O’Brien, 2001; Volden, 2006). Our study focuses on the microscopic decisions that happen at the
street level (or, more appropriately given our case, the bench level). How are the individual deci-
sions made by public employees directly dependent on the information and insights they get from
others outside of their immediate organization?
An important research stream that informs our study is the literature on the diffusion of innova-
tion, some of which looks at diffusion among public sector organizations (cf. Rogers, 1995). There
is a direct linkage between our argument with respect to “searching” for answers and the diffusion
literature, because the spread of information that is novel to the recipient might be viewed as inno-
vative from the recipient’s point of view. In our study, we look at knowledge diffusion from a dif-
ferent angle and focus on the individual knowledge-sourcing strategies (Gray & Meister, 2004). In
other words, we describe innovation as “search, when you don’t know what you are looking for”
(Girard & Stark, 2007). There is an active component to diffusion when applied to the decisions of
public administrators: Innovative knowledge is not passively transported through the social system,
but actively sought for based on individual choices. This qualitative viewpoint of the procedural
microlevel decision includes the individual’s choice of source and the reason for the selection of

4 The American Review of Public Administration XX(X)
the chosen specific source. Researchers, such as Nebus (2006), have suggested the use of individual
choice models in theories of social structure (Coleman & Fararo, 1992; Lindenberg, 1985, 1990).
Much of this research is grounded in the organizational behavior literature, to which we now turn.
Networks of Practice
In the knowledge-based view of the organization (Grant, 1996), knowledge is viewed as a scarce
resource, and its creation and use is what determines the competitive advantage of an organization.1
To be useful to someone other than its creator(s), knowledge needs to be shared. The sharing of
knowledge is relatively easy when the knowledge to be shared is explicit, that is, verbalized, writ-
ten, drawn, or otherwise articulated (Nonaka, 1994). A classical example in bureaucracies is given
by “the files” (Weber, 1921/1968), today joined by large databases. Tacit (or implicit) knowledge,
however, is more difficult to share: Polanyi (1966) asserted that individuals know more than they
can explain; that is, individuals have knowledge that is nonverbalized, intuitive, and unarticulated.
Tacit knowledge has been defined as hard to communicate, deeply rooted in action, involvement,
and commitment within a specific context (Polanyi, 1962); “a continuous activity of knowing”
(Nonaka, 1994, p. 16); or, more practically, as “the way things are done around here” (Spender,
1996). Furthermore, as tacit knowledge is not articulated (and thus likely resides in an individual’s
brain), its transfer requires the interaction between individuals, whereas explicit knowledge can
be retrieved from nonhuman sources such as reference databases (Schreiber & Carley, 2003).
Similarly, although computer-mediated communication has been found to facilitate knowledge
transfer (e.g., Constant, Sproull, & Kiesler, 1996; Kiesler & Sproull, 1992; Sproull & Kiesler,
1986), technology cannot substitute for face to face interaction (e.g., Hinds & Bailey, 2003).
Not all knowledge required to pursue an organization’s goals is readily available within the
organizational boundaries (Anand, Glick, & Manz, 2002; Grant, 1996; Spender & Grant, 1996).
Therefore, members of an organization often rely on knowledge from external third parties (Anand
et al., 2002). Research shows that professionals rely on their communities of peers when it comes
to specialized knowledge (Cross, Parker, Prusak, & Borgatti, 2001; Orlikowski, 2002), and among
those, they turn to those individuals they trust, are friends with or whom they respect (Ibarra &
Andrews, 1993; Krackhardt & Kilduff, 1999). Within the organizational literature, the idea that
there exist knowledge-based “communities of practice” has gained substantial currency. Com-
munities of practice are bounded groupings of professionals who engage in similar practices and
have frequent occasions to interact with each other (Lave & Wenger, 1991; Wenger, 1998). They
learn from one another through a process termed legitimate peripheral participation (Lave &
Wenger, 1991), where, much like in an apprenticeship, individuals who are new to the practice
only engage in it peripherally before acquiring in-depth knowledge from more experienced indi-
viduals, which over time permits them to fully adopt the practice.2
A related but broader concept is that of networks of practice, which can be defined as a set of
loosely coupled relationships between individuals who are not necessarily collocated but engage
in practices that share a certain degree of similarity (Vaast & Walsham, 2009). Although members
of a network of practice may never have met face to face, or not even be aware of each other, they
share common practices and tend to have an interest in similar issues (Brown & Duguid, 2001;
Vaast & Walsham, 2009). A network of practice is generally more open than a community of
practice and the relationships among its members are looser—they share an overall knowledge
domain, but the redundancies in what they know and do are not as strong (Vaast & Walsham, 2009).
We refer to the system that is the subject of our study as a network of practice because it fits the
above description of such a network well: DNA forensic scientists working in government crime
labs scattered across the country are mostly aware of each other, but they may never have met face
to face. They follow the same federal rules and regulations, but develop practices that are molded
by their local context and may therefore be partly similar or different (Vaast & Walsham, 2009).

Binz-Scharf et al. 5
However, regardless of their organizational affiliation, they share a significant degree of similarity
in their practices and could benefit from communicating with each other and exchanging ideas
about their practices (Duguid, 2005).
It is important to notice that most of the research on communities and networks of practice has
been conducted in private sector settings (a notable exception is Eglene et al.’s (2007) study on
public sector knowledge networks). Knowledge sharing in the public sector meets some specific
challenges, such as tighter budget restrictions than in the private sector; data privacy and confi-
dentiality, and related security issues, which, due to mandatory regulations, are a top priority for
government; as well as traditionally rigid hierarchical structures and organizational boundaries
that stand in stark contrast to the informal structure of knowledge networks (Bardach, 1999;
Binz-Scharf, 2008; Fountain, 2001). However, “networks and bureaucracy coexist and interact”
in this setting (Eglene et al., 2007, p. 92, citing O’Toole, 1997): Innovation happens through infor-
mal, mostly horizontal, collaboration, whereas formal, vertical, authority remains important in the
bureaucratic government environment (Eglene et al., 2007). Moreover, recent collaborative gov-
ernance literature suggests that as mandates broaden in scope, public sector actors need to reach
across organizational boundaries and include the knowledge and resources of third parties into
their own practices (O’Leary & Bingham, 2009).
Research Design
This article examines knowledge sharing among individuals in multiple public organizations that
are all part of a highly controlled system. These individuals are specialized professionals who each
operate within the reporting structure of the organization they are affiliated with. However, at the
same time, they need to follow the procedural rules and regulations issued by the controlling federal
agency, which sometimes supersede local procedures. The reason for this configuration is a con-
stitutionally mandated dispersion of authority to state governments, and within those, a pressure
to create geographic equity by further dispersing authority to local governments. This geographic
dispersion is not related to a clear separation in the types of issues encountered and the knowledge
required to deal with them. Therefore, the tasks executed and the problems encountered by these
professionals are mostly similar across different organizations. The solutions to problems, however,
vary significantly, as they are the fruit of individual interpretations, based on various constellations
of institutional resources, culture, and other countless characteristics. Our key objective was to
understand how knowledge is actively sought by individuals within this decentralized system.
Specifically, our study posed the following research questions:
Research Question 1: What search strategies do individuals adopt when looking for
information? That is, where and how do people look for answers?
Research Question 2: What factors determine these individual search strategies?
We addressed these questions with an interpretive case study, aiming at generating theory that
emerges from the knowledge-sharing activities occurring in the empirical setting under inquiry
(Emerson, Fretz, & Shaw, 1995). We followed Miles and Huberman’s (1994) conceptual approach
to the design, collection, and analysis of qualitative data. In this approach, the researcher starts out
by distilling ideas into initial categories, which are used to create a conceptual framework. This
framework then serves as a basis for data collection and analysis. In our study, the initial focus
was on the need for public administrators to reach outside of their organization to acquire the
knowledge necessary to accomplish essential tasks.
This focus was developed on the basis of the second author’s intimate knowledge of the DNA
forensics community, the first author’s preliminary observations of and interactions with this com-
munity, as well as extant organizational and networks literature. These initial observations suggested

6 The American Review of Public Administration XX(X)
that there existed an epistemic network of knowledge sharing, and that while the central authority
in the system (the FBI) did indeed play a pivotal role, that it did not mediate most knowledge
transfer among laboratories. Even brief exposure to the system indicated, unsurprisingly, that it
has a distinctive and varied cast of characters that have settled into distinctive roles in the system;
and that laboratories and regions have converged on sometimes different norms and practices.
The categories that emerged were centered around factors that influence an individual’s choice
among various knowledge sources. Figure 1 presents our initial conceptual framework of search
strategies and their determinants. According to this framework, an individual’s search behavior
is contextualized by “exogenous” factors at the community level, such as the perception of the
community’s boundaries, and shared norms within the community. At the organization level, the
size, jurisdiction, and location of the organization (and, related to the latter, the proximity to other
organizations doing similar work) might have an influence on an individual’s search strategies,
as does the lab’s organizational culture. For example, strictly observed reporting structures might
discourage an individual from “shopping” for an answer instead of going to her supervisor. How-
ever, individual characteristics play into such a decision as well: Personality, expertise, seniority,
and reputational concerns, among others, likely determine one’s strategies for searching for
answers. Finally, the choice of sources, as well as the sequence in which one might approach
these, can be expected to differ according to the type of knowledge sought (e.g., advice, opinion,
confirmation, and whether it is tacit or explicit) and the source characteristics, such as access to
and availability of a certain source.
Nevertheless, the framework, although providing some initial structure, does not control the
analysis. It is prone to modification as the data collection and analysis unfold, and its final structure
might differ significantly from what it looked like at the outset. Therefore, the analysis is data
driven, aiming at understanding the viewpoints of the individuals who participate in the study.
It is important to note at this point that the theoretical concepts we present here emerged from our
study, and are thus a result of our data interpretation, rather than a starting point for data collection
Individual factors
- Personality
- Expertise
Reputation-
Source
characteristics
- Availability
- Access
- Expertise
Relational factors
- Trust
- Reciprocity
- History
- Shared norms
- Boundaries
Community factors
Lab factors
- Size, location
- Jurisdiction
- Org. culture
Type of knowledge
sought
-
-
-
Confirmation
Opinion
Advice
- Choice of sources
- Sequence
Search strategies
Figure 1. Conceptual framework

Binz-Scharf et al. 7
and analysis. This is a distinct characteristic of grounded, interpretive research, where the theoretical
concepts and framework are grounded in, and emerge from, the data, instead of being drawn from
prior theory to guide the collection and analysis of data (Suddaby, 2006). In this research paradigm,
it is common to report data and their interpretation before the theoretical concepts are introduced
(Suddaby, 2006, p. 637, citing Dact, 1985). We have chosen, however, to follow a more traditional
presentation strategy and to introduce the theoretical concepts before we turn to data reporting and
interpretation, in the hope that doing so will clarify in advance our argument and theoretical contribu-
tions, while keeping in mind that these concepts actually emerged from the study itself in a process
of going back and forth between data and relevant literature (Strauss, 1987; Suddaby, 2006).
Field Setting
Given that this is an exploratory study aiming at illuminating the presence and dynamics of
knowledge-sharing processes within a dispersed policy community, we selected our setting in
such a way that we could expect with a relatively high degree of certainty to be able to observe
and analyze the phenomenon of interest here. The case we have chosen is that of the professionals
in government crime laboratories involved in the forensic DNA analysis. The work of forensic
scientists is knowledge intensive, specialized, and highly complex, and it is subject to constantly
changing technology (Bieber, 2004). The minimum educational requirement to be a forensic
scientist3 is a bachelor of science degree, most commonly in chemistry or biology. Many DNA
analysts also have a master of science degree (often mandatory for supervisory positions) in
chemistry, biology, or forensic science, and several among them have a PhD in these or related
disciplines. The profession has experienced an enormous rise in popularity, thanks to the televi-
sion series “CSI”, a phenomenon insiders refer to as the “CSI effect” (Willing, 2004).
There are DNA laboratories at the federal, state, and local levels. These labs are generally
nonhierarchically arranged—that is, state labs generally do not have authority over local labs, nor
do federal over state labs. However, there are elements of hierarchy within the system, with respect
to the following: (a) the infrastructure (where, for example, the FBI sets the rules regarding access
to the national database, as discussed below); (b) resources (e.g., the federal government provides
significant resources to state and local laboratories); and (c) federal regulation (which is in part
based on FBI requirements, following from point “a”).
Neither the size of a single lab nor the number of labs in a given state necessarily reflect the size
(or population) of the state they reside in. Rather, the configuration is a result of historical develop-
ment, in part due to state and local jurisdictions that attribute more or less importance to DNA analysis
(which is reflected in the budget given to labs), institutional entrepreneurship (a successful lab is
more likely to expand), and geographical necessity (a large state might feel the functional or political
need for dispersed laboratory capacity).
Federal law enables states to upload certain DNA profiles into a database system called the
Combined DNA Index System (CODIS), managed by the FBI. The database contains the DNA
profiles of qualified offenders (in most states, convicted felons) and profiles from crime scenes.
The objective of the database is to link known individuals to crimes as well as crimes to each other.4
The state component of the database is known as the State DNA Index System (SDIS), where a
subset of these profiles is also uploaded into the National DNA Index System (NDIS).5 The FBI
distinguishes between DNA testing labs and DNA databasing labs, that is, some local labs do not
directly upload their profiles into the CODIS system, but send them to another lab in the same state
that does databasing.
Although procedural rules and regulations vary between labs and across jurisdictions, the core
work of forensic scientists in government crime labs is essentially the same. It comprises tasks
such as determining the usefulness of a DNA sample provided by crime scene investigators, the

8 The American Review of Public Administration XX(X)
preparation of a sample for analysis, the interpretation of DNA mixtures (e.g., when the DNA of
two or more individuals is present in a sample), creating DNA profiles, and uploading the DNA
profiles of convicted offenders into a database. The vast majority of procedures involved in these
tasks is described in painstaking detail in the labs’ manuals, which are constantly updated to reflect
the rapid change of technology and the resulting new procedures. Furthermore, the FBI issues
regulations that govern access to CODIS. Lab procedures are regularly audited under the auspices
of the FBI, and need to conform to standards set by the National Institute for Standards and Tech-
nology (NIST). Despite this high level of codification, there remains a considerable amount
of uncertainty in the work of DNA forensic scientists. For example, as we mentioned above, the
processing of DNA mixtures requires interpretation. There are a number of statistics that can be
used in this case, and in fact, different labs prescribe the use of different statistics (Butler, 2005).
Another example regards the adoption of newly available technologies. Apart from the decision
on the technology per se, there is a host of private vendors to choose from, and the quality of the
purchased material as well as the available support from vendors can significantly influence both
the efficiency and efficacy of a lab. Finally, there is a wide array of discretionary management
issues, ranging from human resource issues to decisions around the design of new laboratory space.
Dealing with these uncertainties often requires coming up with creative solutions, which result in
the adaptation of preexisting procedures and constitute a form of innovation (Woodman, Sawyer,
& Griffin, 1993).
DNA laboratories in the United States, as a population, offer a good substrate within which to
study knowledge sharing within government. There is a discrete population of laboratories (about
180), with at least one laboratory in every state. In this undeniably dynamic knowledge-intensive
domain, there is substantial discretion at the laboratory level. They present a “petri dish” within
which to study the possibilities of knowledge sharing within a decentralized system of
government.
Data Collection
Our goal was to compile a comprehensive sample of individuals with the most common profes-
sional roles, reflecting behavior across different types and sizes of labs, to understand the entire
system under inquiry. Respondents for this study were selected through purposeful sampling (Yin,
1994) according to the professional roles held by members of the community in a single case study
design with multiple sites. This method of sampling allows for comparability between the respon-
dents and at the same time incorporates the range of different realities that characterize the various
government DNA labs. We initially focused our recruitment efforts on state CODIS administrators,
because we assumed that these individuals had greater needs to connect to their peers in other labs,
as there is only one CODIS administrator per lab. After a first round of telephone interviews, two
of the authors attended several CODIS conferences and interviewed state administrators in person.
Attending these conferences allowed these authors to gain insight into the interaction patterns
among CODIS administrators. Early participants in the study led us sequentially to additional
important respondents within the community (Miles & Huberman, 1994). In particular, it soon
became clear that it would be beneficial to extend the sample to other professional roles as well as
to individuals in local labs. We therefore sought to be introduced to such individuals through our
initial respondents. We stopped recruiting additional respondents when we started getting very
similar responses and therefore had reached saturation in our sample. Our final pool of respondents
consisted of 33 individuals, from 30 labs and 26 states.
The roles in our final sample were those of CODIS administrator, technical manager, and lab
director. The CODIS administrator is the individual in a databasing lab who is the gatekeeper for
the CODIS database and determines what goes into NDIS, if she is a state administrator. The

Binz-Scharf et al. 9
technical manager is in charge of the actual laboratory work. Thus, the technical manager oversees
the development of procedures, and the distribution of tasks. Whereas the positions of CODIS
administrator and technical manager are generally held by different individuals in large labs, in
smaller labs it is common for one individual to wear multiple hats; that is, the CODIS administra-
tor and the technical manager can be the same person. Similarly, the range of roles a lab director
holds varies significantly with the size of the lab. The lab director generally has the greatest
exposure in terms of contacts outside the lab.6 An overview of respondents’ professional roles
and their affiliations is presented in Table 1.
We conducted semistructured, open-ended interviews with these individuals, lasting between
30 minutes and two hours each. The interviews covered the following topics:
Description of work function and work environment: The hierarchical relationships the
respondent is embedded in and the proximity to peers within the lab;
Description of knowledge required for the job: The areas of expertise and the types of knowl-
edge the interviewee required, such as technical or legal knowledge, advice, or opinions;
Habitual knowledge sources: The most commonly used knowledge sources for the identified
types of required knowledge, media and venues used, and difficulties in retrieving knowl-
edge from these specific sources;
Engagement in the community: The behavior of interviewees when approached with a ques-
tion, in particular regarding their motivation to set aside time to answer questions from
colleagues and peers, the content of questions, and reasons for answering certain questions
rather than others.
As our data collection proceeded, we also included specific questions about the nature of ties
to individuals in other labs.
In addition to the abovementioned data collection efforts, two of the authors visited several
crime labs and observed forensic scientists at their workplace. Although these were not full-fledged
ethnographies, the observers were nonetheless able to get a sense of the localized settings and
work practices in these labs (Emerson et al., 2001), which proved of utmost importance to under-
standing the work of forensic scientists as described and done by themselves as well as the situ-
ational variations between laboratories.
Data Analysis
The analytic process was driven by the objective of making sense of the data by identifying recur-
ring patterns. To this end, we followed the coding and analysis process described in Miles and
Huberman (1994). All interviews were transcribed verbatim, and coded in a three-step process.
During the first step, which Miles and Huberman refer to as open coding, each author carefully
reviewed the interview transcripts as they became available and attributed labels (or codes) to
Table 1. Overview of Interviews Conducted
Professional rolea Affiliation: State lab Affiliation: Local lab Total
Lab director 8 2 10
Technical leader 7 3 10
CODIS administrator 8 5 13
Total 23 10 33
a. In cases where individuals hold multiple roles, the highest-ranking role is indicated.

10 The American Review of Public Administration XX(X)
meaningful passages. The authors then met as a team and discussed each label, resulting in a
provisional start list of codes. This list of codes was subsequently used by a team of research
assistants to code all interviews with the help of the qualitative research software package NVIVO
(2006). The research assistants were instructed to code line by line, that is, the smallest coding
unit was to be one line of text. Each member of the research team also wrote memos, or short,
reflective remarks about codes and their relationships, which became part of the data. One author
periodically met with the team of research assistants and discussed issues that arose during indi-
vidual coding. Once the initial coding was completed, a senior research assistant ran a coding
comparison to ensure intercoder reliability. The level of intercoder reliability was 85%, which is
considered acceptable (Smith, Feld, & Franz, 1992).7 Any discrepancies between the individual
coders were discussed until an agreement was reached.
In the next step, referred to as axial (or pattern) coding, similar codes were grouped together
into a smaller number of meta-codes. For example, we clustered all codes that referred to the way
individuals search for knowledge, what sources they turn to, and the rationales behind those
choices, into a meta-code titled “search strategies.” This second level of coding helps generalize
observations and identify “repeatable regularities” (Kaplan, 1964, quoted in Miles and Huberman,
1994, p. 69). In the example above, we were able to compare the search strategies of each respon-
dent and thus better understand emerging patterns of search.
A further step in the analysis was that of constructing matrix displays to study the relationship
between meta-codes and select characteristics of the respondents. To continue our previous example,
we categorized respondents according to their professional roles, their affiliations, the size and
jurisdiction of their labs, their educational background, and their expertise. We then studied how
search strategies varied according to these attributes. This allowed us to find confirmation for our
initial assumption that scientists in smaller labs reach out to individuals in larger labs in search for
answers, as larger labs tend to have more resources and therefore, for example, are more likely to
have newer equipment and experience with it. However, the matrix displays helped us generate
important insights that were somewhat counterintuitive, such as the observation that individuals
who are considered experts in the network are not necessarily affiliated with a large lab, thus mov-
ing certain small labs toward the center of the network of practice.
Throughout the data analysis process, we followed a procedure of doubling back and forth
between our conceptual framework, the data, and the literature (Strauss, 1987). As an illustration
of this procedure, after an initial round of data collection, we discovered that our conceptual frame-
work did not take into account some of the themes that were emerging. We then went back into
the literature and searched for confirmation of these themes. For example, early interviews revealed
that many participants subscribed to listservers, and the opinions on and use of these listservers
varied considerably. We therefore reviewed specific literature on this topic and included it in our
analysis of the data. However, the review of this new literature brought about new conceptual and
empirical angles that we had not previously considered, and led us to modify the interview guideline
to include questions addressing those issues in subsequent fieldwork.
Findings
Searching for answers, it appears, represents a considerable part of the workload in a knowledge-
intensive profession such as that of forensic scientists. Although most of the participants in this
study took their enormous knowledge-sharing efforts for granted as part of their daily work, the
picture that emerged from the individual accounts and our observations was that of a busy advice
network across hierarchical and organizational boundaries, often overcoming, but at times yielding

Binz-Scharf et al. 11
to, classically bureaucratic barriers to knowledge sharing. In what follows, we present our findings,
based on our analysis and interpretation of the fieldwork and the literature.
Access to Knowledge Beyond Organizational Boundaries
The first key question is whether interorganizational knowledge sharing networks matter. That
is, do participants reach out to individuals in other organizations to find answers to questions that
arise on their job? The answer was an unambiguous yes: Accessing knowledge beyond organiza-
tional boundaries occurs regularly, and is essential for individual effectiveness. Every single
participant referred to instances where talking to individuals outside of their immediate organiza-
tion provided information that made them more effective. The account of this participant, a CODIS
administrator, is typical:
[F]or an actual DNA data banking type question there is very little literature on it. It is all
in people’s heads, the information. So most of those issues are handled by calling people
because there simply is not much to look for in print, either as a scientific paper or some
kind of a text book because no such thing exists on this—that I’m aware of anyway—that
is up to date and will give me information about a very specific new problem and so on.
As a CODIS administrator, this person makes decisions, generally speaking, about which DNA
profiles get uploaded into the database. The answer to this type of questions (which the participant
refers to as a “DNA data banking type question”) is not simply a binary one—there are many gray
areas in determining what constitutes a legitimate profile. For example, there could be a mixture
of DNA samples from a crime scene, yielding an “excess” of alleles (genetic markers) at each
locus. The decision regarding whether to upload a profile involves judgments on what constitutes
evidence from a crime scene, interpretation of a particular set of analyses, and interpretation of
the rules for inclusion in the national database. Although many of these procedures are recorded
in a lab’s manuals, these manuals are not exhaustive, especially considering the fast pace at which
DNA technology is evolving. Hence, CODIS administrators often need to reach out to peers who
might have experience with the specific problem the individual is tackling, as this interviewee
makes clear:
The process goes from local manuals, [to] scientific publications, and then [to] peers. And
when I say peers, now I’m talking outside of the laboratory. Because if we do not have it
in our manuals, how to do something, we have to look outward.
The use of “local manuals” maps perfectly to Weber’s (1921/1968) “knowledge of the files” by
bureaucrats. This particular quote highlights the efforts in this location to formally capture local
knowledge. The existence of local documentation is in part mandated by external regulatory pro-
cedures—for example, FBI rules for access to the national database and American Society of Crime
Lab Directors (ASCLD) accreditation require extensive documentation.
The use of “scientific journals” highlights that in this domain there are formal epistemic infor-
mational resources that exist outside of the formal hierarchy that nonetheless may be authoritative.
The fact that it is accepted that there is such authoritative epistemic knowledge creates the potential
for interpersonal information exchange. That is, accepting that someone from outside of the orga-
nization might have insight into how to do things within one’s organization requires accepting that
knowledge transcends the local setting (although for the limits on this, see our observations on
perceptions of local exceptionalism below).

12 The American Review of Public Administration XX(X)
From our perspective, the more general principle that this subject is illuminating with respect
to interpersonal interorganizational networks is that if the information does not exist locally, it is
necessary to consult individuals outside of the organization, as this interviewee indicates:
We’re limited on [statistical] knowledge, so we do need to go outside for that. From a
casework perspective we’d like to think we’ve just about seen it all, here. [. . .] We’ve been
working with it for so long that, we’d like to believe that we can troubleshoot things here,
with the knowledge base that we have. So I would say it’s probably a statistic type issue.
Maybe a kinship issue, relatedness. They’re talking about searching databases now to find
family members, the aspects involved with that. So that’s probably just beyond where our
comfort level is, and we would ask advice on that.8
These data suggest an interplay between standard hierarchical mechanisms for informational
search and network mechanisms. In particular, it appears that individuals sequence their search
processes following a pattern molded by the bureaucratic context in which they function: First, they
resort to the local sources of information within the standard hierarchical structures of the organiza-
tion, and then they shift to informal, external resources.
Knowledge Distribution Within the Network
Given that there are vastly uneven resources within the system, as well as accidental variations
in experience, it is likely that the knowledge will be unevenly distributed, yielding network-wide
reservoirs of expertise. Participants systematically referred to particular corners of the system
regarding their reputation in particular domains. For example, this individual refers to the expertise
of another lab with respect to their recommendations regarding hardware:
[I] rely on places that are very good at researching what is in the community or in the busi-
ness world. The [name of lab] is wonderful. Before they implement something they check
the computer industry, all these different industries before they make a decision. So I rely—
I’ve got a relationship with them that I enjoy. And I call them and ask, “Have you heard of
anything like this?”9
Unsurprisingly, the interviews suggested that there is a large asymmetry in the flow of knowledge
based on the presence or absence of local resources. That is, the asymmetry is clearly driven in large
part by asymmetries in size. We should note, however, that we did find that there were individuals
in smaller labs who had system-wide reputations, and who were clearly important hubs in the broader
network of knowledge sharing. The DNA crime laboratories across the United States constitute a
very heterogeneous population, and that greatly affects the knowledge searching of individuals.
Labs range in size across orders of magnitude—from many labs with just a handful of examiners
to a handful of labs with hundreds of examiners. We observed that within large labs, knowledge
searching typically stays within the labs. However, large labs, by virtue of their sheer volume of
experience, often become critical reservoirs of knowledge for the system, where individuals from
small labs often talk to people in large labs to tap into their experiences, as one interviewee describes:
I find that when you’re a small laboratory and you don’t have the time, money and personnel
to put in to starting new programs, you’re better off to sit back and listen to what everybody
else has done, rather than reinvent the wheel. So that’s kind of my philosophy on it. That’s
kind of how I go about getting information. If somebody has already validated a method or

Binz-Scharf et al. 13
they’re using a particular type of procedure and they say, “Hey, this is the thing to go to,”
I would defer to their judgment.
We also found some degree of hierarchy within the network, where large states often had a
single lab that was much larger than other labs, which then served as the informational hub for the
smaller labs in the state, as one participant highlights as being the case in his state: “The one
resource that we’ve definitely leaned on a lot more than the others is the [name of biggest lab in
state] lab. Based on our experience, they’re the cream of the crop basically.”10
Although reciprocity is often cited as being one of the building blocks of networks (Coleman,
1988; Gouldner, 1960), these large asymmetries in knowledge sharing limit the role that reciprocity
in kind can play in sustaining this knowledge sharing network. More generally, reciprocity cannot
play a major role in sustaining a knowledge-sharing network if knowledge is very unevenly dis-
tributed. That is, the uninformed have little information to offer the knowledgeable, and yet the
most important transactions from the systemic perspective in such a scenario are for the knowl-
edgeable to help the unknowledgeable. Elsewhere we explore with these data what drives voluntary
engagement in knowledge sharing (Mergel et al., 2008). Why, for example, do individuals from
the labs with large reservoirs of knowledge even create ties outside of their organizational bound-
aries (and, in particular, why with labs with far less resources)? Part of the answer clearly lies in
the institutionalized extra network processes that facilitate the creation of ties (Feld, 1981), which
we discuss below.
Mechanisms Facilitating Cross-Organizational Ties
There are system-wide processes that facilitate the creation of interorganizational interpersonal
ties. Several mechanisms are in place that allow for knowledge exchange across labs. One such
mechanism is that of lab audits, or inspections. The inspection process represents an important
control function for the FBI. The FBI mandates adherence to particular practices if a laboratory
is to remain connected to the NDIS. The inspectors typically include examiners from other labs.
The inspections thus have a standard hierarchical function, but with a twist. That is, the FBI has
authority within the network, not because they have formal authority over local DNA labs, but
because they have authority over the infrastructure that labs need access to. Here, our focus is on
the impact that the inspection process has on knowledge sharing, which is considerable. Inspec-
tions require a fairly substantial engagement with the practices of another lab, and thus allow for
the spread of a large amount of tacit knowledge. One individual describes the process:
We draw upon other agencies: Their policies and procedures, and also just general experi-
ence and knowledge of the individual examiners. That includes the FBI. That includes the
neighboring states. It is very common for us to go out and do an audit of another laboratory
and bring some of that knowledge that we have gained in the audit home with us.
Inspectors are thus a key conduit of information among labs—both among the labs that they
inspect as well as back to their home lab.
Conferences played a particularly important role in facilitating the creation of ties. Participants
repeatedly referred to three key conferences. Two of the conferences are annual conferences aimed
specifically at forensic scientists: The annual meeting of the American Academy of Forensic
Scientists (AAFS), and a conference organized by a large vendor (Promega). In addition, the FBI
holds a conference/training aimed specifically at CODIS administrators, allowing them to make
contacts with CODIS administrators from other states. Notably, the FBI provides resources to
state and local labs to attend (again, there is an interweaving of an element of hierarchy in the

14 The American Review of Public Administration XX(X)
network). The participants in this study repeatedly referred to the role these conferences played
in facilitating meeting people from other labs. This intersection of individuals and organizations
repeats itself over the years, building deeper relationships. The development of deep relationships,
in turn, should facilitate instrumental use of those relationships for knowledge exchange, especially
for complex, tacit knowledge (Hansen, 1999).
The Importance of Preexisting Personal Relationships
The social network literature has shown without fail that prior relationships are an important
determinant for choosing a source of advice (e.g., Granovetter, 1973; Krackhardt & Hanson, 1993;
Nelson et al., 2001), which is corroborated by our data. In particular, we observed that individuals
first turn to their friends when looking for advice, as one participant describes:
I’d start with people I know fairly well. And [people] I know that [I have] seen a lot. And
I would call someone [from] our surrounding states, because I’m friends with some [people
from] our surrounding states. So I’ll start with my friends first.
Another participant highlights both the nurturing of relationships through conferences as well
as the subsequent instrumental value of those ties:
You not only get that information during the day, but I think the most important aspect of
the meetings is what happens in the evenings. Those bonds you develop as friends allow
you to develop those contacts over the years. And that has been an instrumental tool for me
in my job. And I think it makes me twice as good as I would be otherwise. I mean my abili-
ties are enhanced tremendously by my contacts.
This finding highlights the intrinsically interpersonal nature of interorganizational knowledge
sharing. It is clear that within organizations, there are often defined channels for resolving ambigu-
ity. However, with a few important exceptions, knowledge sharing between organizations within
this system relied on personal relationships particular individuals had developed over years. One
of the key exceptions involved information about the rules surrounding CODIS, regarding for
example, what was allowed into the database. Here many individuals reported going directly to
the FBI, after following internal channels, as this participant states:
I’d call the FBI [about CODIS rules]. [. . .] But I wouldn’t mess with anybody besides those
people as far as from my perspective. Now we have somebody who’s in charge of that in
our laboratory, so if there’s a general question, I’d probably go to our CODIS administrator
first, ask them before we’d go to the CODIS people themselves. And that’s how we gener-
ally handle that. I wouldn’t necessarily go to another laboratory because those standards
are open to interpretation and I figure, why mess with other peoples’ interpretation, when
the FBI’s the one really making the ground rules. So I’ll just go straight to them.
As can be seen from these quotes, individuals draw on relationships of varying intensity. The oft-
hypothesized trade-off between strong and weak ties is between the greater flow of information from
strong ties and the higher redundancy of information that comes from strong ties (e.g., Burt, 1992;
Granovetter, 1973). The argument is that while strong ties are higher bandwidth, they often provide
redundant information because an individual's strong ties tend to know one another and thus have
overlapping information. Consider this account of a local network characterized by strong ties:

Binz-Scharf et al. 15
[W]ithin our community, we have a private laboratory, just up the street, a big one [name].
We use them a lot for technical knowledge. They are brilliant people up there. [. . .] It’s
because we’ve faced the same challenges, basically. And [. . .] sometimes they ask us, “How
would you do this sample? Could we look at your protocol for this particular type of sample?”
And we’ll say, “Absolutely. Come up. We’ll show you how we do it.” And the reason being
that they are truly intertwined in the same battle we are, trying to get DNA from crime scene
samples. Trying to get DNA profiles published that meet the quality assurance criteria. So
we sometimes interact with them, and sometimes we interact with the neighboring states.
It’s easiest to call [neighboring state A], because they’re our close neighbor. It’s easiest to
call [neighboring state B], because they’re a close neighbor. Not very often that I would
call [a faraway state] and ask them a question.
This local orientation in tie formation can lead to a network of strong, but redundant, ties. Our
data do not speak directly to whether that is what the network looks like, but it should be a natural
consequence of this orientation, and can potentially constitute a barrier to the spread of knowledge.
Obstacles to Information Sharing: Reputational Concerns
Our interviews revealed a number of significant obstacles to seeking information through the
network. One major concern was that a question revealed information about the individual asking
the question: Ignorance.
In fact, we observed that concerns regarding reputation of being knowledgeable constitute a major
obstacle to the interpersonal sharing of information. Individuals generally were concerned about their
reputation among their peers, as one individual states: “Sometimes it’s difficult because you don’t
necessarily always like to admit what you don’t know.” Therefore, before she turns to any other
human for advice, this forensic scientist puts great time and effort into ensuring that her question is
indeed a legitimate one (as defined by community norms). Furthermore, the fear of revealing their
ignorance on an issue often prevents individuals from reaching out to a larger group of people, as is
the case when a question is posed on a listserver, and pushes them to seek out informants with whom
they have a relationship of trust (e.g., friends). The following account illustrates these concerns:
There is a professor in statistics that I’ve taken a couple a classes for training. [. . .] I know
him well enough that I have called him on an occasion or two to access his area of expertise
. . . [So] if it’s a problem I knew for sure we couldn’t handle here, I go directly to the statisti-
cian I mentioned. If it was one that I believe we handled correctly and I’m looking for
confirmation, I’m likely to contact all or some of those folks on [the listserver].
We would attribute this phenomenon in part to the still prevailing culture in public agencies,
where the motivational emphasis lies on mistake avoidance rather than increased productivity
(Schofield, 2001), and peer pressure strongly affects performance (Wilson, 1991).
A second element was the concern about ramifications outside of the network. Recall that all of
this is taking place within an adversarial criminal justice system. Reputation can take the form of
being discredited on the stand as an expert. The evolution of the use of listservers offers a particularly
interesting case study of the role of the reputation outside of the network. As one participant recounted,
a popular list regarding forensics expanded to include an increasing fraction of the outside com-
munity: “So the interesting thing is that the listserver evolved, and it evolved to include people who
are non-forensic people. There were people on that list that were looking to overturn their convic-
tions.” The inclusion of other people on the list—notably defense attorneys—created a concern that
what was said on the list might come back and haunt individuals later. This provided the impetus