Вы находитесь на странице: 1из 32

r Academy of Management Journal

2016, Vol. 59, No. 4, 1308–1338.


http://dx.doi.org/10.5465/amj.2013.1214

COORDINATING KNOWLEDGE CREATION IN


MULTIDISCIPLINARY TEAMS: EVIDENCE FROM
EARLY-STAGE DRUG DISCOVERY
SHIKO M. BEN-MENAHEM
GEORG VON KROGH
ZEYNEP ERDEN
ETH Zurich

ANDREAS SCHNEIDER
Ypsomed

Based on a multi-year field study of early-stage drug discovery project teams at a global
pharmaceutical company, this paper examines how multidisciplinary teams engaged in
knowledge creation combine formal and informal coordination mechanisms when faced
with unpredictable interdependencies among specialists’ knowledge domains. While
multidisciplinary teams are critical for knowledge creation in increasingly specialized
work environments, the coordination literature has been divided with respect to the
extent to which such teams rely on formal coordination structures and informal co-
ordination practices. Our findings show that when interdependencies among knowledge
domains are dynamic and unpredictable, specialists design self-managed (sub-)teams
around collectively held assumptions about interdependencies based on incomplete in-
formation (conjectural interdependencies). These team structures establish the grounds
for informal coordination practices that enable specialists to both manage known in-
terdependencies and reveal new interdependencies. Newly revealed interdependencies
among knowledge domains, in turn, promote structural adaptation. Drawing on these
findings, we advance an integrative model explaining how team-based knowledge crea-
tion relies on the mutual constitution of formal coordination structures and informal
coordination practices. The model contributes to theory on organizational design and
practice-based research on coordination in cross-disciplinary knowledge creation.

Organizations increasingly rely on multidisciplin- Jones, & Uzzi, 2007). Understanding how such teams
ary teams for creating high-yielding knowledge1 at effectively coordinate knowledge creation across
the frontier of science and technology (e.g., Hoegl & knowledge domains—i.e., manage interdependencies
Gemuenden, 2001; Singh & Fleming, 2010; Wuchty, among specialists—is a subject of growing importance

1
We apply Liebeskind’s (1996) definition of knowledge as
“information whose validity has been established through
Authors contributed equally. This research was supported tests of proof” (p. 94). The main purpose of creating knowl-
by a grant from the Swiss National Science Foundation edge through tests of proof is to reduce uncertainty and en-
(SNF 100018-146439). hance an actor’s capacity to act (Huber, 1991). Because the
We thank Scott Sonenshein and three anonymous re- validity of new knowledge can only be established after
viewers for their insightful comments and suggestions the fact, knowledge creation is a fundamentally uncertain
throughout the review process. We gratefully acknowledge endeavor (Liebeskind, 1996). Organizational knowledge
Natalie Reid for her editorial assistance and Fabienne creation refers to “the process of making available and
Vukotic for her research assistance. We also thank Vivianna amplifying knowledge created by individuals as well as
He, Pursey Heugens, Johannes Meuer, Michaéla Schippers, crystallizing and connecting it to an organization’s knowledge
Nicole Rosenkranz, Christian Wedl, as well as participants system” (Nonaka & von Krogh, 2009: 635). This definition
of the faculty seminar series (Professor Tobias Kretschmer) emphasizes that, in an organizational context, knowledge
at Ludwig Maximilian University of Munich for their feed- creation hinges on social interactions and collaboration be-
back on earlier drafts of this work. We further express our tween individual actors (e.g., Grant, 1996; Kogut & Zander,
gratitude to the participants in this study. 1992; Nonaka, 1994).

1308
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express
written permission. Users may print, download, or email articles for individual use only.
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1309

(e.g., Bruns, 2013; Kotha, George, & Srikanth, 2013; 2014; Okhuysen & Bechky, 2009), there is a lack of in-
Majchrzak, More, & Faraj, 2012; von Krogh, Nonaka, & depth analysis of the process through which formal
Rechsteiner, 2012). As the complexity of scientific coordination structures and informal coordination
problems increases, and as efficiency gains from cu- practices interact in multidisciplinary knowledge-
mulative learning drive individuals to specialize creating teams. Without taking these interactions into
(Becker & Murphy, 1992; Jones, 2009), multidisci- account, theory on how teams coordinate knowledge
plinary teams increasingly face unpredictable in- creation remains incomplete. In particular, it is un-
terdependencies among specialists due to high task clear not only how formal structures evolve when
uncertainty (i.e., incomplete information about tasks) designers of team structures face unpredictable and
(Argote, Turner, & Fichman, 1989; Cardinal, Turner, changing interdependencies among knowledge do-
Fern, & Burton, 2011; Gardner, Gino, & Staats, 2012). mains (Grandori & Soda, 2006; Hülsheger, Anderson, &
While it is well established that such conditions pose Salgado, 2009; Puranam & Raveendran, 2013) but
massive challenges to coordination of knowledge work also how informal coordination practices are shaped
(Ancona & Caldwell, 1992; Cronin & Weingart, 2007; by such evolving formal structures (e.g., McEvily
Dougherty, 1992; Faraj & Sproull, 2000; Mell, van et al., 2014).
Knippenberg, & van Ginkel, 2014; Vural, Dahlander, & This paper develops theory on how multidisci-
George, 2013), scholars have developed conflicting plinary teams engaged in knowledge creation com-
theories about the role of formal and informal co- bine formal and informal coordination mechanisms
ordination mechanisms in knowledge creation. when faced with unpredictable interdependencies
Some studies, building on a long tradition of orga- among specialists’ knowledge domains (Grant, 1996;
nizational design research (Galbraith, 1973; Tushman Thompson, 1967). We do so by using a longitudinal
& Nadler, 1978) and structural contingency theory qualitative study of early-stage drug discovery teams
(Burton & Obel, 2004; Donaldson, 2001; Thompson, at DrugCo,3 a global research-driven pharmaceutical
1967), propose that formally designed team structures company. Our analysis illuminates important re-
are necessary for effectively coordinating knowledge lationships between formal structures and informal
intensive work. Such structures specify the grouping coordination practices. To disaggregate the knowl-
and linking of interdependent organizational members edge creation process and manage the integration of
and prioritize information-processing interactions specialists’ efforts, multidisciplinary project teams
(e.g., March & Simon, 1958; Puranam, Raveendran, & delegate responsibility for knowledge creation to
Knudsen, 2012), thus promoting team members’ ability dynamic sub-team structures based on conjectures
to coordinate their activities (Bresman & Zellmer-Bruhn, about interdependencies among knowledge do-
2013; Bunderson & Boumgarden, 2010). In contrast, mains. Within these structures, team members enact
a growing number of studies adopting a practice-based their responsibility for knowledge creation through
perspective (Orlikowski, 2000) focus on the coordina- a set of relatively stable informal coordination prac-
tion of knowledge work in teams as informally emerg- tices, which we conceptualize as anticipatory con-
ing patterns of interactions enacted through specialists’ forming, workflow synchronizing, and cross-domain
everyday practices (e.g., Faraj & Xiao, 2006; Hargadon & triangulating.
Bechky, 2006; Kellogg, Orlikowski, & Yates, 2006). From our findings, we develop an integrative
These studies generally suggest that formally designed model suggesting that formally designed coordina-
coordination structures may stifle knowledge creation tion structures and informal coordination practices
(Okhuysen & Bechky, 2009). are mutually constitutive, thus jointly enabling spe-
The discrepancy between these two perspectives cialists in multidisciplinary teams to coordinate
is symptomatic of a more fundamental disconnect in their knowledge creation efforts. The model pro-
the literatures on the formal and informal aspects of vides new insights into how formal team structures
coordination (e.g., Gulati & Puranam, 2009; Soda & establish the grounds for informal practice-based
Zaheer, 2012).2 While scholars have noted that this coordination of specialists’ knowledge creation ac-
disconnect limits our understanding of how organi- tivities (e.g., Bruns, 2013). It also develops an un-
zations function (e.g., McEvily, Soda, & Tortoriello, derstanding of how specialized team members’
everyday coordination practices lead them to con-
2
A similar debate on formal and informal aspects of tinuously uncover new interdependencies among
organization appears in a related stream of literature on
3
organizational control (e.g., Cardinal, Sitkin, & Long, 2004; Names of the organization, projects, and individuals
Sitkin, Cardinal, & Bijlsma-Frankema, 2010). are pseudonyms.
1310 Academy of Management Journal August

knowledge domains. These newly revealed interde- organization’s structural design should align the or-
pendencies necessitate a reallocation of responsi- ganization’s capacity for information processing with
bilities for knowledge creation, thus driving project the needs for information exchange (Galbraith, 1977;
team members to redesign formal team structures. Tushman & Nadler, 1978). In line with this view, or-
In advancing this model of how multidisciplinary ganization design scholars proposed a structural
teams coordinate the knowledge creation process, contingency theory (Donaldson, 2001; Perrow, 1967;
our study offers a much-needed integration of theory Thompson, 1967) stipulating that as task uncertainty
on formal and informal coordination mechanisms increases and interdependencies among special-
(McEvily et al., 2014; Okhuysen & Bechky, 2009). ists become less predictable, structural designs with
higher capacities for information processing based on
more extensive interaction become more appropriate
LITERATURE REVIEW
(e.g., Argote, 1982; Burns & Stalker, 1961; Sherman &
Relevant theory and research on coordination of Keller, 2011).
knowledge creation in multidisciplinary teams can be Some scholars, however, cast doubt on this trade-
grouped into two perspectives (McEvily et al., 2014; off, suggesting that formal coordination structures
Okhuysen & Bechky, 2009). First, the organizational may benefit knowledge creation in uncertain contexts
design perspective suggests that the division of la- (e.g., Cardinal, 2001; Jelinek & Schoonhoven, 1990;
bor in teams with specialists representing distinct Child & McGrath, 2001; von Krogh, Nonaka, &
knowledge domains necessitates the integration of Rechsteiner, 2012). This argument has strong sup-
individual efforts, giving rise to interdependence port from research demonstrating that formal struc-
among specialists (e.g., Burton & Obel, 2004; tures prompt team members to engage in joint
Lawrence & Lorsch, 1967; Thompson, 1967). Building problem-solving and helping behaviors, thus en-
on classic information processing theory (Galbraith, abling the team to more effectively accomplish highly
1977), organization design scholars argue that the uncertain tasks with high information-processing
coordination of efforts from interdependent special- needs (e.g., Edmondson, 1999, 2003; Valentine &
ists relies on formal structures4 that mandate in- Edmondson, 2015). Studies also suggest that formal
dividuals’ information provision, relationships, roles, structures enable teams to innovate by significantly in-
and responsibilities. Puranam et al. (2012) posit that creasing the potential for information sharing among its
formal structures enable coordination by grouping members (Bunderson & Boumgarden, 2010) and pro-
and prioritizing interactions among organizational moting learning in self-managed teams facing non-
members with epistemic interdependence—that is, routine tasks (Bresman & Zellmer-Bruhn, 2013). Such
members whose optimal actions depend on their ca- findings are consistent with a more general social sci-
pacity to predict what other members will do. Thus, ence notion that formal structural design may enable
by directing information exchange between interde- coordination, communication, and exchange of knowl-
pendent members and enabling them to predict one edge among group members (Lamont & Molnár, 2002).
another’s actions, formal structures allow specialized The second stream of research on coordination
team members to better integrate their individual ef- takes a practice perspective, emphasizing the impor-
fort and prevent coordination failures. tance of the informal emergent aspects of coordination
Early organizational design literature poses a sharp (Okhuysen & Bechky, 2009). This perspective chal-
trade-off between the benefits of formal structures in lenges earlier design approaches, such as structural
driving efficiency and control, and the benefits of in- contingency theory, for suggesting that coordination
formal, emergent relationships in driving creativity necessitates formal structural designs based on pre-
and innovation (Burns & Stalker, 1961; Gittell, 2002; defined interdependencies (Faraj & Xiao, 2006). A
Scott, 1987). Underlying this perspective is a core fundamental argument is that a focus on formal co-
assumption in information processing theory that an ordination structures obscures the increasing demands
placed on members of post-bureaucratic organizations,
4 i.e., dealing with unplanned contingencies and emer-
Although scholars have adopted diverse definitions
gent interdependencies (Kellogg et al., 2006). In con-
of structure, we follow Davis, Eisenhardt, and Bingham
(2009), who conclude that prior definitions “all share an trast to the design perspective, the practice perspective
emphasis on shaping the actions of organizational mem- concentrates on coordination practices as they unfold.
bers,” so that “entities are more structured when they It thus emphasizes the need for a dynamic under-
shape more activities of their constituent elements and standing of emergent, adaptive coordination in teams
thus constrain more action” (2009: 415). engaged in knowledge work (Okhuysen & Bechky,
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1311

2009). By exploring coordination practices in the possibility that existing formal structures may not
context of high task uncertainty, widely distributed only inhibit but also support the integration of spe-
expertise, and fluid interdependencies, practice stud- cialists’ efforts under a variety of unpredictable cir-
ies adopting a dynamic view of coordination add im- cumstances (e.g., Brass, Galaskiewicz, Greve, & Tsai,
portant insights into how teams integrate specialist 2004; Hollenbeck, Ellis, Humphrey, Garza, & Ilgen,
knowledge (e.g., Bruns, 2013; Majchrzak et al., 2012). 2011; Jelinek & Schoonhoven, 1990; Pennings, 1992).
For example, in a study of expertise coordination In line with McEvily et al.’s more general ob-
in medical trauma teams facing high uncertainty servation that “it is essential to clarify the condi-
from fluctuating patient arrival rates, Faraj and Xiao tions under which formal and informal elements
(2006) show that complex and highly interdepen- [of organizations] interact” (2014: 333), this study
dent medical work relied on emergent, partially im- draws on longitudinal qualitative fieldwork to il-
provised coordination practices. Similarly, Bechky luminate how formal coordination structures and
and Okhuysen (2011) show that for unexpected informal coordination practices co-evolve. In so
events, police SWAT teams and film production doing, we contribute a more comprehensive un-
crews coordinated expertise by flexibly shifting derstanding of how knowledge creation is co-
roles, reorganizing routines, and reassembling their ordinated within multidisciplinary teams facing
work. Kellogg et al.’s (2006) study of a web-based unpredictable interdependencies.
marketing organization shows that by enacting
a “trading zone,” specialists coordinated their work
across the boundaries of their professional commu- RESEARCH DESIGN
nities. Within these zones, they made their work
Case Selection and Overview
“visible, legible, and aligned” when facing an un-
certain context. We conducted a single-site case study (Yin, 2003),
Thus both the organizational design and practice which allows for in-depth qualitative investiga-
perspectives offer key insights into the coordination tion of the coordination process. Our study focused
of knowledge work in team-based structures and the on early-stage drug discovery projects at DrugCo,
contingency effect of unpredictability. Yet, as recent a global pharmaceutical company ranked among the
literature reviews show (Kilduff & Brass, 2010; top 10 firms as measured by revenues and global
McEvily et al., 2014; Okhuysen & Bechky, 2009), market share for therapeutic drugs. Drug discovery
research has focused on either the formal or informal occurs in the first five to six years of a 12- to 15-year
aspects of coordination, whereas “the mechanisms development trajectory from the lab to the market
describing the interplay between formal and in- (Pammolli, Magazzini, & Riccaboni, 2011) and ac-
formal elements are less well-understood” (McEvily counts for about one-third (i.e., about $800 to 850
et al., 2014: 335). As a result, the literature presents million) of the total cost of bringing a new drug to
largely static and fragmentary insights into how market (Paul et al., 2010). Decisions made during this
specialists coordinate the knowledge creation pro- stage, such as which disease targets to study, are
cess across knowledge domains. critical for the overall success of commercializing
Specifically, formal structural designs appear crit- a drug (e.g., Munos, 2009; Rishton, 2005).
ical for developing predictive knowledge among Our study of early-stage drug discovery project
specialists in multidisciplinary teams, and thus for teams at DrugCo was guided by our interest in learn-
coordinating the integration of their efforts (Puranam ing how knowledge creation is coordinated in
et al., 2012). However, the design literature has a context where intense task uncertainty involves
largely overlooked the process whereby coordina- unpredictable interdependencies among knowl-
tion unfolds when interdependencies among spe- edge domains. DrugCo is a prototype of large phar-
cialists are partly unknown and change unpredictably maceutical companies with in-house R&D capabilities
(Grandori & Soda, 2006; Puranam & Raveendran, in the early exploratory and pre-clinical stages of
2013; Sherman & Keller, 2011). Practice-based re- drug discovery research. Modern early-stage drug
search offers important insights into how emerging discovery hinges on the efforts of multidisciplinary
interdependencies are informally managed under project teams where specialists from diverse knowl-
uncertainty (e.g., Bechky & Okhuysen, 2011; Faraj & edge domains need to continually coordinate their
Xiao, 2006). Research in this tradition, however, has knowledge creation (Drews, 2000; Sams-Dodd, 2005).
paid little attention to the structural context in Table 1 gives an overview of typical drug-discovery
which coordination practices unfold, and overlooks the knowledge domains.
1312 Academy of Management Journal August

This site also displays pervasive and inherent task mechanisms whereby coordination of knowledge
uncertainty, because drug-discovery specialists fo- creation across interdependent specialized knowl-
cus on what is unknown about the behavior of edge domains unfolds in multidisciplinary teams.
chemical compounds and the biological causes of Due to a personal contact of one of the authors with
a disease (Dougherty & Dunne, 2012; Grandori, 2010; a senior executive at DrugCo, we received full access
Northrup, 2005). A critical concern in drug discov- to examine everyday work in early-stage drug dis-
ery is that of minimizing side effects. Under the covery teams. This access is unique because much of
precautionary principle (Aven, 2011), pharmaceu- the knowledge creation precedes patenting; early-
tical companies are accountable for proving that new stage drug discovery processes are thus typically off
drugs will not cause the public significant harm. limits to outside field researchers.
Scientists’ everyday work must therefore be guided
by a strong sense of obligation to be cautious in the
Data Collection
context of task uncertainty. Thus we expected that
this extreme setting would enable us to elaborate Our early discussions and interviews with key
theory (Lee, Mitchell, & Sablynski, 1999) on the informants showed us that developing an in-depth

TABLE 1
Examples of Knowledge Domains in Drug Discovery Project Teams
Knowledge Domain Expertise Examples of Activities Examples of Artifacts and Tools

Medicinal chemistry Designing and synthesizing Finding ways of effectively Chemical molecules; solvents;
molecular compounds with the synthesizing new compounds; glassware (e.g., distilling column,
aim of developing drug-like designing new compounds with clamps, boiling flasks, funnels);
compounds improved activity, stability, tongs; test tubes
selectivity, or safety profiles
Molecular modeling Computationally designing new Running computational algorithms Computational infrastructure and
molecules binding specifically to to fit molecules to target proteins; hardware; 3D glasses;
the target molecule docking fragments in the binding computational algorithms;
pocket of the target molecule; software; three-dimensional
predicting the binding of rendering of molecules and target
molecules to target proteins; online databases
In-vivo biology Designing and conducting animal Injecting drug molecule in animal Tissue samples; animal models;
experiments to link target protein models; preparing and staining colored pictures of tissues; in-vivo
with disease state of the model tissue sampled from animal laboratories with hutches
models; observing disease state of
animals
High-throughput Developing and operating HTS Scaling up assays; preparing Screening infrastructure; well
screening (HTS) assays to rapidly identify a large samples for HTS screening; plates; pipetting robots;
number of potential drug-like maintaining screening compound libraries; lists with
molecules (hits) infrastructure; interpreting HTS activity profiles for over 10,000
screens compounds
Biochemistry Developing biochemical assays to Producing proteins in bacteria or Proteins; bacteria; bio-reactors used
assess the activity of target mammalian cells; harvesting and for growing bacteria and
proteins purification of proteins; designing mammalian cells; infrastructure
new assay formats (e.g., read-out to purify the proteins; Petri dish
of assays); conducting the assays
Molecular biology Cloning and genetic engineering in Cloning genes in bacteria; DNA; gene sequence; “laminar
bacteria or mammalian cells generating “knock-out” animal flow” workstations; database;
model that misses specific genes; cells; bacteria; gene sequencer;
comparative analysis of genomes/ DNA amplifier
gene sequences
Pre-clinical safety and Assessing toxicology and safety Conducting safety and toxicology Lists and slides with safety profile of
toxicology profile of drug compounds experiments with compounds; compounds; experimental set-up
compiling safety profiles of to conduct safety and toxicology
molecules; anticipating patient experiments; data from clinical
population in clinical trials of benchmark compounds
development
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1313

and rich understanding of how knowledge creation (Carlile, 2002), facilitating situated learning processes
is coordinated required the use of purposeful sam- (Tyre & von Hippel, 1997), and influencing how team
pling (Patton, 2002) of informants within drug dis- members mutually adjust their work (Okhuysen &
covery projects with different conditions. Drawing Bechky, 2009). Table 2 summarizes the five projects
on initial interviews and informal discussions with along these dimensions, with examples of core sci-
accessible senior managers, and reviews of the lit- entific questions.
erature on knowledge creation, coordination, and Data collection occurred in real time from spring of
scientific research, we sampled informants from five 2011 until early winter of 2012. As we progressed in
project teams with maximum variation along four our data collection and analysis, we gradually de-
salient conditions: (1) novelty of the target, (2) stage veloped a sense of how project team members co-
of the trajectory, (3) therapeutic area, and (4) number ordinated the knowledge creation process. Building
of research sites. on our emerging insights, we used purposeful sam-
First, novelty of the target refers to the extent to pling to select informants from all hierarchical levels
which the link between a biological target (e.g., the (project team leaders, senior investigators, investigators,
disease-causing gene) and a final medical indication is and research associates) and all specializations
understood at the outset of the drug discovery project. across each of the three drug discovery depart-
A target is novel if little is known about how a drug ments: therapeutic indication (e.g., pre-clinical
molecule “modulates” it (Eder, Sedrani, & Wiesmann, safety), chemistry (e.g., molecular modeling), and
2014).5 Novelty may influence the length of the pro- drug discovery technologies (e.g., structural bi-
cess (Sams-Dodd, 2005) and shape scientists’ ability ology). In so doing, we remained attentive to the
to foresee interdependencies (Carlile, 2004) through consistency of emerging patterns across these di-
both the availability of external information, such as mensions as well as across the five projects. Table 3
relevant reports and peer-reviewed papers (Edwards, gives an overview of data sources.
Isserlin, Bader, Frye, Willson, & Yu, 2011), and the The primary source was 68 semi-structured in-
extent to which scientists benefit from past in-house terviews with scientists and senior managers directly
experience (Bresman, 2013). Second, as the drug involved in drug discovery.6 Our data comprise 24
discovery process typically lasts five to six years interviews with project team leaders/senior man-
(Pammolli et al., 2011), we selected teams conducting agers, 28 with senior investigators (more experienced
research activities at each stage of the drug discovery laboratory heads with PhDs), 12 with investigators
trajectory, from the early identification and validation (laboratory heads with PhDs), and four with research
of the target to later stages focused on identifying, associates operating the lab equipment. We conducted
characterizing, and optimizing compounds involving 63 interviews at the informants’ work sites (in
potential new drug molecules (Sams-Dodd, 2005). Europe and North America) and five by telephone.7
Selecting projects along these different stages also in- Interviews lasted between 60–90 minutes, and all but
creased variance among team members’ experience of two were recorded and transcribed verbatim. For the
working together. Such experience might impact two cases for which recording was not possible (at
how effectively they coordinate drug discovery (Kotha the informants’ request), we took notes and wrote
et al., 2013). detailed reports immediately afterward (Miles &
Third, distinct therapeutic areas achieve different Huberman, 1994).
clinical success rates (Booth, Glassman, & Ma, 2003). We asked these informants to describe their every-
While success rates cannot be reliably predicted, day activities, responsibilities, project team work-
differences could reflect varying levels of uncertainty flow, interactions and interdependencies with other
in the underlying science (Ringel, Tollman, Hersch, & project team members, and the challenges they faced
Schulze, 2013). Fourth, previous studies suggest that in drug discovery (Appendix A shows the interview
coordination effectiveness in geographically co-located protocol). To keep the interviews close to the in-
teams may differ from those in dispersed teams formants’ practice, we used probing questions (Miles
(e.g., Pinto, Pinto, & Prescott, 1993; Srikanth & & Huberman, 1994), asking for an in-depth account of
Puranam, 2011). Number of research sites indicates recent experiences and events, not for reflections on
the co-location of project team members. Co-location vague concepts (Miller, Cardinal, & Glick, 1997).
can shape coordination by enabling a richer dialogue
6
Twelve informants were interviewed twice.
5 7
We assessed novelty of the target by consulting com- An overview of the interview data and observational
pany specialists and analyzing the portfolio of past projects. data are available from the authors upon request.
1314

TABLE 2
Summary of the Five Project Teams
Stage in Drug Discovery
Project Novelty of Therapeutic No. of Research
Team the Target May 2011 Dec. 2012 Area Sites Exemplary Core Scientific Questions

CanPro1 Medium Lead identification Lead optimization Oncology 1 n Designing and synthesizing new compounds with optimized
pharmacokinetics (i.e., stability in human body) and safety
properties
n Establishing biological experiments in cells and in-vivo to
understand efficacy of compounds in living organisms
CanPro2 Very high Target identification Target identification Oncology 3 n Assessing the “druggability” of target (i.e., the potential to
generate compounds that bind with high affinity to that target)
n Analyzing signaling pathway in cells where the target is
involved
CanPro3 High Target validation Terminated (by Oncology 2 n Designing novel approaches to identify new hits (i.e.,
end 2012) molecular chemical molecules that modulate the activity of
the target)
n Validating the roles that the target plays in the development of
which disease states
AutoPro High Lead identification Lead optimization Autoimmune 1 n Optimizing compounds with regards to activity and selectivity
diseases n Identifying “new chemical matter” (i.e., new series of chemical
Academy of Management Journal

molecules that are distinct with regards to their molecular


structure)
InflaPro Very high Assay development Lead identification Inflammatory 1 n Confining “hit lists” generated through high-throughput
diseases screening (i.e., what are the compounds identified as “hits”
the team wants to take any further?)
n Establishing an in-vivo model (i.e., animal model) linking the
target with the disease molecule

Note: Stage in drug discovery follows Sams-Dodd’s (2005) classification: Target identification: identifying the disease-associated gene or gene product; Target validation:
determining the therapeutic value of the target in the therapeutic indication; Assay development: expressing the target in biochemical or cellular assay systems; Lead identification:
screening and selecting compounds for further optimization; Lead optimization: optimizing compounds for target affinity and selectivity.
August
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1315

TABLE 3
Data Sources Across Projects
Project Team

Data Source CanPro1 CanPro2 CanPro3 AutoPro InflaPro Othera Total

Interviews Therapeutic indication specialists 6 6 4 1 7 1 25


(e.g., pre-clinical safety, in-vivo biology)
Chemistry specialists (e.g., molecular 5 4 1 6 8 1 25
modeling, medicinal chemistry)
Drug discovery technology specialists 5 2 5 3 2 1 18
(e.g., structural biology, high-throughput
screening)
Subtotal 16 12 10 10 17 3 68
Observations Project-team meetings 6 2 3 4 7 – 22
Sub-team meetings 5 2 2 2 5 – 16
Specialists’ informal discussions and 3 4 3 3 4 – 17
everyday lab work
Subtotal 14 8 8 9 16 – 55
Secondary Project team meeting slides; entries in project management database; research organization’s intranet;
data project team meeting minutes; scientific publications on disease background, target, and drug
candidates.

a
Interviews with project team leaders prior to the selection of the five drug discovery projects.

Our data also included non-participant observa- and to disentangle how scientists coordinated work
tions of 55 instances of interactions among project across knowledge domains, we initially engaged in
team members during project team meetings where open coding of raw interview data (Strauss & Corbin,
all specialist members—and occasionally other 1998). To contextualize the interview data, we used
stakeholders, such as senior managers—of a drug field notes from observations and secondary data.
discovery project unite, sub-team meetings of se- Data analysis occurred in real time, starting when
lected project team members from various knowl- transcripts and field notes became available. Two
edge domains, and specialists’ informal discussions authors coded the interview transcripts until the
and everyday bench work within laboratory envi- analysis stopped yielding sufficiently distinct first-
ronments. Due to confidentiality, we could not re- order categories. Each first-order category was labeled
cord these meetings. We took extensive field notes consistently with informants’ terminology (e.g., first-
on site to capture as much as possible of the con- order category “branching out into sub-teams” corre-
versation verbatim, and enriched these notes with sponds to interview excerpt “there is a scientific
further contextual information after the observation question and we need to sit down and talk about it. So
(Miles & Huberman, 1994). Moreover, we collected why don’t we call a sub-team meeting to get those
archival data to contextualize our interviews. Data people that are relevant together and hear everybody’s
included scientific publications, intranet entries, input?”) (Gioia, Corley, & Hamilton, 2012). If coding
documents from the online project management labels conflicted, we discussed and crosschecked
platform, and minutes of meetings. emerging codes to ensure that the data were particular
to a given code and did not become decontextualized.
Second, we used axial coding (Strauss & Corbin,
Data Analysis
1998) to identify similarities and differences in
The combined interview and observation data the first-order categories, and we aggregated corre-
consisted of 930 pages of single-spaced interview sponding categories into second-order themes, giv-
transcripts (over 87 hours of interview recordings) ing them higher-order theoretical labels (e.g.,
and 167 pages of field notes from observations. We “structuring around conjectural interdependencies”).
managed our data using NVivo 9, a software tool for As the research design is aimed at elaborating theory,
qualitative data. we repeatedly consulted the coordination literature to
Data analysis was done in three stages. First, to re- help us interpret the findings in the light of prior work.
construct project teams’ workflow of everyday activities In so doing, we aggregated second-order themes into
1316 Academy of Management Journal August

higher-order theoretical dimensions (Gioia et al., 2012). project team leader in biology (project AutoPro)
This second stage produced a data structure with three described the formation of a drug discovery project
aggregate theoretical dimensions. Figure 1 summarizes as follows:
this structure.
You have an idea for exploratory work to ensure the
Third, we revisited the full data set in search of
target is druggable and. . .that it is really important. . . .
emerging patterns and relationships between the
And then the complexity increases. . .you have to start
themes and theoretical dimensions. As we pro- working with other people, developing assays, run-
gressed toward a deeper understanding of these ning a screen, and setting up a flow chart. But the
patterns and relationships, we developed a preliminary target validation does not really stop at a certain point
model of how knowledge creation was coordi- because we always need to evaluate how good the
nated in the project teams. Finally, to lend increased target is in comparison to other targets. What are its
credibility to our interpretations, we discussed our advantages? Does this target have certain safety as-
emerging model with several key informants within pects? Is it better than other targets in efficacy? What
DrugCo (Miles & Huberman, 1994; Nag, Corley, & are the indications that we want to focus on? You re-
Gioia, 2007). ally have to develop a clear understanding of which
disease you want to target.
FINDINGS This excerpt illustrates that when initial evidence
looks promising, the individual scientist’s explo-
Knowledge creation at DrugCo involves coordina-
ration develops into a more complex project—a
tion of specialists’ work at the project team level and
team-based undertaking comprising specialists from
within emerging sub-teams. Following a brief de-
complementary knowledge domains within the
scription of a drug discovery project’s inception, we
company. The initial individual exploration thus
first explain the two-tier structure within which
leads to core scientific questions, each of which
project team members assume different modes of
foregrounds some missing scientific evidence on
specialist work, either as sub-team “insiders” or sub-
proposed cause-and-effect relationships between
team “outsiders,” each with its distinct responsibilities
the molecular structure of compounds, the activity
for contributing to the knowledge creation process.
of the target, or the disease state in living organ-
We next discuss three informal practices that enabled
isms. These questions underlie the knowledge that
specialists within these sub-teams to manage in-
scientists need to create for the project to progress
terdependencies among their respective knowledge
toward clinical trials and the commercialization of
domains (i.e., informal coordination practices within
a drug.
sub-teams) and progress in knowledge creation. We
The excerpt also shows that a compartmental-
then explain how specialists create formal sub-team
ized approach to knowledge creation is insuf-
structures within the confines of their project team
ficient for resolving core scientific questions:
and dynamically restructure these sub-teams accord-
The expertise for creating knowledge around
ing to conjectural and revealed interdependencies
core scientific questions is situated within mul-
(i.e., formal structuring around interdependencies). In
tiple knowledge domains. As a senior investigator
so doing, project teams accommodated the unpre-
in molecular modeling (project InflaPro) noted,
dictable and continuously changing need for contri-
“With ‘pure’ modeling we are not able to solve any
butions from different knowledge domains as the
big problems in drug discovery. We must carefully
process unfolded. Table 4 provides detailed informa-
integrate our contributions with those from struc-
tion on aggregate dimensions, second-order (emer-
tural biology—the knowledge we have about structure-
gent) themes and their definitions, and additional
activity-relationships, 8 chemical synthesis, and
examples of first-order data.
pharmacology.” Advances on core scientific ques-
tions thus require project teams to coordinate in-
Project Inception terdependencies among specialists from various
knowledge domains.
Drug discovery projects originate when senior
scientists conduct individual laboratory experi-
ments to identify a disease-associated gene, protein, 8
Structure-activity-relationships (SARs) describe the
or signaling pathway in cells (i.e., a target). In this causal linkages between the chemical structure of a drug-
inception phase, specialists investigate, for exam- like molecule (“structure”) and its biological activity on
ple, how the target behaves in the disease state. One the target (“activity”).
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1317

FIGURE 1
Overview of Data Structure

Specialist Work structural biology (project InflaPro) notes, “The


project team meeting is rather formal. Experts that
DrugCo uses a two-tier structure to coordinate the
don’t know particularly much about each other’s
drug discovery process, with project teams forming
work come together [here]. These meetings are
the first tier. Project teams disaggregate the drug
more about information exchange and discussing
discovery process by delegating the responsibility
what to do next at a very high and strategic level.”
for answering core scientific questions to a second
Most in-depth day-to-day collaboration among
tier of formally structured, self-managed sub-teams
specialists unfolds at the sub-team level. The ad-
consisting of specialists from within the confines
vantage of the sub-team structure is that it signifi-
of the project team. 9 As a senior investigator in
cantly reduces the perceived complexity of the
process. As the same senior investigator adds,
9
Each project team includes between three to five sub- “There are so many smaller sub-teams where the
teams at any time. Project teams on average include ca. 50 actual work gets done. . .they do certain experi-
members, with about 20–30 members attending project ments I’m not aware of, I’m simply not on their
team meetings (laboratory technicians do not attend). Sub- mailing list. It would be too much information if I
teams on average include 4–6 members (laboratory heads). were included there as well.”
1318
TABLE 4
Aggregate Dimensions, Second-Order Themes, Definitions, and First-Order Categories
Aggregate Dimension Second-Order Theme Definition First-Order Categories with Examples of Interview Excerpts

Specialist Work Specialist Work of Project team members’ c Bench work in laboratory environment: “My contribution. . .concerns the
Sub-Team Insiders contributions to a sub-team strategy with respect to protein production, crystallization, selection of
to which they belong. what compounds to be crystallized. . . . All of that has to do with my
scientific expertise.” (CanPro1, senior investigator, structural biology)
c Modeling and computational analysis: “I conduct computational
investigation to predict a number of properties of compounds, such as
solubility, stability, or degradation [in the human body]. That concerns
modeling the interactions [between the compound and target] and also
includes statistical analyses.” (InflaPro, senior investigator, structural
biology)
c Updating the project team on findings and problems: “. . .you try to update
all the others [in the project team] on what is going on, which doesn’t
mean there’s no discussion. . . .” (InflaPro, project team leader, medicinal
chemistry)
Specialist Work of Project team members’ c Questioning everyday work: “[In sub-teams] you are quite often doing
Sub-Team Outsiders contributions to a sub-team things in the usual way. Sometimes it’s quite beneficial to get challenged
to which they do not belong. and to rethink the processes.” (InflaPro, project team leader, medicinal
chemistry)
c Suggesting new approaches: “No one [outside our specialization] will tell
you: ‘Okay, you should make this type of modification because it will be
super active [in the animal model].’ Rather, [these outsiders] will be there
to help you to design or to guide you. . . . I think that’s very useful.”
(InflaPro, project team leader, medicinal chemistry)
Informal Coordination Anticipatory Conforming Specialists’ efforts to understand c Orienting domain-specific activities toward cross-domain contributions:
Practices Within the implications of their work “I consider myself a specialist in crystallography with the goal of being
Academy of Management Journal

Sub-Teams for other specialists, and if a drug discovery scientist. My activities in crystallography should be
necessary, compromise their valuable to the discovery of new drugs; not the activities themselves
domain-specific standards of should be at the center stage but how they contribute to answering
excellence to meet cross-domain scientific questions.” (InflaPro, senior investigator, molecular modeling)
requirements. c Compromising domain-specific standards of excellence to meet cross-
domain requirements: “A modeler can have brilliant ideas, but if
[medicinal chemists] are not synthesizing. . .it is useless. Sometimes the
molecular modeler designs the molecule in a way that is difficult to
synthesize. . . . After a discussion with the medicinal chemist or input
from a chemist we arrive at the molecule which is really visible in terms
of synthesis. . . you have to trigger this dialogue.” (CanPro1, senior
investigator, molecular modeling)
Workflow Synchronizing Specialists’ efforts to understand c Understanding temporal interdependencies: “To develop the
temporal interdependencies, experiments I have to wait for the protein. I can’t do without. Similarly,
and to order and pace their the specialist who makes the protein. . .needs to know whether he is
work accordingly. purifying the right thing, whether he loses activity.” (Senior manager,
screening technology)
August
TABLE 4
2016
(Continued)
Aggregate Dimension Second-Order Theme Definition First-Order Categories with Examples of Interview Excerpts

c Ordering and pacing specialist contributions: “Normally, the technology


guys take care of many projects in parallel. The structural biologist, for
example, could work on four to five projects in parallel. But his input is
very critical. . . . They just ask: ‘how long you can wait [for the results]?
Two weeks? One month?’ I then say: ‘For us this is critical, but depends
on you guys’ scheduling. At most we can wait one to two months.’ They
then try to manage it somehow.” (CanPro2, senior investigator,
medicinal chemistry)
Cross-Domain Triangulating Specialists’ efforts to establish c Aligning experimental conditions across specializations: “As I designed
the reliability of domain- my experiments, I was sitting together with the others who designed
specific findings. [complementary experiments] before. We had to ensure that we aligned
several parameters and used the same experimental conditions. For
example, we had to make sure that we used the same substrate [for all
experiments].” (CanPro3, research associate, screening technology)
c Confirming domain-specific results: “I have seen many projects where in-
vivo data [generated by biologists based on cells or living organism] was
not fitting at all with in-vitro data [generated by biochemists based on
biochemical experiments with isolated target]. In our case, I would say,
for [one chemical series] we have a perfect match. In-vivo and in-vitro is
just a perfect match. I think that is the first time in the past ten years I have
seen that. It is very good.” (InflaPro, project team leader, medicinal
chemistry)
Formal Structuring Structuring Around Designing structural arrangements c Deciding which knowledge domains are relevant for answering core
Around Conjectural (i.e., sub-teams) that include scientific questions: “We also had sub-team meetings during the time we
Interdependencies Interdependencies knowledge domains considered worked toward establishing the [new screening experiments]. To do so,
relevant and responsible for we included two medicinal chemists, the biochemistry lab, and the
knowledge creation. structural biologist.” (InflaPro, project team leader, medicinal chemistry)
c Branching out into sub-teams: “There is a scientific question and we need
to sit down and talk about it. So why don’t we call a sub-team meeting to
Ben-Menahem, von Krogh, Erden, and Schneider

get those people that are relevant together and hear everybody’s input?”
(CanPro3, senior investigator, biochemistry)
Restructuring Around Redesigning structural c Including new knowledge domains: “Once we have the structure [of the
Revealed arrangements target] and cannot explain it, we might also have to include [the structural
Interdependencies (i.e., sub-teams) by including biologist] in our sub-team. We then have to work together, coordinate
new efforts, and understand that piece of data.” (CanPro3, senior investigator,
knowledge domains or biochemistry)
excluding c Excluding irrelevant knowledge domains: “In the future, for example,
irrelevant knowledge-domains. biophysics may not be that important anymore. Now we understand how
the molecules bind and we probably won’t go back to finding completely
new compounds. . .” (Senior manager, medicinal chemistry)
1319
1320 Academy of Management Journal August

A core feature of the sub-team structure is that it Specialist work of sub-team insiders also involves
allows specialists to assume different responsibilities members’ responsibility for reporting their findings
in the knowledge creation process. For each core in monthly project team meetings. As a project team
scientific question, the mode of specialist work of leader in medicinal chemistry (project AutoPro) ex-
a project team member is guided by whether the for- plains, “project team meetings are about getting ev-
mal sub-team structure designates him or her as a sub- erybody on the same page and documenting progress.
team insider (i.e., a member of a particular sub-team However, to be honest, the real work is not done in
embedded within the larger project team) or a sub- these meetings. It’s done on an everyday basis.”
team outsider (i.e., a member of the larger project team Updating allows the project team (both sub-team in-
but not of that particular sub-team). siders and outsiders) to jointly discuss how to proceed
Specialist work of sub-team insiders. The spe- with new results, and offers an opportunity for sub-
cialist work of sub-team insiders refers to project team team insiders to receive potentially valuable input on
members’ contributions to a sub-team to which they their work. As one senior investigator in biology
(formally) belong, and involves day-to-day knowl- (project CanPro2) comments:
edge creation activities and routine project team up-
In project team meetings all the different specializations
dates in project team meetings. Specialists’ isolated meet again. Thus it could well happen that an outsider
work (e.g., designing, modeling, conducting, and to your domain has a completely different idea. He looks
interpreting experiment results) is the primary source at an experiment without having conducted it thou-
of input to knowledge creation within sub-teams. As sands of times before. In sub-teams we sometimes make
a senior investigator in medicinal chemistry (project decisions because we did so in the past. Having some-
AutoPro) describes this work: body from the outside increases the chances of seeing
No doubt. . .discussions [with the modeler] are of things differently, and it might result in cross-fertilizing.
pivotal importance for medicinal chemists. Yet at one In the sub-team we then deal again with solving the
point. . .we have to give it a try. We have to go back in issue, given our specialization in that field.
the lab, cook,10 and test the compound. We have to Specialist work of sub-team outsiders. The spe-
generate results. After that, we can discuss things cialist work of sub-team outsiders constitutes the
again. I think that it’s very important. . .that we don’t contributions from project team members to a sub-
neglect the true chemistry activities. We still use most team to which they do not belong. Although spe-
of our time to actually cook the compounds. cialists outside of the sub-team do not formally share
To integrate the findings from individual experi- the sub-team’s responsibilities, they remain pivotal
ments conducted in different knowledge domains, to knowledge creation by questioning sub-team
specialists share their results with other sub-team members’ assumptions underlying experimental
members and jointly decide on further activities. A designs and stimulating them to consider new ap-
project team leader in biology (CanPro2) comments: proaches. As a senior investigator in medicinal
“In sub-teams we really discuss issues in-depth. It’s chemistry (project AutoPro) explains:
not about overview or strategy. . .it’s about science Having an outside view typically helps. If you are
and details.” A senior investigator in biochemistry “inside” you are frequently overwhelmed with details
(project CanPro2) points out that the sub-team struc- and do not see the forest for the trees. Sometimes,
ture offers efficient knowledge exchange among those these questions from outside—which at first sight
specialists deemed necessary for resolving core sci- might appear naive but sometimes really target the
entific questions: very core of the problem—are helpful. [Outsiders]
might feel they don’t contribute something relevant,
I personally like these very focused meetings where
but these contributions are usually highly valued.
only those people involved in that scientific question
are there. [In sub-teams] we try to move things. We go The following excerpt from a team meeting of
straight to the point. The other day we had a meeting project AutoPro illustrates such an exchange. A
on protein purification. One hour later we knew ex- member of the “hit list sub-team” 11 updated the
actly what the others were doing, they knew what we project team on the idea of expanding screening
were doing, and we made some decisions for moving
forward. 11
A “hit” is a compound showing significant activity on
the disease target in the high-throughput screen, identified
10
Slang for “create a chemical reaction.” as “hitting” the target.
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1321

efforts to new molecules. Sven, a structural biologist Our analysis identified three informal coordinating
outside the sub-team, observed: practices that specialists use to manage cross-domain
interdependencies in sub-teams: anticipatory con-
Sven: “Maybe it doesn’t fit here, but I remember having
forming, workflow synchronizing, and cross-domain
worked on a propeller-like compound we once identi-
fied in our experiments but never followed up on. Don’t
triangulating (see Table 5 for examples). While spe-
you think this could be an entry point for you guys?” cialists performed diverse work (e.g., modeling and
experimenting), used specialized jargon, and adhered
Rob (project leader): “Nice new input, you [the sub- to values and standards of excellence unique to their
team] should definitively take Sven’s comment into domains, these informal coordination practices were
consideration. I suggest that you meet in the sub-team consistent across levels, domains, and projects.
again and come up with appropriate decisions.” Anticipatory conforming. Anticipatory con-
Such interventions by outsiders can prove valu- forming refers to specialists’ efforts to understand
able even in cases when they do not hold in-depth their interdependence with other knowledge domains
knowledge about the sub-team’s work. An investi- in terms of the implications of their work for that of
gator in medicinal chemistry (project CanPro2) ex- the specialists, and if necessary, compromise their
plains: “There are [scientific questions] that are domain-specific standards of excellence to meet cross-
outside my field. . .. [But] even if it’s not your field, domain requirements. A senior investigator in high-
you have to ask questions, you can try to understand. throughput screening (project AutoPro) explains:
Even a naive question can trigger new thinking in We have to design our assays such that they support
a scientist who knows his field in-depth. There are the medicinal chemists’ decision-making in the best
never stupid questions.” possible ways. For instance, we first go for binding
As a senior investigator in medicinal chemistry assays with the inactive enzymes. These offer the best
(project InflaPro) points out, incorporating suggestions possible starting point for finding binders of a partic-
from sub-team outsiders is ultimately voluntary: “I ular class—even if we, the assay developers, would
think [an experienced biologist] would be able to sug- prefer working with the active enzyme from the be-
gest things to [the medicinal chemists]. It might well be ginning. The latter would be more interesting for us.
that the chemists didn’t think about it. The chemists, Also, I always have to think about whether the hits we
however, then decide whether or not they would like to identify are interesting from a medicinal chemistry
consider that suggestion and go forward with it.” perspective. For instance, a compound might involve
a very challenging synthesis. These thoughts are very
important for moving on with the problem of finding
Informal Coordination Practices Within
new potential drug-like compounds.
Sub-Teams
The following excerpt exemplifies this comment.
Within sub-teams, specialists need to manage their
Two sub-team members of project InflaPro—Alex,
cross-domain interdependencies in day-to-day activ-
a structural biologist, and Chris, a medicinal chemist—
ities. Managing interdependence is paramount—yet
discussed which of two selection criteria Chris could
inherently demanding—for answering core scientific
use to reduce the hit list: the compound’s “IC50”
questions assigned to the sub-team. As a senior in-
values or the medicinal chemist’s more subjective
vestigator in medicinal chemistry (project AutoPro)
assessment of “chemical attractiveness”:
explains:
Alex: “I don’t get what chemical attractiveness means.
Clearly, most challenges we face are beyond pure
That’s very fuzzy to me.”
chemistry. I would label [such problems] as true drug
discovery activities. . .they are by far the most chal- Chris: “Well, it’s indeed fuzzy. It’s basically the me-
lenging ones. Consider, for example, how to optimize dicinal chemist’s gut feeling. . .hard to put in words or
chemical compounds according to the data we get numbers. It’s whether or not we like a molecule—for
from pharmacokinetics, in-vivo activity, stability, or a variety of reasons.”
other assays. As soon as you improve one parameter,
Alex: “Yes, for a variety of reasons, precisely. That’s
the others drop immediately. It’s like packing a suit-
hard to understand for non-chemists. I would go for
case that is already jam-packed: As we try to close one
IC50 to select the next round of compounds.”
lock, the others for sure won’t be lockable any-
more. . .and nobody knows why [laughter]. These are Chris: “Are you sure? I would go for chemical
the main problems we face. attractiveness.”
1322 Academy of Management Journal August

TABLE 5
Informal Coordinating Practices Within Sub-Teams: Interview Excerpts
Anticipatory Conforming

“Being a good medicinal chemist is simply not enough. It’s not sufficient to plan and cook interesting compounds. It requires much more
involvement with the sub-team. Fitting my knowledge contribution to the other disciplines is precisely what makes this job so exciting.”
(CanPro1, project team leader, medicinal chemistry)
“We have to think about what a ‘good’ compound really means. A good compound actually is not the one looking nicely on my screen and being
highly active. It’s one that still is active and medicinal chemists can actually synthesize.” (AutoPro, investigator, molecular modeling)
“There is absolutely no point in modeling a compound when you know that the chemist won’t it take up. For instance, I worked in a team where
the chemists were clear about not including any Alpha atoms in the molecules—even if it was quite popular at that time. So I of course did not
include any Alpha atoms in the molecules I modeled. I have to accommodate to the preferences of others in the sub-team. . .they are also
involved in answering that scientific question.” (InflaPro, senior investigator, molecular modeling)

Workflow Synchronizing

“We would like to study the pharmacological profile of some compounds. This implies that the pharmacologists should not be waiting with the
animals. We have to ensure that chemistry produces enough compounds so that they can properly scale up production. Similarly, the protein
production specialist must deliver in time to the crystallographer so that he can then start determining the three-dimensional shape of the
target protein.” (InflaPro, senior investigator, molecular biology)
“I talk to the crystallographers, and to the person doing nuclear magnetic resonance, so I know what they are expecting of me in the near future.”
(CanPro2, research associate, structural biology)
“We have a close interaction [with the assay development laboratory] because when we finish our compound the next step takes place in [this]
laboratory.” (AutoPro, research associate, medicinal chemistry)

Cross-Domain Triangulating

“It would be a dream scenario if we would find a hit in our biochemical screen that is also independently found to be a hit in the cellular screens.
That would more or less be the validation that the hit we have found also shows cellular activity. However, chances are pretty slim. We are
talking about two experimental set-ups with different sensitivities. Of course . . .essentially we design the assays such that they should inform
each other. In fact, the goal is to get hits coming from the cellular assays and to test them in the biochemical assay and vice versa.” (CanPro2,
project team leader, biochemistry)
“In case the team wants to know [how the drug molecule binds to the target], they send it to me. I perform the crystallization and might confirm
that what the modeler and the chemist predicted was perfectly accurate. Yet there are also situations where this is not the case...” (CanPro3,
investigator, structural biology)
“The goal is to make the validation [of the target] scientifically more sound. . . . Over time, the goal is to gain confidence that what you are doing
really has an effect at the end and delivers value to the patient.” (AutoPro, senior investigator, medicinal chemistry)

Alex: “But why?” effective resolution of the sub-team’s core scientific


question (how to reduce the hit list).
Chris: “We could go for IC50 as well. That’s very fine
with me—anyway, from a chemistry point of view we
Conversely, when sub-team members are un-
would prefer IC50. It’s much easier to justify. . . . aware of or disregard the requirements of other sub-
However, do you really want to crystallize yet another team members, progress on core scientific questions
benzyl amide?” [overall laughter]12 is at risk. For example, in a sub-team meeting of
project InflaPro, molecular biologist Adam pre-
This example illustrates specialists’ vigilance sented the results of his analysis to structural bi-
about the requirements and possible challenges of ologist Mike and biochemist Roger (both relying on
other sub-team members’ work. The medicinal Adam’s findings):
chemist preferred selecting compounds based on the
“IC50” criterion. However, anticipating that apply- Adam: “You can say that the [amino acid X] corre-
ing this criterion would complicate the work of the sponds to [type A] and [amino acid Y] corresponds to
structural biologist (who was unfamiliar with that the [type B]. We could use that for differentiation.”
criterion), the medicinal chemist “relaxed” his Mike: “I’m not sure about that.”
domain-specific standards of excellence to allow an
Roger: “No, Adam is right.”
12
Compounds belonging to the structural class of “ben- [Adam continues explaining how to differentiate be-
zyl amides” were known for being tough nuts to crack. tween the two amino acids.]
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1323

Mike: “You forget that this site is masked. We cannot people have the right information and understand
use it.” what is expected of them.

Adam: “Ah, OK, I didn’t know that. I just looked at the Cross-domain discussions on temporal interde-
sequence.” pendencies aid specialists in providing meaningful
contributions to the sub-team. In the following ex-
Roger: “So what can we conclude with regards to
species differences? What species would you recom-
cerpt (project InflaPro), in-vivo biologist Tim dis-
mend for toxicity studies?” cusses his planning with fellow sub-team members:

Adam: “That’s why you wanted to know about spe- Tim: “I might start working on developing new bio-
cies differences. I don’t know. I didn’t look into markers in mice now. I have some free resources in the
that. . . .” lab.”

Roger: “So the conclusion is that we first have to know Linda: “We don’t know yet whether we really need
in depth the binding mode and then select the these biomarkers in mice or humans. We have to wait
species.” for others’ input here.”

[As Mike explains his requirements, Roger suggests Ralph: “We need to act according to the project pri-
ending this discussion.] orities and think about that at a later stage. . .”

Roger: “In the interest of time, could we do that off- By pointing out the temporal interdependence
line? Because, besides you two, nobody can follow at between Tim’s and other specialists’ work, his col-
the moment.” [overall laughter] leagues ensured the timeliness of Tim’s contribution
by deferring his intended activities.
Adam: “No, but it’s important. Now I’m starting to Cross-domain triangulating. Given the precau-
understand what you really want.” tionary principle in drug discovery, the value of
During Adam’s updating, it became apparent that knowledge created within the bounds of specialists’
he did not know all cross-domain requirements knowledge domains derives from the degree to
when conducting his experiments—and thus could which specialists consider such knowledge reliable.
not contribute to the core scientific question. In establishing the reliability of their contribu-
Workflow synchronizing. Workflow synchroniz- tions, specialists are dependent on other knowledge
ing refers to specialists’ efforts to understand tem- domains. Cross-domain triangulating refers to spe-
poral interdependencies, and order and pace their cialists’ efforts to establish the reliability of domain-
work accordingly. A project team leader in medici- specific findings. The following statement from
nal chemistry (project CanPro3) explains: a senior investigator in molecular modeling (project
CanPro1) illustrates how the uncertainty in the
The pharmacologist tells me that he needs to grow characteristics of novel compounds and their bi-
some tumor in mice and that this will take him three ological effects necessitates such triangulation:
weeks after injecting the cancer cells into the mouse.
The tumor then starts to grow, and in three weeks the In biochemical assays you have many artifacts. . .
mice are ready for treatment with the compound. If some of the compounds that we are screening are
you wait for more than three weeks, the mouse dies. not pure, some are doing strange things to proteins.
So you have to have the compounds ready three Especially when you don’t know anything about
weeks from now. That’s essentially what we dis- the target—when it is a really new target—you
cuss. . .whether I am ready with the compounds then. always wonder: Is this something real? The
modeler—if he knows the pocket and its target very
Another project team leader in biology (project well—can really make a connection with the struc-
AutoPro) further explains that, to prioritize their ture. . .because he can see what it is like in 3D and
work, specialists need to understand each other’s check on the computer whether your idea was right
needs: or not.

When we do multi-disciplinary activities. . .people Specialists use cross-domain triangulating to


have like ten different things they need to achieve scrutinize the findings and assumptions in their own
every day. . .and so what might be my priority may not work. In so doing, they either gain confidence that
be the priority for the colleague, and vice versa. That’s their output can feed into later stages of the pro-
why. . .I think it’s always complex, but communica- cess or become aware of inconsistent findings that
tion is a big one. . .just really ensuring that the right necessitate more work. In describing an in-vitro
1324 Academy of Management Journal August

experiment in a mouse model, a senior investigator domains are most optimal for moving the drug dis-
in molecular biology (project InflaPro) explains how covery process forward. Formal structuring around
cross-triangulation involves close interaction be- interdependencies refers to the process by which
tween sub-team members: project team members collectively decide which of
the specialized knowledge domains from within
At the very beginning, it was really actually going
back and forth. . .we had many discussions because
the project team’s confines are relevant and inter-
when I come up with a question, [the X-ray specialist] dependent for resolving a particular core scientific
looks into a sequence and comes back to me to make question, and then structure ad hoc sub-teams by
sure—when he starts to look into the structure— selecting and grouping specialists accordingly. Proj-
whether he understood the problem correctly, you ect teams thus continuously (re-)establish the grounds
know, that he has a structure in front of his eyes. Then for substantive interactions in sub-teams.
we discussed. . . . We looked at the structure together Structuring around conjectural interdependencies.
and he explained to me. . .[and] also [asked me] what I Decisions to include specialists in sub-teams are
think that this amino acid could be. . . . Of course at the primarily driven by project team members’ collec-
end [the X-ray specialist] is the one solving the tive appraisal of the need for expertise with respect
structure, and he is the expert in the field, but it was to specific core scientific questions. The following
really going back and forth, and questions came excerpt illustrates the formation of a sub-team in
iteratively. project CanPro1: Bill, a senior investigator in high-
Finally, triangulating findings across specialized throughput screening, updated the project team
domains involves sensitivity to the pragmatics of (including project leaders Ann and Ryan) on his
specialists’ language and dedicated efforts to resolve insights from a large screening of 1.5 million mol-
misunderstandings. As a project team leader in me- ecules, which resulted in the identification of
dicinal chemistry (project InflaPro) comments: roughly 25,000 compound hits. For practical use in
future experiments, the hit list must be reduced to
Between chemists and biology we have to have the fewer than 5,000 compounds:
same platform of discussion, of language, just to be
sure that they can exactly understand what I want. Ann: “Should we start anticipating the criteria for
Sometimes, as chemists, we are using biological restricting the hit list to 5,000 compounds now?”
words in the wrong sense. For example, last time [the
Ryan: “Yes, we need a sub-team here to come up with
biologist] said this compound was “nicely absorbed,”
some criteria for decision-making.”
and now she says that the compound was “badly
absorbed.” For me it is nicely absorbed when it is Ann: “Ryan and I should probably think about who
metabolized. [She] said: “No, for me ‘nicely absorbed’ should be included. . . .”
means that you retrieve the parent compound in the
Bill: “We need to have a chemist join that sub-team.
blood after some time, it does not mean that it is
Different chemists think about the same problem in
absorbed and then you have metabolism. That is not
different ways. Here we need different viewpoints.”
the same.” So you have this small discrepancy in
language between specializations and have to be sure As the need for expertise is contingent on the
you are on the same page. largely unknown nature of novel scientific problems
By engaging in a discussion on the domain- and unpredictable outcomes of sub-team specialists’
specific meaning of “nicely absorbed,” the chemist knowledge creation efforts, structuring around in-
and the biologist could prevent that the chemist terdependencies among knowledge domains is es-
wrongly interpreted the biologist’s findings. As this sentially a conjectural undertaking—as specialists
excerpt shows, triangulating with sensitivity to the are well aware. Asked how cross-domain interac-
particulars of their counterpart’s knowledge domain tions emerge, a project team leader in medicinal
is a key requirement for effective cross-disciplinary chemistry (project AutoPro) explains:
interactions. It is very important to realize that compared to other
industries, where the process is really well defined, in
our business this is not the case. I like to compare it
Formal Structuring Around Interdependencies
with [a local newspaper] we visited a few years ago.
New core scientific questions and progress on There it was very obvious that [while] of course they
existing questions prompt project teams to continu- don’t know what they will write [about] the next
ously consider which combinations of knowledge day. . .it’s very certain that the newspaper will be
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1325

published the next morning, no matter what is in Cam: “Were rats really the best species to look at?
there. . . . So everything has its well-defined stages: At Mice could be similar to humans.”
this time of the day you have to have this, and then it
Lucy: “We assumed rats to be similar to mice and thus
goes to press, and at night it gets distributed, and the
to humans. I wouldn’t have expected any differences
next day you always have your newspaper. And for us
between rodents.”
it is basically: [In] the morning you come. . .you think
you have a brilliant idea, you try to do an experiment, Cam: “Probably there is a difference. We could have
and in the evening or maybe days later you realize that a look at it, if you want us to.”
this was not the right approach. . .you cannot plan for
this. [Y]ou try the first approach, and realize that this Lucy: “Very good point. We definitely should test
is not the way it works, so you have to look for alter- whether a rat responds similar to a mouse. Could you
natives. . . . You ask around, you contact different do that?”
people. . . . There’s no right answer, you get different Exercising their specialist work responsibilities
opinions...and go in one direction or the other. So it’s in this way—Lucy as a sub-team insider and Cam as
not that one exactly knows what comes next. a sub-team outsider—thus revealed an unpredicted
As this excerpt shows, the uncertain nature of the interdependence between their respective knowl-
drug discovery process imposes limits on specialists’ edge domains. To answer its core scientific question,
ability to predict which knowledge domains become Lucy’s sub-team now appeared to rely on Cam’s do-
relevant at what time. main specific knowledge on how to test the poten-
Structuring around conjectural interdependencies tially dissimilar response of rodents. This newly
was a consistent pattern of activities whereby spe- revealed interdependence triggered the restructur-
cialists across DrugCo’s project teams formally co- ing of the sub-team by including Cam in Lucy’s sub-
ordinate knowledge creation. Figure 2A shows how team after the meeting.
an initial conjecture about interdependencies among Table 6 shows examples of obstacles that were
specialized knowledge domains relates to sub-team resolved through sub-team restructuring around
formation. In this example, a sub-team consisting of revealed interdependencies, and implications of
specialists in medicinal chemistry, biochemistry, restructuring for knowledge creation. As the project
and in-vivo biology is formed within the broader CanPro3 example shows, unpredicted interdepen-
project team, which also includes specialists in mo- dencies even appeared during standard procedures.
lecular modeling, pharmacology, etc. Figure 2B illustrates a restructured sub-team (com-
Restructuring around revealed interdependencies. pared to the initial structure in Figure 2A)—with
Sub-teams are formed with the intention of combining molecular modeling now included but in-vivo bi-
those specializations that can jointly create knowl- ology excluded.
edge until a core scientific question is answered. In Restructuring sub-teams during the project team
practice, however, specialists often need to re- meetings was critical for overcoming obstacles in
structure the initial sub-team when new findings or drug discovery. The decision to restructure could be
obstacles reveal interdependencies with knowledge initiated and influenced both by sub-team insiders
domains not yet represented. As a senior manager and sub-team outsiders, and depended on the extent
observes: “[S]ome of the team members become less to which project team members could recognize
important, some become more important, some that the problems with the current sub-team structure,
are less important now can become more important at and whether the members agree to move to change
a later stage, so there is a coming in and going out.” the structure. Yet the ability to restructure stood in
While specialists are mindful of the changing impor- constant tension with the potential hazard that sub-
tance of knowledge domains throughout the drug dis- team members would tenaciously search for solu-
covery process, interdependencies among specialists
tions within the safety of their current configuration.
arise unpredictably.
In the following excerpt (project CanPro1), medici-
For example, in the following excerpt from project
nal chemist Helen and toxicologist Zoe update the
CanPro1, the sub-team was using a rat model to study the
project team on their sub-team’s difficulties with
potential interference of the target with a human meta-
understanding why important compounds test pos-
bolic reaction. While sub-team member Lucy, an in-vivo
itive on toxicity:
biologist, proposed an experimental design to the project
team, Cam, a pre-clinical safety specialist outside her Leon (project team leader): “So your recommendation
sub-team (but on the project team), commented: would be to optimize the compounds based on other
1326 Academy of Management Journal August

FIGURE 2
Formal Structuring Around Interdependencies

parameters and then get back to the [toxicity test] cross-domain triangulating). Given these findings,
again?” we propose a model of how formal coordination
Helen: “It’s really important data. If the compound
structures and informal coordination practices in-
really is positive here all my options are gone. I want terweave in multidisciplinary self-managed teams
to give it another try [to optimize the compound].” creating knowledge under unpredictable interde-
pendence (see Figure 3). First, we explain how the
Zoe: “We [toxicology specialists] can take that up formal structure establishes a foundation for spe-
offline in a meeting with the chemists.” cialist contributions to knowledge creation and fos-
Paul (senior manager): “I don’t think we should let the ters the emergence of informal coordination practices
chemists further work on that. The risk is too high. that enable interdependent specialists to integrate
This is too much.” their efforts. Second, we show how informal coordi-
nation practices can trigger changes in formal coordi-
Zoe: “I would really like to understand what makes nation structures. The model captures the process
these compounds become positive here. We have to through which multidisciplinary project teams man-
understand that in more detail.”
age the continuously changing demand for expertise
Sub-team members Helen and Zoe’s suggestion to by dynamically restructuring those sub-teams to
further modify the compounds and redoing the tox- which the project team had delegated specific parts of
icity tests illustrates that relinquishing conjectural the knowledge creation process.
interdependencies often proves difficult. In such
cases where uncertainty emerges from the sub-team’s
The Mutual Constitution of Formal Coordination
findings without pointing to alternative courses of
Structures and Informal Coordination Practices
action, specialists tended to continue to model and
experiment within existing structures. From formal coordination structures to informal
coordination practices. Formal structuring in self-
managed multidisciplinary project teams involves
EMERGENT THEORETICAL MODEL
the selection and grouping of project team members
The findings revealed distinct types of formal and into sub-teams. This process is guided by project
informal coordination: formal structuring (of sub- team members’ collectively held assumptions about
teams) around interdependencies, and informal interdependencies among their respective knowl-
coordination practices within sub-teams (i.e., antic- edge domains for knowledge creation. Two mecha-
ipatory conforming, workflow synchronizing, and nisms underlie the link between the formal structure
TABLE 6
2016
Examples of Restructuring Around Revealed Interdependencies
Restructuring Around Revealed
Project Team Obstacle Interdependencies Implications for Knowledge Creation

InflaPro (project team leader, biology) Standard experiments did not result in A group of specialists in high-throughput “[The screen] generated new chemical
identification of any new compounds: screening—who could run assays with matter for the chemists to work on. . .and
“Finding new compounds for [this new read-out—was included in the sub- that was something really good because
target] is really challenging.” team. it’s usually a big effort to run a high-
throughput screen and yes, they did
it. . .it worked out.”
AutoPro (research associate, medicinal Fluorescent “tracer molecule” that tracks A specialist in structural biology—who “[The structural biologist] was like: ‘Oh,
chemistry) the location of the target protein in had the unique domain knowledge to yeah the spacer is not long
tissues did not show any activity: “One investigate how the tracer molecule enough. . .there might be some
part [of the molecule] should bind to [the might mechanistically bind to the target interaction between the [target] and the
target] and the other part should be based on three-dimensional fluorophore. Maybe the spacer should
fluorescent. The molecule is one models—was included in the sub-team. be bigger because on the other molecules
hundred percent pure, but it’s just not the spacer is bigger.’ So now we
fluorescent. So what is happening redesigned the molecule with a bigger
there?” spacer and another way of connecting.”
CanPro1 (senior investigator, molecular Experiments to identify new active A specialist in molecular modeling was Molecular modeling resulted in
modeling) compounds yielded no results (i.e., the added to the sub-team to run identifying new
sub-team ran out of molecules that computational algorithms to identify compounds—complementary to the
medicinal chemists could further complementary compounds. ones identified based on biochemical
optimize): “The difficult part now is to assays: “[Molecular modeling] was the
find potent molecules in a biochemical most powerful approach for developing
compound. . .it’s a really new target so the right molecules. . .the molecular
you have to start from scratch. Usually modeler had the most important
it’s rather difficult to get compounds contribution here, I would say.”
with high potency. However, this is
necessary for having a chance of seeing
some results in the following steps.”
CanPro2 (senior investigator, assay Hit list with hundreds of compounds had A specialist in structural biology was Different specializations included in the
Ben-Menahem, von Krogh, Erden, and Schneider

development) to be reduced to have a workable set of included in the sub-team to check sub-team conducted complementary
compounds for optimization: “. . .the whether there could be complementary experiments to triangulate the structural
laboratory showed a hit list from that selection criteria applied for confining biologist’s findings across knowledge
biochemical screen that included 776 the number of compounds. domains: “It turned out there was
compounds.” a reactive site on the surface of the
protein that was giving us a fair amount
of background [i.e., false positives].”
CanPro3 (investigator, structural Purification of the target—a routine A specialist in protein purification was “As we needed a purified target for [further
biology) activity—could not be carried out with included in the sub-team to develop experimentation], it just happened that
the typical protocols: “[The target] was a new protocol for effectively purifying he was probably the best person to carry
very difficult in terms of purification. . . . the target. out this activity. He could make a high
Normally [purification] is very fast and quality target protein very quickly.”
we all did that during our PhD and
postdoc time. We [i.e., the non-
specialists] typically do everything on
our own.”
1327
1328 Academy of Management Journal August

FIGURE 3
A Model of the Mutual Constitution of Formal Structuring and Informal Coordination Practices

and the informal practices that specialists use to that the specialist contributions expected from
manage interdependence in multidisciplinary teams: a sub-team insider (e.g., computational analysis and
(a) delegating responsibility for knowledge creation in modeling) differed substantially from those ex-
relevant knowledge domains and (b) developing team pected from sub-team outsiders (e.g., questioning
members’ awareness of interdependencies. everyday work and suggesting new approaches).
First, formal structuring allows project team mem- Once project team specialists joined a sub-team,
bers to decompose tasks and delegate responsibility their responsibilities changed to include a direct
for parts of the knowledge creation process to sub- involvement in resolving core scientific questions.
team members. Project teams thus reduce the overall Second, in creating a formal structure that stipulates
complexity of coordination among specialists from (sub-)team membership and defines the configuration
different knowledge domains. By associating mem- of interdependent knowledge domains, specialists
bership in different categories of team structures shape their shared mental representation of the web of
with different responsibilities for knowledge creation interdependencies in which their work is embedded.
(e.g., establishing the safety of a new compound), the Using the category of a sub-team, specialists develop
formal structure conveys information about how a social awareness of how their contributions relate to
project team members are expected to contribute those of other specialists in the collective endeavor
(i.e., in terms of norms and behavioral patterns; cf. (Tomasello, Carpenter, Call, Behne, & Moll, 2005), and
Bechky, 2006). This part of the formal structure infer which interactions to prioritize (Baumeister &
functions as a cue for specialists to cognitively asso- Masicampo, 2010; Puranam et al., 2012). Developing
ciate themselves as sub-team insiders or outsiders. awareness—albeit conjectural—of interdependencies
Such self-categorization (McCall & Simmons, 1978; is particularly relevant for coordinating knowledge
Stryker, 1980) prompts specialists to enact a corre- creation in a multidisciplinary and uncertain context.
sponding mode of contribution that involves either It elevates the target of specialists’ efforts to the
actively creating in-depth domain-specific knowl- collective team level, thus mitigating the danger of
edge or employing their expertise in an outsider role specialists’ narrowly interpreting their knowledge
(Turner, Oakes, Haslam, & McGarty, 1994). creation responsibilities within their own domain. As
At DrugCo, specialists across the five project teams specialists often engage in knowledge creation activi-
had an implicit yet broadly shared understanding ties in isolation, without taking interdependencies into
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1329

account, such a narrow domain-specific interpretation and the value of their findings across disciplinary
of responsibility is a common problem in multidisci- boundaries. By seeking confirmation from other
plinary work, severely limiting the relevance of spe- knowledge domains, specialists check whether their
cialists’ contributions and hampering integration. responsibility for contributing to the team’s collective
Given these two mechanisms, the relationship effort is fulfilled. These informal coordination prac-
between the formal structure and the emergence of tices allow specialists to make their work more
informal coordination practices can be understood predictable across domains, thus facilitating the
in terms of specialists’ efforts to manage conjectural integration of their contributions with those of other
interdependencies in a way that allows them to fulfill team members in a vigilant, timely, and collectively
their responsibility for knowledge creation. Spe- validated way.
cialists are the primary individuals responsible for Second, informal coordination practices encour-
creating knowledge while under scrutiny by other age specialists to redesign formal coordination
team members whose needs and evaluation criteria structures. The need for restructuring might arise
they can only partly anticipate—due to both task from progress in knowledge creation leading to new
uncertainty and lack of common ground between scientific questions that prove difficult to answer
knowledge domains (Lerner & Tetlock, 1999). Spe- within the current configuration of knowledge do-
cialists thus tend to exert themselves to develop mains. Alternatively, specialists might encounter
a deep, accurate understanding of how they can difficulties in producing predictive knowledge
contribute to the team’s overall tasks (e.g., De Dreu, (Puranam et al., 2012) and fulfilling their responsi-
Nijstad, & Van Knippenberg, 2008). As a result, they bility for knowledge creation while using informal
may intensify their engagement with other sub-team coordination practices within the context of the
members to search for, disseminate, and integrate existing formal structure. In such cases, they may be
information (e.g., De Dreu, Koole, & Steinel, 2000). unable to meet the needs of other specialists (Pfeffer
Our analysis shows that, to this end, sub-team & Sutton, 2000), attempt to provide contributions at
members draw on the three informal coordination inappropriate points in time, or repeatedly produce
practices that we conceptualized as anticipatory dissonant findings for unknown reasons.
conforming, workflow synchronizing, and cross- Experiencing such limitations of the formal struc-
domain triangulating. ture encourages specialists to take a more skeptical
From informal coordination practices to formal attitude to initially conjectured interdependencies
coordination structures. Two important features of and collectively reconsider the relevance of knowl-
informal coordination practices stand out. First, in- edge domains beyond the confines of the present (sub-)
formal coordination practices help shape specialists’ team structure. Such a response can be explained by
work such that it contributes to the sub-team’s col- specialists’ need to reduce the fear of creating invalid
lective task. In so doing, informal coordination prac- knowledge and avoid costly misapprehensions when
tices enable specialists to cope with the unpredictable the stakes involved in their activities are high (e.g.,
nature of cross-domain interdependencies. Specifi- Mayseless & Kruglanski, 1987). In our case, running
cally, anticipatory conforming involves specialists’ unnecessary or ill-timed experiments was very costly
vigilance in exercising their responsibility for and could severely disrupt progress of the drug dis-
knowledge creation by contemplating how applying covery project. Moreover, drawing false conclusions
domain-specific norms and behaviors for knowledge about the efficacy and safety of new compounds could
creation impacts the work of others. Such a vigilant significantly harm the public, the firm, and individual
approach compels seemingly interdependent spe- scientists (Aven, 2011).
cialists to specify the nature of their shared tasks and In DrugCo, sub-team specialists routinely reported
shape domain-specific knowledge creation efforts their findings (or lack thereof) to the project team dur-
such that they contribute to achieving the team’s ob- ing monthly meetings. This allowed project team
jective. Through workflow synchronizing, specialists members from a wider set of knowledge domains to
create a mutual understanding of temporal interde- exercise their “outsider” responsibility to knowledge
pendencies among team members. They can thus at- creation by posing questions to sub-team members and
tend to their domain-specific responsibilities for providing suggestions. In many instances we observed,
knowledge creation in a timely manner and flexibly the constructive comments and questions of project
readjust the team’s overall workflow when new in- team members revealed interdependencies among
sights emerge. Cross-domain triangulating enables knowledge domains spanning the sub-team’s bound-
specialists to scrutinize the validity of their approach aries. This process prompts specialists to formally
1330 Academy of Management Journal August

transition into (and out of) sub-teams whenever doing problem of unforeseeable outcomes of the knowledge
so appears valuable for progressing in collective creation process. Our study thus suggests that theoret-
knowledge creation. In such a way, project teams dy- ical accounts of how multidisciplinary teams co-
namically activate specialists’ responsibility for knowl- ordinate knowledge creation cannot be reduced to
edge creation and resolve incongruencies between the explaining how designers of team structures optimally
formal team structure and newly discovered interde- allocate tasks among specialists and integrate members’
pendencies. This continuous structural adaptation is efforts within static systems of interdependencies.
critical for coordinating specialists’ contributions in This study contributes to understanding how
multidisciplinary teams when task uncertainty is so structural design unfolds under uncertain conditions
intense that no one can anticipate either the interde- that cause organizational members to face unknown
pendencies or the outcomes. and unpredictable interdependencies (e.g., Cardinal
et al., 2011; Grandori & Soda, 2006; Miles, Snow,
Mathews, Miles, & Coleman, 1997; Sherman & Keller,
DISCUSSION
2011). Our model proposes that the process of de-
Despite their potential for making radically new signing and adapting formal structures around con-
and valuable discoveries, multidisciplinary teams jectural and revealed interdependencies can result
face significant coordination challenges involving from specialists’ self-managed knowledge creation
the division of labor and integration of specialists’ efforts. Within structures formally designed around
efforts (e.g., Cronin & Weingart, 2007; Dougherty, conjectural interdependencies, specialists enact in-
1992; Latour & Woolgar, 1986). Motivated by an in- formal coordination practices, which not only help
creasing disconnect between design and practice- them to develop predictive knowledge, but also pro-
based perspectives on these “universal problems of pel them to collectively search for and discover new
organizing” (Puranam, Alexy, & Reitzig, 2014), this interdependencies. To accommodate newly revealed
paper contributes integrative theory on how special- interdependencies, project teams flexibly reallocate
ists in multidisciplinary teams coordinate knowledge the responsibility for knowledge creation among
creation in the face of unpredictable interdepen- project team members from different knowledge do-
dencies. By simultaneously examining formal and mains, thereby adjusting the (sub-)team composition.
informal coordination mechanisms, this study offers Our theorizing about why this design dynamic oc-
new insights into their mutual constitution, with im- curs in a context of knowledge creation under un-
plications for research on both organizational design certainty brings together the literatures on team
and practice-based coordination. design and team-level information processing rooted
in lay epistemic theory (Kruglanski, 1989; Kruglanski &
Webster, 1996). In line with motivated information
Implications for Research on Organizational
processing theory (De Dreu et al., 2000), our findings
Design
show that bearing the responsibility for knowledge
Our study contributes to design-based theories of creation without possessing the necessary predictive
coordination and knowledge creation. Traditional knowledge to integrate their activities (Puranam et al.,
contingency perspectives on organizational design 2012) propels team members’ “willingness to ex-
emphasize the merits of creating team structures that pend effort to achieve a thorough, rich, and accurate
match the team’s information-processing requirements understanding of the world” (De Dreu et al., 2008: 23).
arising from its task environment (e.g., Donaldson, Furthermore, we show that such “collective in-
2001; Galbraith, 1977). An implicit assumption in re- formation processing” efforts have important implica-
search drawing on this structural contingency frame- tions for overcoming obstacles in team knowledge
work is that organizational designers choose team creation (Schippers, Edmondson, & West, 2014) by
structures by accurately assessing interdependencies increasing members’ understanding of not only the
among organizational members. Contrary to this as- nature and intensity (Knudsen & Srikanth, 2014;
sumption, our study shows that in the context of Puranam & Raveendran, 2013; Sherman & Keller,
knowledge creation, organizational members may ex- 2011) but also the scope of latent interdepen-
perience significant difficulties in designing optimal dencies that shape formal structures.
formal team structures around interdependencies. Our findings also enrich the literature on structural
Specifically, our findings highlight the inability to un- adaptation in teams. A longitudinal interpretation of
derstand the relevance of certain knowledge domains arguments underlying structural contingency theory
ex ante (i.e., during the initial design stage), and the (Thompson, 1967) leaves open many issues about
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1331

how structural designs adapt over time as task con- primarily on “high-reliability” or “fast response” or-
ditions change (Cronin, Weingart, & Todorova, 2011; ganizations (e.g., Faraj & Xiao, 2006; Klein, Ziegert,
Hollenbeck et al., 2011). The relatively few studies Knight, & Xiao, 2006; Majchrzak, Jarvenpaa, &
that explore the process and outcomes of structural Hollingshead, 2007; Weick & Roberts, 1993)—
adaptation conclude that appropriate coordination emphasizes the critical role of informal flexibility
and communication behaviors are necessary for enhancing and improvisational coordination mecha-
teams to adapt (Hollenbeck et al., 2011; Moon et al., nisms. Our theoretical model confirms these obser-
2004). Yet whereas prior empirical evidence on vations and adds that to cope with the changing
structural adaptation primarily concerns structures nature, intensity, and scope of interdependencies
with fixed membership and relatively broad, over- emerging in knowledge creation, team members
lapping knowledge domains (Hollenbeck, Beersma, & practice coordination through a combination of an-
Schouten, 2012), our study offers in-depth insights ticipatory conforming, workflow synchronizing, and
into structural adaptation through ongoing flexible cross-domain triangulating.
(re)configuration of highly specialized knowledge However, our model also differs from past work
domains. that primarily focused on how informal practices
This feature of our model displays close links to compensate for inertial forces of formal, hierarchical
Bigley and Roberts’ (2001) findings from a study of role structures (Klein et al., 2006) but that paid lim-
fire departments. They found that, to adapt largely ited attention to the potential interaction between
formalized organizational structure to complex and informal coordination practices and formal co-
volatile task environments, designers tend to rely on ordination structure (Okhuysen & Bechky, 2009). In
variable structuring mechanisms, such as assigning contrast, we emphasize that informal coordination
roles and tasks to human resources, reassigning practices are inherently embedded within and cir-
personnel to different positions, or deferring in- cumscribed by formal structures (Valentine &
formal decision-making authority to more qualified Edmondson, 2015), and demonstrate that formal
individuals. structures provide critically important—albeit
However, in Bigley and Roberts’ study, designers of tentative—guidance for specialists to practice co-
organizational structures could draw on reasonably ordination in multidisciplinary knowledge-creating
well-defined tasks and interdependencies among di- teams.
versely skilled individuals and so implement struc- By illuminating this relationship, our model fos-
tural change in a top-down fashion. Our theoretical ters the integration of practice-based research with
model complements and extends these findings in the the literature on the enabling effects of formal
context of cross-disciplinary knowledge creation by structure (e.g., Adler & Borys, 1996; Bresman &
showing that structural adaptation does not necessi- Zellmer-Bruhn, 2013; Cardinal, 2001; Jelinek &
tate an “omniscient designer” (Puranam & Raveendran, Schoonhoven, 1990). Specifically, we propose that
2013) “inscribing” his or her intentions on a team by the need for flexibility and adaptability in multidis-
structuring it (Orlikowski, 2008). Nor does it require ciplinary knowledge-creating teams arising from
team members to hold an accurate representation of unpredictable interdependencies cannot be fully
the optimal interdependencies that connect them met by informal coordination practices (e.g., Bruns,
(Puranam & Swamy, 2011). Instead, we propose that 2013). Our model suggests that in complex multi-
specialists’ conjectural, tentative representations of disciplinary teams comprising specialized indi-
interdependence, in conjunction with their informal viduals facing highly uncertain tasks, knowledge
coordination practice, provide a basis on which creation relies on formal team and sub-team struc-
structural team designs in highly uncertain contexts tures as a source of flexibility. Specialists draw on
can evolve in the direction of fit (Cardinal et al., 2011; formal structures not only to make sense of their
Siggelkow, 2002). sub-team responsibilities (cf. Valentine & Edmondson,
2015) but also to develop a vital awareness of the
changing web of interdependence within which
Implications for Practice-Based Research on
their knowledge creation efforts have a collective
Coordination
bearing.
This study also contributes to practice-based re- Without the ability to derive these cues from the
search focusing on how coordination of expertise un- formal team and sub-team structures, using co-
folds in situations with high complexity, rapid change, ordination practices to develop cross-domain pre-
and uncertainty. This stream of research—drawing dictive knowledge and integrate specialists’ efforts
1332 Academy of Management Journal August

would become highly time consuming and costly. light of this study’s boundary conditions and limita-
The decomposition of teams into sub-teams around tions. Our study focuses on self-managed teams with
conjectural interdependencies among specialists the discretion and power to change their structures.
greatly reduces the complexity of collective knowl- Previous studies have suggested that the impact of
edge creation, and thus establishes the grounds for structures can depend on the extent to which em-
informal coordination practices. Sub-team structur- ployees are involved in their formation and the degree
ing allows specialists to allocate attention to their to which these structures are aligned with their own
specific responsibilities in knowledge creation and goals (Adler & Borys, 1996; Langfred, 2007). There-
manage a narrower scope of conjectural interdepen- fore, the autonomy of project team members in
dencies relevant for the uncertain tasks at hand. The forming structural arrangements should be consid-
model shows that such complexity reduction is an ered a relevant boundary condition to the relation-
important condition for coordination practices to af- ship between informal coordination practices and
fect the managing of both current and emerging in- the structural adaptations explained in our model.
terdependencies. The coordination practices that we Moreover, given the importance of the precaution-
uncovered demand intense personal interaction in ary principle in drug discovery and scientists’ high
smaller groups if they are to effectively ensure ev- awareness of the potential safety implications of
eryday coordination among sub-team members, adapt mistakes, our research setting might reflect specific
coordination structures, and safeguard knowledge characteristics of members’ attitudes toward task un-
creation within higher-level organizational struc- certainty. Similarly, previous research has suggested
tures. This feature of our model contributes a valuable that distinctive characteristics of the science-based
understanding of how the dynamics of two-tier context (e.g., high task uncertainty and complexity)
structuring interacts with coordination practices. are important contingencies in project design
On a related note, as practice scholars have often (Cardinal et al., 2011; Pavitt, 1999).
sought to uncover fallacies in the logic underlying Given the five- to six-year length of a typical drug
the design perspective (Okhuysen & Bechky, 2009), discovery trajectory, our limited observation interval
they may have paid insufficient attention to the could not cover the complete trajectory of any one
conditions under which coordination practices en- project. Although we were able to study co-
able effective organizational structures. Notable ex- ordination in the early to late stages of drug discov-
ceptions include Brown and Duguid (1991), who ery, we are cautious about making claims about
proposed that effective organizational structures causal relationships between our findings and the
should be designed around social practices wher- ultimate success or failure of the drug discovery
ever these practices are effective for solving organi- projects we studied. Future studies should examine
zational problems.13 Following a reverse logic, our the extent to which the structural adaptations and
theoretical model shows how coordination practices informal coordination practices we observed impact
manifest the limitations of an existing configuration project performance. Moreover, as in any single-
of knowledge domains and thus trigger structural site study, the distinctive characteristics of the
adaptation. We therefore argue that to fully under- case company mean that further empirical research
stand how coordination of knowledge creation should examine to what extent our theoretical model
across a team’s knowledge domains unfolds when is transferrable to different firms in the same industry
interdependencies are unpredictable, managers and or to varied sites in which multidisciplinary teams
organization theorists need to consider the recursive face similar coordination challenges.
effects between the team’s formal structure and the Another avenue for future research relates to our
emergence of informal coordination practices. finding that as teams of specialists collaborate to cre-
ate knowledge, interdependencies among knowledge
domains change over time. This finding is important
Boundary Conditions, Limitations, and Directions
for the literature on the relationship between in-
for Future Research
novation performance and distance among specialist
Although our findings yield valuable insights into knowledge domains (e.g., Fleming, 2004; Kotha et al.,
a phenomenon that has remained largely inaccessible 2013), because it implies that knowledge distance is
to scholarly fieldwork, they should be interpreted in a dynamic aspect of multidisciplinary teams. Further
investigation of the evolution of knowledge distance
13
For an alternative analysis see: von Hippel and von within teams might thus add important new insights
Krogh (2016). to the coordination literature.
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1333

Practical Implications specialists in sub-teams to engage in informal co-


ordination practices of anticipatory conforming,
This study has three important practical implica-
workflow synchronizing, and cross-domain tri-
tions. First, it indicates that a mix of coordination
angulating. These informal coordination practices
structures and practices are necessary for knowledge
facilitate the integration of knowledge creation ef-
creation in multidisciplinary teams facing intense
forts and propel specialists to reveal new interde-
task uncertainty. Managers must provide sufficient
pendencies that establish the ground for structural
autonomy and resources for such project teams to
adaptation. Given the mutual constitution of these
self-manage interdependencies. Specialists must have
formal and informal coordination mechanisms in
time and slack resources both to formally restructure
cross-disciplinary knowledge creation, this study
the sub-teams and recruit specialists as needed, and
underscores the necessity of further integrating or-
to develop subtle, interpersonal coordination prac-
ganizational design and practice-based perspectives
tices. Second, to manage cross-disciplinary interde-
on coordination.
pendencies and enable vigilant, timely, and collectively
validated integration of efforts, the specialists we
studied sometimes had to forgo their domain specific REFERENCES
standards of excellence for the sake of the team’s
collective outcomes (cf. von Krogh, Haefliger, Spaeth, Adler, P. S., & Borys, B. 1996. Two types of bureaucracy:
Enabling and coercive. Administrative Science Quar-
& Wallin, 2012). This suggests that performance ap-
terly, 41: 61–89.
praisal in fundamental discovery needs to take
a comprehensive view of specialists’ contributions to Ancona, D. G., & Caldwell, D. F. 1992. Bridging the
the team’s shared objectives. Importantly, managers boundary: External activity and performance in orga-
should avoid the “appraisal trap,” whereby they nizational teams. Administrative Science Quarterly,
37: 634–665.
evaluate individual effort and skill based exclusively
on predefined standards within knowledge domains. Argote, L. 1982. Input uncertainty and organizational co-
Third, we found that sub-team outsiders—while ordination in hospital emergency units. Administra-
not considered directly responsible for knowledge tive Science Quarterly, 27: 420–434.
creation—contribute substantially to the work of Argote, L., Turner, M. E., & Fichman, M. 1989. To cen-
sub-team insiders. Managers cannot force specialists tralize or not to centralize: The effects of uncertainty
to externalize their (tacit) knowledge (Osterloh & and threat on group structure and performance. Or-
Frey, 2000), especially if these specialists do not feel ganizational Behavior and Human Decision Pro-
responsible for a particular task. Instead, managers cesses, 43: 58–74.
should encourage them to mutually support one Aven, T. 2011. On different types of uncertainties in the
another’s work and take an interest in the project’s context of the precautionary principle. Risk Analysis,
overall progress through, for example, mentoring 31: 1515–1525.
and training programs, project debriefings, and pe- Baumeister, R. F., & Masicampo, E. J. 2010. Conscious
riodic social events. thought is for facilitating social and cultural in-
teractions: How mental simulations serve the animal-
culture interface. Psychological Review, 117: 945–971.
CONCLUSION Bechky, B. A. 2006. Gaffers, gofers, and grips: Role-based
This study examined how multidisciplinary teams coordination in temporary organizations. Organiza-
coordinate knowledge creation while facing un- tion Science, 17: 3–21.
predictable interdependencies among knowledge Bechky, B. A., & Okhuysen, G. A. 2011. Expecting the
domains. Our findings from early-stage drug dis- unexpected? How SWAT officers and film crews
covery teams suggest a dynamic relationship be- handle surprises. Academy of Management Journal,
tween formal coordination structures and informal 54: 239–261.
coordination practices. By continuously designing Becker, G. S., & Murphy, K. M. 1992. The division of labor,
formal sub-team structures around conjectural in- coordination costs, and knowledge. The Quarterly
terdependencies, project teams adaptively delegate Journal of Economics, 107: 1137–1160.
responsibility for knowledge creation to relevant Bigley, G. A., & Roberts, K. H. 2001. The incident command
specialists. We show that formal structures not only system: High-reliability organizing for complex and
reduce the complexity of collective knowledge volatile task environments. Academy of Manage-
creation at the project team level but also enable ment Journal, 44: 1281–1299.
1334 Academy of Management Journal August

Booth, B., Glassman, R., & Ma, P. 2003. Oncology’s trials. functionally diverse teams. Academy of Manage-
Nature Reviews. Drug Discovery, 2: 609–610. ment Review, 48: 1107–1123.
Brass, D. J., Galaskiewicz, J., Greve, H. R., & Tsai, W. 2004. Cronin, M. A., Weingart, L. R., & Todorova, G. 2011. Dy-
Taking stock of networks and organizations: A multi- namics in groups: Are we there yet? The Academy of
level perspective. Academy of Management Journal, Management Annals, 5: 571–612.
47: 795–817. Davis, J. P., Eisenhardt, K. M., & Bingham, C. B. 2009. Op-
Bresman, H. 2013. Changing routines: A process model of timal structure, market dynamism, and the strategy of
vicarious group learning in pharmaceutical R&D. simple rules. Administrative Science Quarterly, 54:
Academy of Management Journal, 56: 35–61. 413–452.
Bresman, H., & Zellmer-Bruhn, M. 2013. The structural De Dreu, C. K. W., Koole, S., & Steinel, W. 2000. Unfixing
context of team learning: Effects of organizational and the fixed pie: A motivated information-processing
team structure on internal and external learning. Or- account of integrative negotiation. Journal of Per-
ganization Science, 24: 1120–1139. sonality and Social Psychology, 79: 975–987.
Brown, J. S., & Duguid, P. 1991. Organizational learning De Dreu, C. K. W., Nijstad, B. A., & Van Knippenberg, D.
and communities-of-practice: Toward a unified view 2008. Motivated information processing in group
of working, learning, and innovation. Organization judgement and decision making. Personality and
Science, 2: 40–57. Social Psychology Review, 12: 22–49.
Bruns, H. C. 2013. Working alone together: Coordination in Donaldson, L. 2001. The contingency theory of organi-
collaboration across domains of expertise. Academy zations. Thousand Oaks, CA: Sage.
of Management Journal, 56: 62–83.
Dougherty, D. 1992. Interpretive barriers to successful
Bunderson, J. S., & Boumgarden, P. 2010. Structure and product innovation in large firms. Organization Sci-
learning in self-managed teams: Why “bureaucratic” ence, 3: 179–202.
teams can be better learners. Organization Science,
Dougherty, D., & Dunne, D. D. 2012. Digital science and
21: 609–624.
knowledge boundaries in complex innovation. Or-
Burns, T. E., & Stalker, G. M. 1961. The management of ganization Science, 23: 1467–1484.
innovation. London, UK: Tavistock.
Drews, J. 2000. Drug discovery: A historical perspective.
Burton, R. M., & Obel, B. 2004. Strategic organizational Science, 287: 1960–1964.
diagnosis and design: The dynamics of fit. New York,
Eder, J., Sedrani, R., & Wiesmann, C. 2014. The discovery of
NY: Springer.
first-in-class drugs: Origins and evolution. National
Cardinal, L. B. 2001. Technological innovation in the Review, 13: 577–587.
pharmaceutical industry: The use of organizational
Edmondson, A. C. 1999. Psychological safety and learn-
control in managing research and development. Or-
ing behavior in work teams. Administrative Science
ganization Science, 12: 19–36.
Quarterly, 44: 350–383.
Cardinal, L. B., Sitkin, S. B., & Long, C. P. 2004. Balancing and
Edmondson, A. C. 2003. Speaking up in the operating
rebalancing in the creation and evolution of organiza-
room: How team leaders promote learning in in-
tional control. Organization Science, 15: 411–431.
terdisciplinary action teams. Journal of Management
Cardinal, L. B., Turner, S. F., Fern, M. J., & Burton, R. M. Studies, 40: 1420–1452.
2011. Organizing for product development across
Edwards, A. M., Isserlin, R., Bader, G. D., Frye, S. V.,
technological environments: Performance trade-offs
Willson, T. M., & Yu, F. H. 2011. Too many roads not
and priorities. Organization Science, 22: 1000–1025.
taken. Nature, 470: 163–165.
Carlile, P. R. 2002. A pragmatic view of knowledge and
Faraj, S., & Sproull, L. 2000. Coordinating expertise in
boundaries: Boundary objects in new product devel-
opment. Organization Science, 13: 442–455. software development teams. Management Science,
46: 1554–1568.
Carlile, P. R. 2004. Transferring, translating, and transforming:
An integrative framework for managing knowledge Faraj, S., & Xiao, Y. 2006. Coordination in fast-response
across boundaries. Organization Science, 15: 555–568. organizations. Management Science, 52: 1155–1169.

Child, J., & McGrath, R. G. 2001. Organizations unfettered: Fleming, L. 2004. Perfecting cross-pollination. Harvard
Organizational form in an information intensive Business Review, 82: 22–24.
economy. Academy of Management Journal, 44: Galbraith, J. R. 1973. Designing complex organizations.
1135–1148. Reading, MA: Addison-Wesley.
Cronin, M. A., & Weingart, L. R. 2007. Representa- Galbraith, J. R. 1977. Organization design. Reading, MA:
tional gaps, information processing, and conflict in Addison-Wesley.
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1335

Gardner, H. K., Gino, F., & Staats, B. R. 2012. Dynamically Kellogg, K. C., Orlikowski, W. J., & Yates, J. A. 2006. Life in
integrating knowledge in teams: Transforming re- the trading zone: Structuring coordination across
sources into performance. Academy of Management boundaries in postbureaucratic organizations. Orga-
Journal, 55: 998–1022. nization Science, 17: 22–44.
Gioia, D. A., Corley, K. G., & Hamilton, A. L. 2012. Seeking Kilduff, M., & Brass, D. J. 2010. Organizational social net-
qualitative rigor in inductive research: Notes on the Gioia work research: Core ideas and key debates. Academy
methodology. Organizational Research Methods, 16: of Management Annals, 4: 317–357.
15–31. Klein, K. J., Ziegert, J. C., Knight, A. P., & Xiao, Y. 2006.
Gittell, J. H. 2002. Coordinating mechanisms in care pro- Dynamic delegation: Hierarchical, shared and dein-
vider groups: Relational coordination as a mediator dividualized leadership in extreme action teams.
and input uncertainty as a moderator of performance Administrative Science Quarterly, 51: 590–621.
effects. Management Science, 48: 1408–1426. Knudsen, T., & Srikanth, K. 2014. Coordinated exploration
Grandori, A. 2010. A rational heuristic model of economic organizing joint search by multiple specialists to
decision making. Rationality and Society, 22: 477–504. overcome mutual confusion and joint myopia. Ad-
Grandori, A., & Soda, G. 2006. A relational approach to orga- ministrative Science Quarterly, 59: 409–441.
nisation design. Industry and Innovation, 13: 151–172. Kogut, B., & Zander, U. 1992. Knowledge of the firm,
Grant, R. M. 1996. Toward a knowledge-based theory of the combinative capabilities, and the replication of tech-
firm. Strategic Management Journal, 17: 109–122. nology. Organization Science, 3: 383–397.

Gulati, R., & Puranam, P. 2009. Renewal through re- Kotha, R., George, G., & Srikanth, K. 2013. Bridging the
organization: The value of inconsistencies between mutual knowledge gap: Coordination and the com-
formal and informal organization. Organization Sci- mercialization of university science. Academy of Man-
ence, 20: 422–440. agement Journal, 56: 498–524.

Hargadon, A. B., & Bechky, B. A. 2006. When collections of Kruglanski, A. W. 1989. Lay epistemics and human
creatives become creative collectives: A field study of knowledge: Cognitive and motivational bases. New
problem solving at work. Organization Science, 17: York, NY: Springer.
484–500. Kruglanski, A. W., & Webster, D. M. 1996. Motivated
Hoegl, M., & Gemuenden, H. G. 2001. Teamwork quality closing of the mind: “Seizing” and “freezing”. Psy-
and the success of innovative projects: A theoretical chological Review, 103: 263–283.
concept and empirical evidence. Organization Sci- Lamont, M., & Molnár, V. 2002. The study of boundaries in
ence, 12: 435–449. the social sciences. Annual Review of Sociology, 28:
Hollenbeck, J. R., Beersma, B., & Schouten, M. 2012. Be- 167–195.
yond team types and taxonomies: A dimensional scal- Langfred, C. W. 2007. The downside of self-management:
ing conceptualization for team description. Academy A longitudinal study of the effects of conflict on
of Management Review, 37: 82–106. trust, autonomy, and task interdependence in self-
Hollenbeck, J. R., Ellis, A. P. J., Humphrey, S. E., Garza, A., & managing teams. Academy of Management Journal,
Ilgen, D. R. 2011. Asymmetry in structural adaptation: 50: 885–900.
The differential impact of centralizing versus decentral- Latour, B., & Woolgar, S. 1986. Laboratory life: The con-
izing team decision-making structures. Organizational struction of social facts. Princeton, NJ: Princeton
Behavior and Human Decision Processes, 114: 64–74. University Press.
Huber, G. P. 1991. Organizational learning: The contrib- Lawrence, P. R., & Lorsch, J. W. 1967. Differentiation and
uting processes and the literatures. Organization integration in complex organizations. Administrative
Science, 2: 88–115. Science Quarterly, 12: 1–47.
Hülsheger, U. R., Anderson, N., & Salgado, J. F. 2009. Team- Lee, T. W., Mitchell, T. R., & Sablynski, C. J. 1999. Quali-
level predictors of innovation at work: A comprehen- tative research in organizational and vocational psy-
sive meta-analysis spanning three decades of research. chology, 1979–1999. Journal of Vocational Behavior,
The Journal of Applied Psychology, 94: 1128–1145. 55: 161–187.
Jelinek, M., & Schoonhoven, C. B. 1990. The innovation Lerner, J. S., & Tetlock, P. E. 1999. Accounting for the ef-
marathon: Lessons from high technology firms. Ox- fects of accountability. Psychological Bulletin, 125:
ford, UK: Wiley-Blackwell. 255–275.
Jones, B. F. 2009. The burden of knowledge and the “death Liebeskind, J. P. 1996. Knowledge, strategy, and the theory
of renaissance man”: Is innovation getting harder? The of the firm. Strategic Management Journal, 17:
Review of Economic Studies, 76: 283–317. 93–107.
1336 Academy of Management Journal August

Majchrzak, A., Jarvenpaa, S. L., & Hollingshead, A. B. 2007. Okhuysen, G. A., & Bechky, B. A. 2009. Coordination in
Coordinating expertise among emergent groups respond- organizations: An integrative perspective. The Acad-
ing to disasters. Organization Science, 18: 147–161. emy of Management Annals, 3: 463–502.
Majchrzak, A., More, P. H. B., & Faraj, S. 2012. Tran- Orlikowski, W. J. 2000. Using technology and constituting
scending knowledge differences in cross-functional structures: A practice lens for studying technology in
teams. Organization Science, 23: 951–970. organizations. Organization Science, 11: 404–428.
March, J. G., & Simon, H. A. 1958. Organizations. New Orlikowski, W. J. 2008. Using technology and constituting
York, NY: Wiley. structures: A practice lens for studying technology
Mayseless, O., & Kruglanski, A. W. 1987. What makes you in organizations. In M. S. Ackerman, C. A. Halverson,
so sure? Effects of epistemic motivations on judg- T. Erickson, & W. A. Kellogg (Eds.), Resources, co-
mental confidence. Organizational Behavior and evolution and artifacts: 255–305. London, UK:
Human Decision Processes, 39: 162–183. Springer.

McCall, G. J., & Simmons, J. L. 1978. Identities and in- Osterloh, M., & Frey, B. S. 2000. Motivation, knowledge
teractions. New York, NY: Free Press. transfer, and organizational forms. Organization Sci-
ence, 11: 538–550.
McEvily, B., Soda, G., & Tortoriello, M. 2014. More for-
mally: Rediscovering the missing link between formal Pammolli, F., Magazzini, L., & Riccaboni, M. 2011. The
organization and informal social structure. The Acad- productivity crisis in pharmaceutical R&D. Nature
emy of Management Annals, 8: 299–345. Reviews. Drug Discovery, 10: 428–438.

Mell, J. N., van Knippenberg, D., & van Ginkel, W. P. 2014. Patton, M. Q. 2002. Qualitative research and evaluation
The catalyst effect: The impact of transactive memory methods (3rd ed.). Thousand Oaks, CA: Sage.
system structure on team performance. Academy of Paul, S. M., Mytelka, D. S., Dunwiddie, C. T., Persinger,
Management Journal, 57: 1154–1173. C. C., Munos, B. H., Lindborg, S. R., & Schacht, A. L.
Miles, M. B., & Huberman, A. M. 1994. Qualitative data 2010. How to improve R&D productivity: The phar-
analysis: An expanded sourcebook (2nd ed.). Thou- maceutical industry’s grand challenge. Nature Re-
sand Oaks, CA: Sage. views. Drug Discovery, 9: 203–214.

Miles, R. E., Snow, C. S., Mathews, J. A., Miles, G., & Pavitt, K. 1999. Technology, management and systems of
Coleman, H. J., Jr.. 1997. Organizing in the knowledge innovation. Cheltenham, UK: Edward Elgar.
age: Anticipating the cellular form. The Academy of Pennings, J. M. 1992. Structural contingency theory: A
Management Perspectives, 11: 7–20. reappraisal. In B. M. Staw & L. L. Cummings (Eds.),
Miller, C. C., Cardinal, L. B., & Glick, W. H. 1997. Retro- Research in organizational behavior, vol. 14: 267–309.
spective reports in organizational research: A reexami- Greenwich, CT: JAI Press.
nation of recent evidence. Academy of Management Perrow, C. 1967. A framework for the comparative analysis
Journal, 40: 189–204. of organizations. American Sociological Review, 32:
Moon, H., Hollenbeck, J. R., Humphrey, S. E., Ilgen, D. R., 194–208.
West, B., Ellis, A. P. J., & Porter, C. O. L. H. 2004. Asym- Pfeffer, J., & Sutton, R. I. 2000. The knowing-doing gap:
metric adaptability: Dynamic team structures as one-way How smart companies turn knowledge into action.
streets. Academy of Management Journal, 47: 681–695. Boston, MA: Harvard Business School Press.
Munos, B. 2009. Lessons from 60 years of pharmaceutical in- Pinto, M. B., Pinto, J. K., & Prescott, J. E. 1993. Antecedents
novation. Nature Reviews. Drug Discovery, 8: 959–968. and consequences of project team cross-functional
Nag, R., Corley, K. G., & Gioia, D. A. 2007. The intersection cooperation. Management Science, 39: 1281–1297.
of organizational identity, knowledge, and practice: Puranam, P., Alexy, O., & Reitzig, M. 2014. What’s “new”
Attempting strategic change via knowledge grafting. about new forms of organizing? Academy of Man-
Academy of Management Journal, 50: 821–847. agement Review, 39: 162–180.
Nonaka, I. 1994. A dynamic theory of organizational Puranam, P., & Raveendran, M. 2013. Interdependence and
knowledge creation. Organization Science, 5: 14–37. organization design. In A. Grandori (Ed.), Handbook
Nonaka, I., & von Krogh, G. 2009. Tacit knowledge and of economic organization: 193–209. Northampton,
knowledge conversion: Controversy and advance- MA: Edward Elgar.
ment in organizational knowledge creation theory. Puranam, P., Raveendran, M., & Knudsen, T. 2012. Organi-
Organization Science, 20: 635–652. zation design: The epistemic interdependence perspec-
Northrup, J. 2005. The pharmaceutical sector. In L. R. tive. Academy of Management Review, 37: 419–440.
Burns (Ed.), The business of healthcare innovation: Puranam, P., & Swamy, M. 2011. Expedition without
27–102. New York, NY: Cambridge University Press. maps: Why faulty initial representations may be
2016 Ben-Menahem, von Krogh, Erden, and Schneider 1337

useful in joint discovery problems. Working Paper, Turner, J. C., Oakes, P. J., Haslam, S. A., & McGarty, C. 1994.
London Business School, London. Self and collective: Cognition and social context.
Ringel, M., Tollman, P., Hersch, G., & Schulze, U. 2013. Personality and Social Psychology Bulletin, 20:
Does size matter in R&D productivity? If not, what 454–463.
does? Nature Reviews. Drug Discovery, 12: 901–902. Tushman, M. L., & Nadler, D. A. 1978. Information pro-
Rishton, G. M. 2005. Failure and success in modern drug cessing as an integrative concept in organizational de-
discovery: Guiding principles in the establishment of sign. Academy of Management Review, 3: 613–624.
high probability of success drug discovery organiza- Tyre, M. J., & von Hippel, E. 1997. The situated nature of
tions. Medicinal Chemistry (Shariqah, United Arab adaptive learning in organizations. Organization
Emirates), 1: 519–527. Science, 8: 71–83.
Sams-Dodd, F. 2005. Target-based drug discovery: Is some- Valentine, M. A., & Edmondson, A. C. 2015. Team scaf-
thing wrong? Drug Discovery Today, 10: 139–147. folds: How mesolevel structures enable role-based
Schippers, M., Edmondson, A. C., & West, M. A. 2014. coordination in temporary groups. Organization
Team reflexivity as an antidote to team information- Science, 26: 405–422.
processing failures. Small Group Research, 45: von Hippel, E., & von Krogh, G. 2016. Identifying viable
731–769. “need-solution pairs”: Problem solving without prob-
Scott, W. R. 1987. Organizations: Rational, natural, and lem formulation. Organization Science, 27: 207–221.
open systems (2nd ed.). Englewood Cliffs, NJ: Prentice von Krogh, G., Haefliger, S., Spaeth, S., & Wallin, M. 2012.
Hall. Carrots and rainbows: Motivation and social practice
Sherman, J. D., & Keller, R. T. 2011. Suboptimal assessment in open source software development. Management
of interunit task interdependence: Modes of integra- Information Systems Quarterly, 36: 649–676.
tion and information processing for coordination per- von Krogh, G., Nonaka, I., & Rechsteiner, L. 2012. Leadership in
formance. Organization Science, 22: 245–261.
organizational knowledge creation: A review and frame-
Siggelkow, N. 2002. Evolution towards fit. Administrative work. Journal of Management Studies, 49: 240–277.
Science Quarterly, 47: 125–159.
Vural, O. M., Dahlander, L., & George, G. 2013. Collabora-
Singh, J., & Fleming, L. 2010. Lone inventors as sources of tive benefits and coordination costs: Learning and
breakthroughs: Myth or reality? Management Sci- capability development in science. Strategic Entre-
ence, 56: 41–56. preneurship Journal, 7: 122–137.
Sitkin, S. B., Cardinal, L. B., & Bijlsma-Frankema, K. M. Weick, K. E., & Roberts, K. H. 1993. Collective mind in
(Eds.) 2010. Organizational control. Cambridge, UK: organizations: Heedful interrelating on flight decks.
Cambridge University Press. Administrative Science Quarterly, 38: 357–381.
Soda, G., & Zaheer, A. 2012. A network perspective on Wuchty, S., Jones, B. F., & Uzzi, B. 2007. The increasing
organizational architecture: Performance effects of the dominance of teams in production of knowledge.
interplay of formal and informal organization. Stra- Science, 316: 1036–1039.
tegic Management Journal, 33: 751–771.
Yin, R. K. 2003. Case study research: Design and methods
Srikanth, K., & Puranam, P. 2011. Integrating distributed (3rd ed.). Thousand Oaks, CA: Sage.
work: Comparing task design, communication, and
tacit coordination mechanisms. Strategic Manage-
ment Journal, 32: 849–875.
Strauss, A. L., & Corbin, J. M. 1998. Basics of qualitative
research: Techniques and procedures for de- Shiko M. Ben-Menahem (benmenahem@ethz.ch) is a senior
veloping grounded theory (2nd ed.). Thousand Oaks, researcher and lecturer in the Department of Management,
CA: Sage Publications. Technology, and Economics at ETH Zurich. He received his
PhD from Rotterdam School of Management, Erasmus
Stryker, S. 1980. Symbolic interactionism: A social struc- University. His research interests include organizational
tural version. Menlo Park, CA: Benjamin Cummings. and team design, coordination of knowledge intensive work,
Thompson, J. D. 1967. Organizations in action: Social and organizational renewal and innovation.
science bases of administrative theory. New York, Georg von Krogh (gvkrogh@ethz.ch) is the Chaired Professor
NY: McGraw-Hill. of Strategic Management and Innovation at ETH Zurich. He
Tomasello, M., Carpenter, M., Call, J., Behne, T., & Moll, H. is also a member of the National Research Commission of the
2005. Understanding and sharing intentions: The or- Swiss National Science Foundation. Georg has published
igins of cultural cognition. Behavioral and Brain widely on topics such as organizational knowledge creation
Sciences, 28: 675–735. and technological innovation. His current research focuses
1338 Academy of Management Journal August

on the effectiveness and efficiency of different approaches to • Follow up: How do they differ? How are collabo-
innovation within and between firms. rations established? What are your experiences in
Zeynep Erden (zerden@ethz.ch) is a senior researcher and working in each of these forms of collaboration?
lecturer in the Department of Management, Technology, and How does it influence your work?
Economics at ETH Zurich. She received her PhD in man- • What enables you to work efficiently with other
agement and organization theory from ETH Zurich. Her re- disciplines?
search interests include creation and coordination of • In collaborating with others, how do you develop
knowledge in teams, implementation of knowledge and in- a shared view of the work and tasks you will
novation strategies in organizations, and their impact on firm perform?
performance, especially in the pharmaceutical industry. • How do you ensure that project team members
Andreas Schneider (andreas.schneider@ypsomed.com) is “are in the know”?
a business development manager at Ypsomed Group. He • What information do you share with other team
received his PhD in management and organization theory members?
from ETH Zurich. His research explores collaborative • What material do you exchange with other
knowledge creation practices and approaches to new disciplines when interacting? Tell me about
product development. a situation that shows how you exchange
materials.
• How do you process and prepare information be-
fore sharing it with other team members?
APPENDIX A • How do other project team members enrich what
you know?
INTERVIEW PROTOCOL • To what extent do you need to understand other
scientific disciplines to do your work?
• Can you tell me about your work and contribution • Can you tell me about the technologies you use in
to [this project]? this project? How do these technologies relate to
• What does a typical workday look like for you? workflow in the project?
• Which scientific questions do you try to answer in • Can you guide me through an example of how you
this project? draw conclusions from data?
• Follow up: Which scientific disciplines take part • Can you tell me about how you collaborate with
in this process? others when drawing conclusions from data?
• Which other scientific disciplines did you interact • Questions about project-teams and sub-teams
with until now to fulfill your work on this project? when these were specifically mentioned as forms
How do you expect that to develop? of collaboration:
• How do project team members possibly differ? • What feedback do people give each other in sub-
• People here at [DrugCo] talk a lot about “integrated team meetings, project-team meetings?
drug discovery”: What does integration from your • How do you contribute to sub-team meetings,
point of view imply? How does integration hap- project-team meetings?
pen in [your project]? Note: Core questions are listed. The interview protocol
• How are decisions made about who does what and was adapted as we progressed in our data collection to
when? ensure its alignment with emerging insights. As a result,
• Can you tell me about the different ways people different participants were asked somewhat different in-
collaborate in your project? terview questions.
Copyright of Academy of Management Journal is the property of Academy of Management
and its content may not be copied or emailed to multiple sites or posted to a listserv without
the copyright holder's express written permission. However, users may print, download, or
email articles for individual use.

Вам также может понравиться