1
--- Accepted by Research Policy (Dec. 2014) ---
Organizational Design of University Laboratories: Task Allocation and Lab Performance
in Japanese Bioscience Laboratories
Sotaro SHIBAYAMA a,b,*
Yasunori BABA a
John P. WALSH c,d
a University of Tokyo. Research Center for Advanced Science and Technology.
4-6-1 Komaba, Meguro-ku, Tokyo 153-8904, Japan
Tel/Fax: +81-3-5452-5371
b University of Tokyo. Department of Technology Management for Innovation.
7-3-1 Hongo, Bunky-ku, Tokyo 113-8656, Japan
c Georgia Institute of Technology. School of Public Policy.
685 Cherry Street, Atlanta, GA 30332-0345, USA.
d National Graduate Institute for Policy Studies.
7 Chome-22-1 Roppongi, Minato, Tokyo 106-0032, Japan.
* Corresponding author.
2
Abstract
A university laboratory is a fundamental unit of scientific production, but optimizing its
organizational design is a formidable task for lab heads, who play potentially conflicting roles of
manager, educator, and researcher. Drawing on cross-sectional data from a questionnaire survey
and bibliometric data on Japanese biology professors, this study investigates task allocation
inside laboratories. Results show a general pattern that lab heads play managerial roles and
members (e.g., students) are engaged in labor-intensive tasks (e.g., experiment), while revealing
a substantial variation among laboratories. Further examining how this variation is related to
lab-level scientific productivity, this study finds that productive task allocation differs by context.
In particular, results suggest that significant task overlap across status hierarchies is more
productive for basic research, and that rigidly separated task allocation is more productive in
applied research. However, optimal task allocation, with regard to scientific productivity, might
conflict with other goals of academic organizations, particularly training of future scientists. The
paper concludes with a discussion of the policy implications of these findings.
Keywords
Laboratory; task allocation; organizational design; scientific productivity; biology
3
1. Introduction
Since the modern economy relies heavily on scientific production in the academic sector,
the organizational design of academic research organizations is a critical agenda for science and
technology policy (Etzkowitz and Leydesdorff, 2000; Stephan, 1996). Academic science,
especially in natural sciences and engineering, is usually undertaken in laboratories that consist
of a lab head (also called principal investigator) and member researchers under his or her
supervision. Unlike temporary collaboration, the continuous nature of laboratories allows lab
heads with a long-range plan to set research goals, arrange a portfolio of research projects, make
investments in facilities, and accumulate and reuse a local knowledge base (Carayol and Matt,
2006; Knorr-Cetina, 1999; Latour and Woolgar, 1979; Owen-Smith, 2001). For these reasons,
prior work has suggested that a laboratory is the appropriate unit when analyzing the nature of
scientific production (Carayol and Matt, 2006; Latour and Woolgar, 1979).
Studies of the organizational design of laboratories, whether in academia or in industry,
date back to the 1950s. Among others, Pelz and Andrews (1966) examined the relationship
between scientific production and a series of organizational factors, broadly covering various
scientific fields and sectors. Subsequent literature in the sociology of science has further
investigated the roles of organizational factors such as communication, coordination, leadership,
and organizational prestige in scientific research (e.g., Allison and Long, 1990; Andrews, 1979;
Heinze et al., 2009; Hollingsworth and Hollingsworth, 2000; Long and McGinnis, 1981;
Zuckermann, 1977). Literature from other disciplinary perspectives has also advanced
understanding in specific aspects of organization; for example, the social psychology literature
studies creativity and its antecedents (e.g., Amabile, 1996) and the organization management
literature examines the motivation of researchers (e.g., Agarwal and Ohyama, 2012; Sauermann
and Stephan, 2012).
While these studies have informed how various organizational factors can affect
scientific production, they have paid limited attention to a peculiarity of university laboratories.
Academic science heavily depends on junior researchers, including students, who are often short
of experience and need training (Knorr-Cetina, 1999; Owen-Smith, 2001). Obviously,
universities are responsible not only for scientific production but also for education (Hackett,
1990), and thus, lab heads are obliged to train young members, although these two missions of
research and education could be in conflict (e.g., Fox, 1992). This is a major challenge for lab
4
heads, who have to organize the lab considering potentially incompatible goals of research and
education when deciding on task allocation for the lab head and its members. This division of
labor and potentially conflicting relationships between a lab head and members have been noted
in a few studies in the sociology of education (e.g., Delamont and Atkinson, 2001; Delamont et
al., 1997; Salonius, 2008) but analyses of their implications for scientific production have been
limited.
To fill these gaps in the literature, this study examines the organizational design of
university laboratories, highlighting the roles of lab heads and members. Investigating task
allocation in the lab context requires in-depth understanding of the distinctive activities in lab
work. In this regard, prior ethnographies of academic laboratories have illustrated in great detail
how academic science operates in one or a few specific laboratories (Knorr-Cetina, 1999; Latour
and Woolgar, 1979; Owen-Smith, 2001; Salonius, 2008). Typically, they describe task allocation
in academic laboratories as lab heads being the manager, who is busy planning, fund-raising, and
supervising members, with members being the workers, concentrating on conducting
experiments and other laborious tasks. To advance this simplified model of task allocation, we
draw on the above-outlined literature on the organization of research groups (e.g., Hollingsworth
and Hollingsworth, 2000; Pelz and Andrews, 1966; Sauermann and Stephan, 2012). In particular,
we examine two forms of possible deviation from the typical task allocation: 1) whether lab
members should engage not only in labor-intensive tasks but also in upstream tasks, and 2)
whether lab heads should engage also in labor-intensive tasks rather than staying away from the
bench like a pure manager. We argue that the optimal task allocation depends on context (Cyert
and March, 1963; Simon, 1957). In particular, we hypothesize that the pattern of task allocation
should be differentiated depending on the orientation of research in terms of being basic vs.
applied.
Drawing on interviews with 30 researchers and a questionnaire survey of 396 lab heads
from Japanese universities in the field of biology, we first draw a general picture of task
allocation in university laboratories. We find it basically consistent with the stylized view of task
allocation, but we also observe considerable variation. Second, we examine the effect of task
allocation on scientific productivity and its contingency on research orientation. Based on our
empirical results, we discuss implications for science policies.
5
2. Theory and Hypothesis
2.1. Social organization of lab work
Research activities in natural sciences are usually undertaken in laboratories that consist
of a lab head and some members under the lab head’s supervision (e.g., Carayol and Matt, 2006;
Latour and Woolgar, 1979; Owen-Smith, 2001). Lab heads are usually professors, and members
include students, postdoctoral researchers (postdocs), junior faculty, and technicians. Unlike
temporary collaboration, laboratories are characterized by a continuous form of teamwork. Lab
heads can pursue relatively long-term goals. They arrange a portfolio of research projects, some
of which may be challenging but with potentially great impact and others of which are less novel
but with limited risk, so that they can constantly produce at least minimal expected output
(Knorr-Cetina, 1999). Laboratories allow division of labor. Particularly in biology, since a
project often involves multiple techniques (Knorr-Cetina, 1999; Latour and Woolgar, 1979),
coordinating researchers with different expertise is essential. Lab tasks are also vertically divided.
Lab heads are usually responsible for setting up the research environment (e.g., funding,
equipment, and recruitment) and coordinating a series of projects, while members engage in
executing specific projects (Traweek, 1988). In addition, laboratories function as a place of
education and training. Young researchers typically consider their lab experience as an
opportunity to acquire research techniques, which will prepare them for future employment
(Delamont and Atkinson, 2001; Delamont et al., 1997).
In terms of task allocation, prior literature has mainly focused on the role of lab heads
and assumed that lab heads are occupied with upstream tasks. In a report on the career design of
American life scientists, the National Research Council (1998) mentions that “[a] principal
investigator builds a research group by defining the scientific questions to be addressed,
specifying the methods to be used, obtaining necessary funding, finding the suitable research
environment, and attracting the research personnel…. The research personnel in the group
usually work on more specific tasks that pertain to the construction of research tools or the
acquisition and analysis of data.” Similary, Knorr-Cetina (1999) finds that in the field of
molecular biology researchers often stop bench work after becoming lab heads. The role of
members, on the other hand, has been relatively understudied. A few studies in the sociology of
education, focusing on postgraduate education, have examined the division of labor between lab
heads and PhD students (Delamont and Atkinson, 2001; Delamont et al., 1997; Salonius, 2008).
6
Delamont et al. (1997), drawing on ethnographic research in British universities, suggest that lab
heads are responsible for identifying research projects and assigning them to students. Becher et
al. (1994) also point out that determining research subjects is rarely the responsibility of students.
Since mastering technical skills is the most important goal during the student’s lab experience
(Delamont and Atkinson, 2001), engaging in technical tasks seems to be regarded as the students’
primary role.
To further the discussion of task allocation, we distinguish three phases of the research
process. In general, scientific research starts from setting a research question and developing a
research plan; then, the question is tested by experiments, simulations, and other approaches; and
finally, the test results are interpreted and used to advance extant knowledge (Nightingale, 1998).
This last phase often raises new questions for future research, and the whole process is repeated.
We split this process into two phases: 1) planning, or determining research subjects and
hypotheses, and 2) execution, or testing the hypotheses, usually by experiment and data analyses
in biology. In addition, we consider the phase of 3) writing scientific papers. Planning and
execution are iterated until sufficient results are accumulated that make up a story as a paper. For
these three phases, lab ethnographies and sociology of education literature generally suggest that
lab heads are the primary player in planning and members in execution, but they are less clear
about task allocation in writing (Delamont and Atkinson, 2001; Knorr-Cetina, 1999; Latour and
Woolgar, 1979). Based on these studies, the following section first describes the general features
and rationales of task allocation for each phase. Then, we add competing arguments from the
literature on the organization of research groups (e.g., Hollingsworth and Hollingsworth, 2000;
Pelz and Andrews, 1966; Sauermann and Stephan, 2012).
2.2. Rationales of task allocation
2.2.1. Execution phase
Since biology is strongly driven by empiricism (Bertalanffy et al., 1962), biological
research heavily depends on experiments, except for purely computational or theoretical
subfields. In the execution phase, researchers attempt to transform some material substances into
interpretable information, which often takes the forms of figures or tables, and some of them are
used in the next writing phase as “results” in publications (Latour and Woolgar, 1979: Ch. 2).
This transformation may be processed manually or through devices such as DNA sequencers.
7
Some experimental procedures follow established protocols and may be made available as
commercial kits or somewhat automated, while others are less established and researchers may
have to start from developing and optimizing protocols. Importantly, even a well-established
procedure often requires substantial tacit knowledge to perform properly, such that researchers
inexperienced in that task could repeat fail to get results (Knorr-Cetina, 1999; Latour and
Woolgar, 1979). Still, lab techniques in biology are fairly standardized (Whitley, 1984), and even
PhD students can learn them in a reasonably short time. It is also noteworthy that biological
experiments usually draw on living organisms such as bacteria, cell lines, mice, and even humans
in clinical research. These organisms are highly complex and often produce unexpected results,
requiring enormous efforts of trial and error by experimenters. In addition, since living
organisms need maintenance, often on a daily basis, researchers tend to be chained to
laboratories. Experiments could take a range of times, from minutes to overnight and from weeks
to months, and researchers have to schedule their tasks depending on the life cycle of these
organisms. For these reasons, execution tasks are highly labor-intensive and time-consuming,
where members seem to have comparative advantage over lab heads (Delamont and Atkinson,
2001; Delamont et al., 1997).
Despite this stylized argument, lab heads’ co-participation could improve productivity
for a few reasons. Because lab members spend most of their time at the bench, a lab head’s
engagement in execution implies the collocation of a lab head and members. This can facilitate
communication and team coherence, both essential drivers of creativity and team coordination
(Amabile, 1996; Pelz and Andrews, 1966; Stapleton, 2004: Ch.2). Lab heads can communicate
with members through various channels such as lab meeting, but casual conversation at the
bench should lower communication barriers and may allow lab heads to obtain information
otherwise inaccessible.
More specifically, collocation enables lab heads to closely monitor members’ routine
activities. Through co-presence, lab heads could find out about and help solve problems in
members’ execution tasks in a timely fashion. This could have substantial impact on the progress
of members’ work. Biological experiment could easily take weeks or months until yielding
results, and inexperienced members could be unaware of critical mistakes that are obvious to
experienced researchers. In fact, Andrews (1979) finds that lab heads’ participation in scientific
work helps detect problems and improve team coherence. Teasley et al. (2002) show that the
8
productivity of software development projects is significantly improved when project teams are
collocated because possible bottlenecks are easily spotted and needed members are always
accessible.
Not only members’ problems but also their success are easily identified by collocation.
Biological research is often exploratory, where unintended results may bring about
breakthroughs (Merton, 1968; Merton and Barber, 2004). However, members tend to overlook
unexpected findings due to the lack of a holistic perspective and might even deliberately do so
for fear that lab heads dislike hearing unexpected news (Barber and Fox, 1958; Van Angel, 1992).
In this regard, lab heads’ close monitoring can be especially helpful. In fact, Shimizu et al.
(2012), drawing on a survey of natural scientists in the US and Japan, find that serendipitous
findings are deterred if the managerial role and the execution role are played by different
scientists.
Another rationale is technical catch-up. That is, lab heads stay able to work with the
latest lab equipment and techniques (including acquiring the necessary tacit knowledge). One lab
head we interviewed emphasized this point.
When interpreting experimental results, researchers have to distinguish true
from false signs of discoveries and to find out hidden serendipitous signs.
They may be obvious for experimenters but not for non-experimenters. It is
not rare that pure-manager lab heads misinterpret experimental results and
make silly instructions to their members. Unfortunately, members often have
to follow the instructions and tend to blindly do so especially when the lab
head is renowned.
Because most experimental techniques take a great deal of tacit knowledge, it is difficult even for
experienced lab heads to follow state-of-the-art techniques if they are not doing any experimental
work directly. Knorr-Cetina (1999) suggests that an experimental technique is a “package” of
protocol, material objects, and researchers. Lab heads may be able to understand new techniques
theoretically, but only experimenters know the knack of techniques. Lab heads will become
technically obsolete if they distance themselves from bench work (Salonius, 2008), which can be
a serious impediment when experimental techniques are rapidly advancing, as in biology.
9
2.2.2. Planning phase
Before execution, research projects must be planned. The planning phase starts with
choosing research subjects and identifying specific questions. Ultimately, the product of research
needs to be published, which is a highly competitive process (Merton, 1973). Thus, researchers
carefully choose their subjects so that they can publish more quickly and with a greater impact
than their competitors do, with prior research an essential source of information for planning.
Once research questions are settled, researchers have to translate their hypotheses into a
technically operational plan for the execution phase, which obviously requires technical
knowledge. Overall, researchers have to integrate and process various types of knowledge to
develop a feasible plan. Thus, lab heads, with greater intellectual capabilities and longer
experience, seem to have a comparative advantage for this phase (Delamont and Atkinson, 2001;
Knorr-Cetina, 1999).
The stereotyped task allocation that lab heads plan and members execute appears
reasonable considering members’ limited experience and capabilities. However, this is exactly
why they need training. Although mastering experimental skills may be the first priority for
young members (Delamont and Atkinson, 2001), learning how to design and coordinate research
projects should be indispensable. Thus, one may argue that members must be engaged in the
whole process of research for educational purposes, even if it may compromise lab performance.
In the long term, well-trained researchers should better serve scientific communities. Thus, lab
heads face the dilemma of whether to prioritize research productivity or to give their
subordinates training opportunities at the cost of productivity (Hackett, 1990). Many of our
interviewees referred to this dilemma, suggesting that there are two types of laboratories: one
where members are treated like blue-collar workers in a factory, and the other where members
are trained as future lab heads. Criticizing the former type of laboratories, one interviewee stated:
In natural sciences, it may be common that young members, especially
students, are exploited to produce experimental results as if they were
technicians. However, I believe that such an approach cannot develop good
researchers. … I believe that universities are the place for education, and thus,
students must be respected more than professors.
Although the above argument implies that members should be given training
10
opportunities in phases other than execution, it still maintains the assumption that this would
compromise productivity. However, we argue that members’ engagement in planning can
contribute to productivity. First, the creativity literature suggests that autonomy and
independence facilitates divergent thinking and exploratory problem-solving, which are critical
for creative performance (Amabile, 1996; Woodman et al., 1993). Pelz and Andrews (1966:
Ch.2) suggest that lack of autonomy is detrimental to scientific productivity, though excessive
autonomy can lead to isolation, and in particular, that sharing research goals among teammates
contributes to productivity.
Second, as often discussed in connection with autonomy, motivation is also
indispensable for creativity (Amabile, 1996; Ford, 1996; Pelz and Andrews, 1966: Ch.6). Roach
and Sauermann (2010) find that PhD students in science and engineering show strong preference
for freedom to choose research projects. Even inexperienced members desire to be involved in
decision making. If autonomy in project selection is given, members can ascribe their success or
failure to their own actions, which stimulates their intrinsic motivation (Hackman and Oldham,
1976). Thus, involving members from the outset of the research process can encourage them to
seriously engage in later phases, but isolation from the planning phase can demotivate them. One
lab head we interviewed argued:
Because the execution phase in life science research is painstakingly
laborious, members could not go through it without strong intrinsic
motivation. In this regard, having members engage in planning is effective. I
try to respect members’ choice of research topics even if they seem likely to
fail, hoping for them to reach a serendipitous discovery.
Third, members’ engagement can contribute to coordination across phases, which,
coupled with autonomy, is critical for scientific productivity (Pelz and Andrews, 1966: Ch.2).
Though the three phases may appear to proceed linearly, upstream tasks are often revised based
on feedback from downstream tasks. In the planning phase, research plans may have to be
adjusted in accordance with experimental results. If members, who are the main player in the
execution phase, are involved in the planning phase, they should be able to understand the aim of
execution tasks, efficiently inform lab heads of experimental results, and properly fulfill changed
plans.
11
2.2.3. Writing phase
Writing is the process of producing a scientific paper for publication using the results
obtained in the execution phase. Although the process of writing a paper may span all three
phases in that literature may be reviewed during planning and figures and tables are created
during execution (Latour and Woolgar, 1979: Ch. 2), we narrowly define it as the process of
writing after results are produced. Still, writing is more than a mechanical process of
summarizing experimental results and can be an intellectual process of interpreting results,
placing them adequately in the context of concurrent scientific debates, and creating a story that
interests peers. These tasks take a set of skills. First, since biological research projects usually
involve multiple members with different technical expertise, the execution phase produces
various types of results, among which appropriate ones need to be selected, as the first step of
writing. This coordination role seems manageable only by lab heads who have the authority and
a holistic viewpoint beyond each member’s. Second, biological research is often serendipitous,
and experimental results are unpredictable (Merton, 1968; Merton and Barber, 2004; Whitley,
1984: Ch.4). To write a paper with unexpected results, authors may have to start over from
literature review and revise the original storyline. Thus, the quality of papers can be greatly
affected by authors’ theoretical knowledge. In addition, the ability to find out serendipitous
results depends on the observer’s knowledge (Seymore, 2009). As Pasteur said, “Fortune favors
the prepared mind.” Lab heads may be better able to see the potential in experimental results.
Third, the writing process can involve informal communication with other researchers. Before
submitting a paper for peer review, authors attempt to increase the likelihood of acceptance by
incorporating the knowledge of leading researchers in the field. Some of our interviewees
emphasized that negotiations with journal editors are also indispensable. Thus, this phase seems
to require both intellectual and social skills that are better exercised by experienced lab heads
than by young members.
However, this argument depends on how much value needs to be added in this phase.
For example, if the research goal is practical, not much theoretical knowledge may be
additionally required. If experimental results are predictable, the storyline of a paper can be fixed
before execution and the value-added in this phase can be limited (Whitley, 1984). Then,
members’ writing may be justified since their labor cost is lower. In addition, assuming that
12
members are the main player of execution, lab heads’ writing can be inefficient because it incurs
communication costs and might suffer from imprecise or incomplete knowledge transfer. Thus,
members may be able to write papers more efficiently than lab heads under some circumstances.
2.3. Contextual contingency
The previous section outlines competing rationales for task allocation in each phase. The
overall impact of task allocation on scientific productivity is thus determined by the balance of
these rationales, which we argue is contingent on organizational context. Prior literature has
identified various contextual factors that explain scientific productivity, such as leadership,
organizational prestige, size, and age (Allison and Long, 1990; Heinze et al., 2009; Long and
McGinnis, 1981; Pelz and Andrews, 1966). We suppose that these factors not only directly affect
scientific productivity but also can change the effect of task allocation on scientific productivity.
Among other contextual factors, we formulate hypotheses on contingency on research
orientations, comparing basic vs. applied research.
Biology is a broad discipline and is related to several research fields such as medicine,
agriculture, and pharmaceuticals. This diversity in research goals can be relevant in coordinating
lab work (Sauermann and Stephan, 2012; Whitley, 1984). Some researchers seek general
understanding of certain phenomena while others are guided by consideration for practical use,
where the former is called basic and the latter applied (Stokes, 1997). Calvert (2004) contrasts
basic and applied research, identifying their characteristics. First, basic research is unpredictable,
where researchers aim to find a new concept or push the boundaries of existing knowledge. This
feature in basic research leads to an exploratory approach compared to a more confirmatory
approach in applied research. Second, basic research is general in that results can be used for a
wide range of instances and phenomena while applied research helps solve a specific problem.
Third, basic research is driven by the theoretical dynamics inside the discipline. This is also
related to the generality as theories involve statements of general principles. We assume that
these features of basic or applied research affect the productivity of different task allocations.1 In
what follows, we discuss how research orientations change the impact of different task
1 Research areas may be a result of strategic choice, but we assume that this choice is less dynamic or occurs only in
the longer term. For example, the basicness of research is substantially affected by disciplines, and thus, by the
department laboratories belong to. Thus, in the short term, research areas could be regarded as a given contextual
factor.
13
allocations in each phase.
2.3.1. Member’s engagement in planning
In the planning phase, we have referred to autonomy, motivation, and coordination as
the rationales of members’ engagement. We argue that these factors play a more important role in
basic research than in applied research. Intrinsic motivation and autonomy are regarded as
antecedents of creativity in general. The mechanism behind this argument is that they facilitate
exploratory and divergent thinking (Amabile, 1996), which is more relevant to basic research
(Calvert, 2004). For example, Sauermann and Cohen (2010), based on a survey of industry
researchers, show that intrinsic motivation contributes to productivity to a greater extent in
upstream R&D activities than in downstream. In addition, the exploratory nature of basic
research implies that plans in basic research are prone to frequent updating. Based on
experimental results, researchers have to frequently adjust their research plan; thus the feedback
loop between planning and execution should be tightly linked (Nelson, 1959). This can be
streamlined by members’ engagement in the planning phase. In contrast, this potential benefit
seems limited in applied research, where the goal of research is clearer and members can stick to
the original plan. Thus, we hypothesize:
Hypothesis 1: Members’ engagement in planning has a more positive effect on productivity
in basic research than in applied research.
2.3.2. Lab head’s engagement in execution
We have pointed out collocation and technical catch-up as possible justifications for lab
heads’ engagement in execution, which we argue are particularly important in basic research. As
discussed above, the exploratory nature and abstract goals in basic research imply that research
plans tend to be loosely predetermined and frequently updated. Exploratory research encourages
autonomous trials and errors, but this incurs the risk that members are stuck in trivial problems
or unpromising lines of research. This calls for good communication between a lab head and
members and lab heads’ direct monitoring even though it is costly. Furthermore, since the
success of basic research depends more on unplanned findings, a keen eye for serendipitous
signs in experimental results is essential. Given that young members have limited capabilities in
this respect, lab heads’ collocation may be needed. Finally, to effectively monitor members and
14
detect interesting signs from noisy experimental results, lab heads need to maintain their
technical expertise by catching up with the latest technologies. In sum, we hypothesize:
Hypothesis 2: Lab heads’ engagement in execution has a more positive effect on
productivity in basic research than in applied research.
2.3.3. Task allocation in writing
We have discussed that the optimal task allocation in writing depends on the balance
between the expected value-added in writing and labor cost, which we argue differ by research
orientations. The ultimate goal of applied research is, by definition, application (Calvert, 2004;
Stokes, 1997). Thus, practically useful findings can be valued and published even if they do not
advance theoretical understanding. For example, research on clinical medicine can be published
if it proves the efficacy of a drug substance but does not elucidate its mechanism. On the other
hand, basic research aims to advance knowledge and tends to refer to general and abstract
concepts (Calvert, 2004; Stokes, 1997). Thus, researchers have to understand the up-to-date
theoretical debate and incorporate their findings in it. In this regard, lab heads’ advantage over
young members in writing is greater in basic research.
The unpredictable nature of basic research could foster this tendency. Basic research
takes a more exploratory approach and applied research a more confirmatory approach (Calvert,
2004). Basic research often starts from a broad question without having a precisely testable
hypothesis, and experimental results might be applied to a diverse range of scientific discussion.
If researchers have a broad range of knowledge not only about the originally intended areas but
also about surrounding areas, serendipitous discoveries are more likely. In this regard, writing in
basic research can be more a creative process of generating a novel story. For example, this
interviewee suggested the necessity of substantial knowledge and experience in writing in basic
research.
Serendipitous discoveries are important in biological, particularly basic,
research. I think that even young researchers could find unintended results if
they are careful enough. However, it does not guarantee publication. For
publication, serendipitous discoveries must be theorized and proved in
accordance with the extant theories. This takes substantial knowledge and is
not feasible for inexperienced researchers.
15
On the other hand, applied research tends to have a clear focus on application (Calvert, 2004). To
the extent that the research goal is specific, room for creative interpretation is limited. Thus,
writing in applied research can be a process of summarizing experimental results according to
the predetermined plan. In this case, inexpensive members’ labor lends them a comparative
advantage. In summary, we hypothesize:
Hypothesis 3: Lab heads’ engagement in writing has a more positive effect on productivity
in basic research than in applied research, while members’ engagement in writing has a more
positive effect on productivity in applied research than in basic research.
3. Data & Method
3.1. Sample and data
To test our hypotheses, we conducted a questionnaire survey of lab heads of Japanese
biology laboratories and collected their publication data from the Web of Science (WoS), which
yielded cross-sectional data of 396 laboratories.
To begin, we interviewed 30 Japanese researchers to investigate the context of
university laboratories and to design the survey instrument. The sampling frame of the survey
was prepared with the following criteria. First, we chose researchers currently in the position of a
full professor. Japanese universities have a three-level promotion system with full professors at
the top, followed by associate professors, and then, by assistants or lecturers. Before becoming
an assistant or lecturer, researchers tend to experience a few years as a postdoc. Typical biology
laboratories consist of senior staff (full and/or associate professor), who are the lab head, and
members, including a few junior faculty members (assistant or lecturer), postdocs, students, and
technicians. Unlike American universities, junior faculty members are often under the
supervision of lab heads. Some associate professors have independent laboratories, and others
work in the same laboratory with a full professor, often co-supervising a laboratory. Second, we
chose researchers who have received a national research grant in the field of biology at least once
in the previous three years (2007-2009), which implies that they are active researchers.2
2 We prepared our sampling list using the government’s database of Grants-in-Aid for Scientific Research (GiA)
(https://kaken.nii.ac.jp/en/). This sampling strategy is based on the assumption that academics who received no grant
for three years are not researchers. Although doing research without receiving this particular grant is possible,
previous research shows that it is not common (Shibayama, 2011), as GiA is the primary funding source for
16
Drawing on the list of grant recipients, we prepared a sampling frame of 1,378 lab heads. After
re-examining their research fields and affiliations with public information, we identified 900 lab
heads in 56 universities as a final sample. We mailed the survey and collected 396 responses
(response rate = 44%) after two waves of requests, May through July 2010.3
We collected bibliometric data for the 396 respondents. We primarily drew on
publications in 2007-2011 to measure the current productivity of each laboratory. Because lab
heads in biology usually become an author of all the papers produced in their laboratories,
publications including lab heads as an author should cover most publications from their
laboratory. As search criteria for the WoS, we used the names and affiliations of the respondents
and research areas being in life sciences. After downloading all matched data, we excluded false
matches on the basis of authors’ full names, affiliations, and so forth.4
3.2. Measurement
Lab productivity. We prepared two measures of productivity at the lab level. First, we
counted the number of publications authored by our respondents in 2007-2011 (pub count).
Second, to measure a more quality-focused aspect of scientific production, we drew on citation
counts. To address the age effect of citation, we summed up citation counts using a three-year
window for each of the papers (citation count).5 Then, we divided pub count and citation count
by the number of lab staff (lab heads and junior researchers) and use their logarithms in
regressions.6 Because these two measures are highly correlated and the regression results are
similar, we mainly report the results based on the citation count.
Organizational structure and task allocation. We asked respondents about the number
of senior staff (full and associate professors), junior researchers (assistant professor, lecturer, and
Japanese academics and widely awarded (http://www.jsps.go.jp/j-grantsinaid/index.html). 3 To examine non-response bias, we randomly selected 50 non-respondents and found no significant difference
between the response and non-response groups in publication productivity, organizational rank, or gender (p > 0.1). 4 We assumed that authors whose full name and affiliation are the same are identical. Since the majority of
publication data in 2007-2011 include author full names, we first excluded false hits whose authors share only
initials but not full names. Approximately 20% of our publication data in 2007-2011 do not have author full names,
and we determined whether their authors were our respondents or not by comparing cited references and coauthors’
names with publication data whose author full names are available. 5 Since newer publications tend to have fewer citations, we fix the time window of citation at three years (i.e.,
citations within three years after publication are counted). We drew on publications up to 2011 for the same reason:
citation counts would be unreliable for too recent publications. 6 We also computed lab productivity using fractional count of publications and citations based on the number of
authors (Lindsey, 1980). The results are qualitatively similar (Table S1 in Supplementary Results), so we present
results with ordinary measures.
17
postdocs), PhD students, and technicians in each laboratory. The total number of these
researchers is used as lab size. On average, a laboratory consists of 1.6 senior staff, 2.6 junior
researchers, 2.8 PhD students, and 1.1 technicians. To examine task allocation, we define six
research tasks: 1) choosing a subject, 2) formulating a hypothesis, 3) planning experiment, 4)
doing experiment, 5) analyzing data, and 6) writing papers. We argue that 1) – 3) corresponds to
the planning phase, 4) and 5) to the execution phase, and 6) to the writing phase. For each of
these six tasks, we asked the extent of involvement of full professors (our respondents), associate
professors, junior researchers, and PhD students, respectively. The responses take a three-point
scale: 0: no role, 1: minor role, and 2: main role, where we allow a “main role” to be played by
more than one rank group. When a laboratory is co-supervised by full and associate professors,
we take the maximum value of their engagement as the lab head’s engagement. For regression
analyses, we prepared the following three variables corresponding to the three hypotheses
respectively. First, we averaged the six items regarding the planning of junior researchers and
PhDs (members’ planning). Second, we averaged the two items on lab head’s engagement in
execution (lab head’s execution). Third, we took the average of researcher’s writing, PhD’s
writing, and the inverse of lab head’s writing (members’ writing (& lab head’s not writing)). In
the following analyses, we focus on laboratories that have at least one junior researchers and one
PhD student to consistently compute the task allocation measures.7 Of the 396 laboratories, 305
laboratories (77%) satisfy this condition with a mean size of 9.2 and the standard deviation of
5.2.
Basic research. To measure the research orientation of each laboratory, we asked
“which describes your research goal, basic (aiming at advancement of theory and knowledge) or
applied (aiming at solving problems in the real society)?” with a five-point scale, 1) mostly basic,
2) more basic than applied, 3) both to a similar extent, 4) more applied than basic, and 5) mostly
applied. Of our respondents, 55% chose 1), implying that their research goal was completely
basic. For these laboratories, a dummy variable is coded one, and other laboratories are regarded
7 For our measurement design, we cannot properly compute the task allocation variables in laboratories without
junior researchers or PhDs. Since excluding these laboratories could cause a bias, we conducted a sensitivity
analysis to confirm that it does not affect our results (Table S2 in Supplementary Results). We further analyze the
nature of the dropped laboratories and find that they are significantly smaller than laboratories included. By
additionally examining the recent career of all lab heads, we also find that the lab heads of dropped laboratories are
more likely to retire soon after our survey. Lab heads about to retire tend to avoid employing members to be ready
for closing their laboratories.
18
as applied with the dummy coded zero (basic research). To validate this subjective measurement,
we drew on the type of journals where the respondents publish their papers. Based on a
classification of the “basicness” of journals (Narin et al., 1976), we calculated the percentage of
respondents’ papers in basic journals and confirmed that basic laboratories tend to publish in
basic journals (r = .41, p < .001). Second, we surveyed the number of patent applications in
2009–2010 and confirmed that basic laboratories have significantly fewer patents than applied
laboratories (.31 vs. .85 applications per year; p < .001). Third, we identified the research field in
which each respondent received the majority of his national grants. We categorized these fields
into basic and applied fields8 and confirmed that this measure is correlated with basic research (r
= .41, p < .001). In addition, we tested the assumption that basic research is more exploratory
and applied research is more confirmatory (Calvert, 2004). We surveyed “which describes the
quality of your research, exploratory or confirmatory?” with a similar five-point scale, finding
that this is positively correlated with the measurement of being basic vs. applied (r =.23, p
< .001).
Control variables. The productivity of individual researchers should affect both lab
productivity and task allocation. To incorporate lab head’s individual productivity, we drew on
publications authored by the respondents during the five years before they obtained a tenured
position (i.e., before they opened their own laboratory). For these publications, similarly to the
lab productivity measures, we computed publication count and citation count with a three-year
citation window (pre-tenure pub count and pre-tenure citation count). As a measure of
organizational prestige, we prepared a dummy variable for the top seven pre-imperial universities
(top 7 univ) because they enjoy exceptionally high prestige both in research and in education
among other Japanese universities (Kneller, 2007).9 As measures of research input, we include
per-staff research budget (JPY in million: budget/#staff) and lab head’s average hours spent on
research activities (time for research). Time for research was measured with a six-point scale, 1)
less than 10 hours, 2) 10-20 hours, 3) 20-30 hours, 4) 30-40 hours, 5) 40-50 hours, and 6) 50
hours or longer per week. Several measures for individual attributes are included. We controlled
8 Basic fields include basic medicine, neuroscience, genome science, etc. while applied fields include agricultural
science, pharmaceutical, clinical medicine, etc. 9 As of 2010, Japan had 778 national, regional and private universities with four-year degree programs
(http://www.e-stat.go.jp/). Among them, national universities are the primary player in academic research while
most private universities are education-oriented. Among 86 national universities, the top seven (Universities of
Tokyo, Kyoto, Osaka, Tohoku, Hokkaido, Kyushu, and Nagoya) are designated as pre-imperial (Kneller, 2007).
19
for the number of years since lab heads opened their laboratory (lab age). We asked about
experience of research abroad with a six-point scale, 1) none, 2) less than half a year, 3) one year,
4) 2 years, 5) 3 years, and 6) 4 years or more (foreign experience). If the current laboratory is
where they obtained their degree, a dummy variable is coded one (inbred). If a lab head had a
medical doctor degree, a dummy variable is coded one (medical doctor). If a lab head is female,
a dummy variable is coded one (female). In addition, we identified the research field of each
laboratory on the basis of fields attributed to the lab heads’ past research grants, and in regression
analyses, we included 21 dummy variables for these fields.10
Table 1 presents the descriptive
statistics and correlation matrix of these variables.
<< Insert Table 1 about here >>
4. Results
4.1. Variation in task allocation
Figure 1 illustrates the extent of engagement in six research tasks by three ranks. It
indicates that the planning phase is primarily conducted by lab heads and execution by members
(junior researchers and PhDs). This confirms the stylized model of task allocation (National
Research Council, 1998; Delamont and Atkinson, 2001; Knorr-Cetina, 1999). However, the
extent of lab heads’ engagement in execution and that of members’ planning show substantial
variation, suggesting that the division of labor is not uniform across labs. As for the writing
phase, lab heads are highly committed but members are also engaged; only about 10% of
members play no role in writing.
<< Insert Figure 1 about here >>
Then, we analyze the extent of deviation from the typical task allocation in Table 2. In
each phase, we identified laboratories where the main role of each phase is played by the atypical
rank. For simplicity, we took the average of junior researchers’ and PhDs’ engagement as
members’ engagement. In about one-third of the laboratories, members are engaged in planning
(Row 1), and in about a half of the laboratories, lab heads are engaged in execution (Row 2) and
members are engaged in writing (Row 3). Again, these results suggest that the division of labor is
10
Fields include neuroscience, environmental science, agricultural science, pharmaceutical, basic medicine, clinical
medicine, structural biochemistry, biophysics, molecular biology, cell biology, and so forth.
20
not uniform. In particular, we do not see a strict division of labor such that lab heads plan and
members execute. Further, we break down the patterns of task allocation into the eight
possibilities of three rank-phase combinations (Rows 4-11). Rows 4 and 5 are supposedly a
typical pattern, which account for 35%, but other patterns are not rare (12-20%) except Rows 8
and 10.
<< Insert Table 2 about here >>
Finally, to validate our three variables of task allocation, we run a factor analysis and a
varimax rotation for the 18 measures of task allocation (three ranks x six task types). We obtain a
six-factor solution based on the Kaiser-Guttman criterion (i.e., eigenvalues greater than one).
Appendix 1 presents the factor loadings of the six factors. This solution is consistent with our
instrument design in that tasks intended to be in the same phase are actually attributed to the
same factor. For example, Factors 1 and 3 highlight the measures for three planning tasks, and
Factors 4 and 5 for two execution tasks (indicated in bold italic). Furthermore, our three
variables of task allocation are found in three distinctive factors; Factor 1 corresponds to
members’ planning, Factor 5 to lab head’s execution, and Factor 6 to members’ writing (and lab
head’s not writing).
4.2. Prediction of scientific productivity
Tables 3 and 4 present the results of regression analyses with per-staff citation count and
pub count as the dependent variables, respectively. For citation count (Table 3), Model 1, based
on the whole sample, shows a weakly positive effect for members’ planning (b = .241, p < .1)
and insignificant effects for lab head’s execution and members’ writing (p > .1). As for control
variables, pre-tenure citation count shows strongly positive effects in all models and female
shows strongly negative effects in Models 1-3. To investigate the contextual contingency, we
split the sample into basic and applied laboratories (Models 3 and 4). Members’ engagement in
planning shows a significantly positive effect in basic laboratories (b = .476, p < .05) and a
positive, though insignificant, effect in applied laboratories (b = .280, p > .1). Interestingly, lab
head’s execution shows opposite signs between two subsamples: positive in basic laboratories (b
= .481, p < .05) but negative in applied laboratories (b = -.385, p < .05). This implies that the cost
of lab head’s execution might exceed its benefit in applied laboratories. Finally, members’
21
writing shows a significantly negative effect only in basic laboratories (b = -.527, p < .1). In
order to statistically compare the coefficients between basic and applied laboratories, we draw on
two approaches. First, we add interaction terms for task allocation and basic research in Model 2
for the whole sample. The results show a strongly significant interaction effect for lab head’s
execution (b = .895, p < .001). The other two interaction terms are insignificant although their
signs agree with our expectation. Second, we directly compare the coefficients between Models 3
and 4, allowing the coefficients of all independent variables to differ between the two
subsamples.11
The result does not show statistical difference for members’ planning (p > .1) but
show strongly significant difference for lab head’s execution (p < .001) and weakly for members’
writing (p < .1). These results support Hypothesis 2 and weakly supports Hypothesis 3, but do
not support Hypothesis 1.
Table 4 shows qualitatively similar results with publication count as the dependent
variable. Comparing basic and applied laboratories, Model 2 shows significant interaction effects
for lab head’s execution (b = .429, p < .001) and for members’ writing (b = -.327, p < .1). In
addition, direct comparison between Models 3 and 4 show significant differences for all three
task allocation measures (p < .1, p < .01 and p < .05, respectively). Thus, all the hypotheses are
supported when publication count is used as the dependent variable.
The contrast between Tables 3 and 4 suggests that the effect of task allocation could be
different for the qualitative and quantitative aspects of scientific production. Figure 2 illustrates
the association between task allocation and the two productivity measures in terms of the three
task allocation measures. It suggests that in applied laboratories members’ planning may
contribute to the quality (citation count) but not to the quantity (pub count) of publication, and
likewise, members’ writing may increase quantity but may not matter for quality.
<< Insert Tables 3 and 4 and Figure 2 about here >>
5. Discussion and Conclusions
5.1. Findings and implications
Drawing on a sample of Japanese biology laboratories, this study first examines the
patterns of task allocation. While our results confirm the prior assumption that lab heads
11
Namely, we included interaction terms with basic research for all independent variables (not only focal variables
but also other control variables). In addition, we drew on seemingly unrelated estimation technique with the STATA
command suest (Weesie, 1999). Both methods yield qualitatively similar results.
22
primarily engage in planning tasks and members in execution tasks (Delamont and Atkinson,
2001; Delamont et al., 1997; Salonius, 2008), they also show some variation in task allocation.
These results contribute to drawing a general picture of organizational design of academic
laboratories, developing prior literature based on ethnographies (Knorr-Cetina, 1999; Latour and
Woolgar, 1979; Owen-Smith, 2001; Salonius, 2008).
Second, this study investigates how variation in task allocation affects scientific
productivity in different contexts. To this end, we build a framework of three research phases
drawing on lab ethnographies (e.g., Knorr-Cetina, 1999; Latour and Woolgar, 1979; Salonius,
2008). Furthermore, we discuss the rationales of task allocation between lab heads and members
based on literature on the organization of research groups (e.g., Hollingsworth and Hollingsworth,
2000; Pelz and Andrews, 1966; Sauermann and Stephan, 2012). Our data highlight productive
patterns of task allocation in each research phase. In planning, although lab heads are the primary
decision makers (Delamont and Atkinson, 2001; Knorr-Cetina, 1999), members’ participation
does contribute to productivity, possibly because autonomy stimulates members’ intrinsic
motivation (Amabile, 1996; Roach and Sauermann, 2010) and may encourage their effort in later
phases. Comparing basic and applied research, members’ engagement in planning is more
important in basic research due to its exploratory nature and the relevance of intrinsic motivation.
In execution, members are believed to be the primary player because lab heads are too expensive
for labor-intensive tasks (Delamont and Atkinson, 2001; Delamont et al., 1997). However, we
hypothesize that the cost of lab head’s labor can be justified by the benefit of collocation and
technical catching up (Andrews, 1979; Teasley et al., 2002). Our results confirm this hypothesis
in basic laboratories, possibly because sharing workspace with members, having frequent
discussion, and updating research plans in a timely fashion are essential for exploratory research.
In contrast, the cost seems greater than the benefit in applied research, where the research is
more confirmatory and more likely to follow predetermined plans. In writing, the results show
that lab heads but not members should be the primary player, particularly in basic laboratories.
Since basic research is more theory-driven and exploratory (Calvert, 2004), lab heads may better
serve writing tasks with their longer experience and holistic scientific perspective.
Academic laboratories are peculiar in that they are responsible not only for research but
also for education. However, these two goals can be incompatible (e.g., Fox, 1992; Hackett,
1990) and cause a dilemma for lab heads, who may have to prioritize either members’ training or
23
scientific productivity at the cost of the other. Indeed, our results suggest that task allocation
optimized for research productivity can be different from that for education. For example, even
though young members should be taught how to write a paper, our results suggest that it
compromises productivity in basic research. This conflict between research and education may
be becoming more serious as science policies increasingly emphasize research productivity and
researchers have been under pressure for short-term evaluation based on publications. Obviously,
the training of future researchers is indispensable in sustaining the science system. Therefore,
science policies should take actions to better balance research and education.
Interestingly, our results also suggest that research productivity and training members
can be partially compatible by showing that members’ participation in planning improves
productivity. Nevertheless, as the stylized task allocation assumes (Delamont and Atkinson,
2001; Knorr-Cetina, 1999; Latour and Woolgar, 1979), we observe that members are not engaged
in planning in many laboratories. This may be because of the hierarchical lab structure in Japan,
modelled on the German chair system (Arimoto, 2011), where a vertical division of labor is clear,
in comparison to flatter organizations observed in the US and other Western countries (Kneller,
2007). If this is the case, science policies might need to restructure the lab design or to change
employment practices to allow greater participation for junior members. In fact, the Japanese
government has embraced such a vision (Kneller, 2007), but it seems yet to be realized.
5.2. Future directions and limitations
This section discusses some potential directions of future research and limitations of the
current study. First, this study focuses on research orientations as a contextual factor, but other
organizational contexts can affect the optimal form of organizational design. For example, we
also explored the impact of organizational prestige, lab size, and lab age (Heinze et al., 2009;
Levin and Stephan, 1989; Tunzelmann et al., 2003), finding some contingency effects.12
For
example, we find that members’ planning has a stronger effect in prestigious universities than in
non-prestigious universities, probably because the ability of students is correlated with university
prestige due to the nature of the admission system in Japan (Kneller, 2007). We also find that lab
head’s writing is more effective in small laboratories and in old laboratories, especially in basic
research. The former may be because lab heads in large laboratories cannot efficiently handle the
12 Table S3 in Supplementary Results
24
writing task for too many members (Delamont and Atkinson, 2001; Salonius, 2008). The latter
may suggest that accumulated knowledge in senior lab heads lends them advantage in writing.
These results imply that the contingency of task allocation should be investigated with multiple
contextual factors taken into account, for which more robust tests are needed in future research.
Second, although we analyze the contingency of task allocation by examining the
interaction effects of research orientation and task allocation, a different form of contingency is
plausible. That is, if lab heads know the optimal forms of task allocation under a certain context,
they should be able to adapt their organizational design to the context (Drazin and Vandeven,
1985). Tables 1 and 2 imply that this might be the case. For example, we find that members’
writing is more common in applied laboratories than in basic laboratories. Thus, lab heads in
applied laboratories may understand the benefit of members’ writing and actually have them
write. On the other hand, the results also suggest that the productive task allocation is not always
followed. For example, although lab head’s execution increases productivity in basic laboratories
and decreases it in applied laboratories, actual task allocation shows no sign of adaptation. Since
these results are limited due to the nature of our cross-sectional data, future research should draw
on dynamic analyses to further the understanding of adaptation.
Third, this study focuses on the laboratory as a unit of analysis, but a laboratory usually
engages in multiple projects, and project can be regarded as another important unit of analysis in
examining scientific production. It is plausible that task allocation in one project is different from
that in another. We attempt to address some potentially confounding factors attributed to projects
such as project size, external collaboration, and interdisciplinarity.13
Controlling for these factors,
we confirm that our results remain qualitatively unchanged. Nevertheless, future research is
needed to closely inspect the attributes of projects in addition to those of laboratories.
Fourth, since this study draws on a specific sample of Japanese biology laboratories,
further research is needed for generalization. Countries can differ in the degree of emphasis on
the practical application of academic science and in the practice of PhD training, both of which
are related to the focal concepts of this study. For the former, a few studies have shown
comparable figures between Japan and the US regarding academics’ practical orientation (e.g.,
patenting, commercial activities) (e.g., Shibayama et al., 2012). For the latter, statistics of the
OECD (2013) suggest, for example, that comparable proportions of PhD graduates are employed
13
Table S4 in Supplementary Results
25
in industry in the US, Japan, and some European countries. Although these lend some confidence
to the representativeness of our sample, it is still possible that our results are specific to the
context. The generalizability of our results from biology to other fields of science also needs
further investigation.
Fifth, we cannot fully address the issue of endogeneity. It is plausible that task allocation
is determined by the ability of lab heads and members. In this regard, members’ ability is
difficult to measure. From our results, we assume that the pattern of task allocation is determined
by lab head’s policy and preferences14
and is rather stable over time and across projects, but
longitudinal analyses are needed for more robust tests of causality in future research.
Sixth, we made a dichotomous distinction between basic and applied research, but
research area is more continuous in reality. A laboratory can engage in both research areas. In
particular, since recent science policies emphasize practical application (Etzkowitz and
Leydesdorff, 2000), basic laboratories are under pressure to engage also in more applied research.
Stokes (1997) suggested that a basic-applied combinatorial approach (so-called Pasteur’s
Quadrant) is fundamentally different from pure basic and pure applied. In this regard,
project-level analyses may be helpful. Related to this point, our argument is based on the
assumption that research areas are a predetermined context rather than a result of strategic choice.
We believe that this is an acceptable assumption in the short term. However, future research
should investigate the dynamics between organizational structure and the longer-term trajectory
of research areas.
5.3. Concluding remark
This study presents evidence on the division of labor in Japanese biology laboratories,
and shows that while the stereotype of the lab head planning and the members executing is not
rare, other combinations are also quite common. In addition, these “off-diagonal” labs can be
more or less effective than the “normal” structure, depending on the research orientation. These
findings suggest a variety of follow-up questions about optimal task allocation, the need to
balance the sometimes conflicting goals of research productivity and effective education, and the
tensions between intrinsic motivations and efficient production. Our findings also suggest the
14
To investigate lab heads’ training policy and how it affects the pattern of task allocation, we asked several
questions on lab head’s policies regarding student training, and we observed significant correlations between lab
head’s policies and task allocation patterns (Table S5 in Supplementary Result).
26
utility of taking an organizational studies approach to the analysis of scientific labs. Such an
approach can help generate new research questions and contribute important results to the
science of science policy.
Acknowledgments
We acknowledge financial support from the Konosuke Matsushita Memorial Foundation,
Grant-in-Aid for Research Activity Start-up of the Japan Society for the Promotion of Science
(#23810004), and Grant-in-Aid for Scientific Research (B) Program (#20330077) form The
Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan. We
acknowledge the U.S. National Science Foundation and the Patent Board for offering us the
journal classification data. We appreciate thorough and encouraging comments from three
anonymous referees. We acknowledge Ms. Aya Sasaki for technical support.
References
Agarwal, R., Ohyama, A. 2012. Industry or academia, basic or applied? Career choices and
earnings trajectories of scientists. Management Science.
Allison, P. D., Long, J. S. 1990. Departmental effects on scientific productivity. American
Sociological Review, 55: 469-478.
Amabile, T. M. 1996. Creativity in context: Update to the social psychology of creativity.
Boulder, CO: Westview Press.
Andrews, R. 1979. Scientific productivity: The effectiveness of research groups in six
countries. Cambridge: Cambridge University Press.
Arimoto, A. 2011. Japan: Effects of changing governance and management on the academic
profession. In Locke, W. & Cummings, W. & Fisher, D. (Eds.), Changing governance
and management in higher education: The perspectives of the academy: 281-320.
Dordrecht: Springer.
Barber, B., Fox, R. C. 1958. The case of the floppy-eared rabbits: An instance of serendipity
27
gained and serendipity lost. American Journal of Sociology, 64: 128-136.
Becher, T., Henkel, M., Kogan, M. 1994. Graduate education in britain. London: Jessica
Kingsley.
Bertalanffy, L. V., Rapoport, A., Meier, R. L. 1962. General systems yearbook. Washington,
D.C.: Society for General Systems Research.
Calvert, J. 2004. The idea of 'basic research' in language and practice. Minerva, 42: 251-268.
Carayol, N., Matt, M. 2006. Individual and collective determinants of academic scientists'
productivity. Information Economics and Policy, 18: 55-72.
Council, N. R. 1998. Trends in the early careers of life scientists. Washington, DC: National
Academy Press.
Cyert, R. M., March, J. G. 1963. A behavioral theory of the firm. Englewood Cliffs, NJ:
Prentice-Hall.
Delamont, S., Parry, O., Atkinson, P. 1997. Critical mass and pedagogic continuity: Studies in
academic habitus. British Journal of Sociology of Education, 18: 533-549.
Delamont, S., Atkinson, P. 2001. Doctoring uncertainty: Mastering craft knowledge. Social
Studies of Science, 31: 87-107.
Drazin, R., Vandeven, A. H. 1985. Alternative forms of fit in contingency theory. Administrative
Science Quarterly, 30: 514-539.
Etzkowitz, H., Leydesdorff, L. 2000. The dynamics of innovation: From national systems and
"mode 2" to a triple helix of university-industry-government relations. Research
Policy, 29: 109-123.
Ford, C. M. 1996. A theory of individual creative action in multiple social domains. Academy of
Management Review, 21: 1112-1142.
Fox, M. F. 1992. Research, teaching, and publication productivity - mutuality versus competition
in academia. Sociology of Education, 65: 293-305.
Hackett, E. J. 1990. Science as a vocation in the 1990s - the changing organizational culture of
academic science. Journal of Higher Education, 61: 241-279.
28
Hackman, J. R., Oldham, G. R. 1976. Motivation through the design of work—test of a theory.
Organizational Behavior and Human Performance, 16: 250-279.
Heinze, T., Shapira, P., Rogers, J. D., Senker, J. M. 2009. Organizational and institutional
influences on creativity in scientific research. Research Policy, 38: 610-623.
Hollingsworth, R., Hollingsworth, E. J. 2000. Major discoveries and biomedical research
organizations: Perspectives on interdisciplinarity, nurturing leadership, and integrated
structure and cultures. In Weingart, P. & Stehr, N. (Eds.), Practising
interdisciplinarity: 215-244. Toronto, Canada: University of Toronto Press.
Kneller, R. 2007. Prospective and retrospective evaluation systems in context: Insights from
japan. In Whitley, R. & Glaser, J. (Eds.), Changing governance of the sciences: The
advent of research evaluation systems. Dordrecht, Netherlands.: Springer.
Knorr-Cetina, K. 1999. Epistemic cultures : How the sciences make knowledge. Cambridge,
MA: Harvard University Press.
Latour, B., Woolgar, S. 1979. Laboratory life: The construction of scientific facts. Princeton,
NJ: Princeton University Press.
Levin, S. G., Stephan, P. E. 1989. Age and research productivity of academic scientists.
Research in Higher Education, 30: 531-549.
Lindsey, D. 1980. Production and citation measures in the sociology of science - the problem of
multiple authorship. Social Studies of Science, 10: 145-162.
Long, J. S., McGinnis, R. 1981. Organizational context and scientific productivity. American
Sociological Review, 46: 422-442.
Merton, R. K. 1968. The bearing of empirical research on sociological theory, Social theory and
social structure: 156-171. Glencoe, IL: The Free Press.
Merton, R. K. 1973. Sociology of science. Chicago: University of Chicago Press.
Merton, R. K., Barber, E. 2004. The travels and adventures of serendipity. A study in
sociological semantics and the sociology of science. Princeton: Princeton University
Press.
Narin, F., Pinski, G., Gee, H. H. 1976. Structure of biomedical literature. Journal of the
29
American Society For Information Science, 27: 25-45.
Nelson, R. R. 1959. The simple economics of basic scientific-research. Journal of Political
Economy, 67: 297-306.
Nightingale, P. 1998. A cognitive model of innovation. Research Policy, 27: 689-709.
Oecd. 2013. Key findings of the oecd-knowinno project on the careers of doctorate holders, Vol.
FP7-Adhoc-2007-13. Paris: OECD.
Owen-Smith, J. 2001. Managing laboratory work through skepticism: Processes of evaluation
and control. American Sociological Review, 66: 427-452.
Pelz, D. C., Andrews, F. M. 1966. Scientists in organizations. New York: Wiley.
Roach, M., Sauermann, H. 2010. A taste for science? Phd scientists' academic orientation and
self-selection into research careers in industry. Research Policy, 39: 422-434.
Salonius, A. 2008. Working in the lab: Social organization of research and training in
biomedical research labs in canada and its relationship to research funding
MCGILL UNIVERSITY, Quebec, Canada.
Sauermann, H., Cohen, W. M. 2010. What makes them tick? Employee motives and firm
innovation. Management Science, 56: 2134-2153.
Sauermann, H., Stephan, P. E. 2012. Conflicting logics? A multidimensional view of industrial
and academic science. Organization Science, forthcoming.
Seymore, S. B. 2009. Serendipity. North Carolina Law Review, 88: 185-211.
Shibayama, S. 2011. Distribution of academic research funds: A case of japanese national
research grant. Scientometrics, 88: 43-60.
Shibayama, S., Walsh, J. P., Baba, Y. 2012. Academic entrepreneurship and exchange of
scientific resources: Material transfer in life sciences and materials science in japanese
universities. American Sociological Review, 77: 804-830.
Shimizu, H., Nirei, M., Maruyama, K. 2012. Management of science, serendipity, and research
performance: Evidence from scientists' survey in the us and japan, Schumpeter
Conference. Brisbane, Australia.
30
Simon, H. A. 1957. Administrative behavior: A study of decision-making processes in
administrative organrzation. New York: Macmilian.
Stapleton, D. H. 2004. Creating a tradition of biomedical research. Contributions to the history
of the rockefeller university. New York: Rockefeller University Press.
Stephan, P. E. 1996. The economics of science. Journal of Economic Literature, 34: 1199-1235.
Stokes, D. E. 1997. Pasteurs quadrant: Basic science and technological innovation
Washington, D.C.: Brookings Institution Press.
Teasley, S. D., Covi, L. A., Krishnan, M. S., Olson, J. S. 2002. Rapid software development
through team collocation. Ieee Transactions on Software Engineering, 28: 671-683.
Traweek, S. 1988. Beamtimes and lifetimes: The world of high energy physicists. Cambridge,
MA: Harvard University Press.
Tunzelmann, N. V., Ranga, M., Martin, B., Geuna, A. 2003. The effects of size on research
performance: A spru review. Brighton, UK: SPRU.
Van Angel, P. 1992. Serendipity: Expect also the unexpected. Creativity and Innovation
Management, 1: 20-32.
Weesie, J. 1999. Seemingly unrelated estimation and the cluster-adjusted sandwich estimator.
Stata Technical Bulletin, 52: 34-47.
Whitley, R. 1984. The intellectual and social organization of the sciences. New York: Oxford
University Press.
Woodman, R. W., Sawyer, J. E., Griffin, R. W. 1993. Toward a theory of organizational creativity.
Academy of Management Review, 18: 293-321.
Zuckermann, H. 1977. Scientific elite: Nobel laureates in the united states. New York: Free
Press.
Figure 2 Prediction of Task Allocation Effect a
a LH stands for lab heads. Based on regression results (Models 3 and 4 in Tables 3 and 4), we predicted the lab productivity in terms of per-staff citation count
(top row) and pub count (bottom row) for basic and applied laboratories. In each phase, we compare two extreme patterns of task allocation. In planning, since a
lab head usually plays the leading role, a lab head’s solo leading vs. co-leading with members is of the primary interest. Similarly, in execution, members’ solo
leading vs. co-leading with a lab head is of concern. In writing, since a lab head’s and members’ roles are negatively correlated, a lab head’s solo leading vs.
members’ solo leading are compared. For the prediction, the mean values are used for all variables except the focal task allocation variables.
Table 1 Descriptive Statistics and Correlation of Variables a
a N=305. Bold italic: p<0.05.
Mean Std.Dev. Min Max 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 ln(Per-staff pub count) 1.608 .640 .000 3.135
2 ln(Per-staff citation count) 3.273 1.166 .000 6.201 .831
3 ln(Pre-tenure pub count) 1.535 .807 .000 3.714 .466 .438
4 ln(Pre-tenure citation count) 2.808 1.683 .000 7.140 .295 .417 .803
5 Time for research 3.770 1.535 1.000 6.000 -.007 .085 .037 .114
6 Foreign experience 3.633 1.548 1.000 6.000 -.062 -.045 -.106 -.039 .084
7 Inbred .105 .307 .000 1.000 .104 .036 .157 .071 -.067 -.071
8 Medical doctor .174 .380 .000 1.000 -.087 .054 -.032 .115 .068 .120 -.016
9 Female .033 .178 .000 1.000 -.151 -.180 -.086 -.029 .148 .103 -.063 .061
10 Budget/#staff 3.960 2.882 .278 17.500 -.024 .136 -.005 .062 .242 .028 -.045 .083 -.059
11 Lab age 13.214 7.747 1.000 35.000 .146 -.004 .114 -.071 -.165 .016 .384 -.049 -.065 -.140
12 Lab size 9.246 5.204 3.000 45.000 -.012 .173 .036 .132 .075 .059 .027 .265 -0.06 .131 .023
13 Top7 univ .489 .501 .000 1.000 .050 .102 .135 .170 .096 -.039 .136 -.050 -0.03 .116 -.042 .199
14 Basic research .540 .499 .000 1.000 -.132 .033 .013 .116 .166 -.013 .016 -.054 0.02 -.023 -.097 .037 .222
15 Members' planning 1.179 .489 .000 2.000 .055 .132 .035 .084 .006 -.065 .029 -.051 -.087 .095 .065 .132 .150 .035
16 Lab head's execution 1.227 .567 .000 2.000 -.048 .025 -.078 -.067 .005 .115 .080 .130 -.009 .052 .094 .022 .028 .037 .074
17 Members' writing (& lab head's not writing) .985 .407 .000 1.667 .072 -.053 .035 -.055 -.236 -.153 .110 -.154 -.099 -.057 .160 .017 .026 -.163 .359 -.093
Table 2 Patterns of Task Allocation a
a YES: main role and NO: otherwise. We computed the mean of the extent of engagement (0: none, 1: minor, and 2:
main role) in related tasks and ranks, and assigned YES if it is 1.5 or greater. The number of laboratories categorized
into each pattern and the percentage of those laboratories (in parentheses) are shown for the whole sample as well as
for the breakdown by research areas and university ranks.
1 YES - - 99 (32%) 55 (34%) 41 (30%) 55 (37%) 44 (28%)
2 - YES - 148 (49%) 78 (48%) 67 (49%) 72 (48%) 76 (49%)
3 - - YES 163 (53%) 81 (50%) 79 (57%) 80 (54%) 83 (53%)
4 NO NO NO 62 (20%) 30 (19%) 32 (23%) 27 (18%) 35 (22%)
5 NO NO YES 45 (15%) 22 (14%) 23 (17%) 23 (15%) 22 (14%)
6 NO YES NO 61 (20%) 38 (23%) 21 (15%) 31 (21%) 30 (19%)
7 NO YES YES 38 (12%) 17 (10%) 21 (15%) 13 (9%) 25 (16%)
8 YES NO NO 11 (4%) 6 (4%) 5 (4%) 5 (3%) 6 (4%)
9 YES NO YES 39 (13%) 26 (16%) 11 (8%) 22 (15%) 17 (11%)
10 YES YES NO 8 (3%) 7 (4%) 1 (1%) 6 (4%) 2 (1%)
11 YES YES YES 41 (13%) 16 (10%) 24 (17%) 22 (15%) 19 (12%)
Top 7
(N = 149)
Not Top 7
(N = 156)
Research Areas University RankMembers'
planning
Lab head's
execution
Members'
writing
All Sample
(N = 305)Basic
(N = 162)
Applied
(N = 138)
Table 3 Prediction of Scientific Productivity: Dependent variable = ln(Per-staff
citation count) a
a Unstandardized coefficients (standard errors in parentheses). Two-tailed test. † p<0.10; * p<0.05; ** p<0.01;***
p<0.001. Ordinary least squared.
Control variable
ln(Pre-tenure citation count) .249 *** (.039) .228 *** (.039) .215 *** (.057) .224 *** (.055)
Time for research .025 (.045) .024 (.044) .032 (.063) .034 (.067)
Foreign experience -.026 (.043) -.060 (.043) -.109 † (.063) .005 (.062)
Inbred .078 (.230) .151 (.225) .571 † (.330) -.344 (.314)
Medical doctor .064 (.207) .024 (.203) .092 (.305) -.202 (.291)
Female -1.083 ** (.355) -1.026 ** (.345) -1.375 ** (.484) -.673 (.581)
Budget/#staff .011 (.024) .008 (.023) .058 (.037) -.031 (.030)
Lab age .005 (.010) .006 (.010) -.014 (.014) .028 † (.015)
Top7 univ .020 (.135) -.052 (.132) -.304 (.197) .176 (.185)
Basic Research .016 (.151) -.812 † (.485)
Task allocation
Members' planning .241 † (.145) .226 (.210) .476 * (.213) .280 (.203)
Lab head's execution .011 (.117) -.418 ** (.160) .481 * (.189) -.385 * (.155)
Members' writing (& lab head's not writing) -.262 (.181) -.023 (.247) -.527 † (.274) .044 (.244)
Interaction
Members' planning x Basic research .162 (.283)
Lab head's execution x Basic research .895 *** (.232)
Members' writing x Basic research -.478 (.344)
F test 3.273 *** 3.672 *** 2.599 *** 2.701 ***
Log likelihood -405.684 -395.549 -213.330 -161.844
N 292 292 156 136
Basic labs Applied labs
Model 1 Model 3 Model 4Model 2
All labs
Table 4 Prediction of Scientific Productivity: Dependent variable = ln(Per-staff
pub count) a
a Unstandardized coefficients (standard errors in parentheses). Two-tailed test. † p<0.10; * p<0.05; ** p<0.01;***
p<0.001. Ordinary least squared.
Control variable
ln(Pre-tenure pub count) .337 *** (.043) .320 *** (.042) .315 *** (.060) .302 *** (.061)
Time for research .019 (.024) .017 (.023) .019 (.031) .037 (.039)
Foreign experience -.002 (.023) -.020 (.023) -.058 † (.032) .026 (.035)
Inbred .040 (.122) .081 (.119) .276 † (.162) -.193 (.182)
Medical doctor -.073 (.109) -.086 (.107) -.052 (.151) -.167 (.167)
Female -.311 † (.187) -.277 (.182) -.240 (.239) -.236 (.335)
Budget/#staff -.018 (.012) -.019 (.012) -.005 (.018) -.030 † (.017)
Lab age .006 (.005) .007 (.005) -.003 (.007) .020 * (.008)
Top7 univ -.007 (.071) -.047 (.070) -.155 (.097) .057 (.107)
Basic Research -.091 (.080) -.547 * (.255)
Task allocation
Members' planning .055 (.076) -.024 (.111) .219 * (.104) -.037 (.118)
Lab head's execution -.046 (.062) -.247 ** (.084) .173 † (.092) -.197 * (.088)
Members' writing (& lab head's not writing) -.035 (.096) .130 (.130) -.225 † (.135) .177 (.141)
Interaction
Members' planning x Basic research .206 (.150)
Lab head's execution x Basic research .429 *** (.122)
Members' writing x Basic research -.327 † (.182)
F test 4.166 *** 4.542 *** 2.779 *** 3.288 ***
Log likelihood -219.014 -208.888 -103.119 -87.426
N 292 292 156 136
Basic labs Applied labs
Model 1 Model 3 Model 4Model 2
All labs
Appendix 1 Factor Analysis of Task Allocation a
a Relatively large factor loadings, conceptually pertinent to each factor, are indicated in bold italic.
Factor1 Factor2 Factor3 Factor4 Factor5 Factor6
Members'
planning
Junior
researcher’s
full responsibility
Lab head’s
planning
PhD’s
execution
Lab head’s
execution
Members'
writing
Subject -.077 .133 .750 .002 -.032 -.088
Hypothesis -.020 .088 .836 .170 .087 .057
Planning .064 -.062 .579 -.007 .498 -.052
Experiment -.006 .046 -.019 -.055 .857 -.004
Analysis .070 .166 .259 .048 .677 -.125
Writing .167 .232 .348 .120 .282 -.629
Subject .666 .423 .104 -.121 .014 .087
Hypothesis .748 .446 .115 -.071 .015 -.051
Planning .544 .594 .107 .060 -.152 -.034
Experiment -.036 .811 .047 .115 .190 -.122
Analysis .154 .822 .035 .206 .029 -.118
Writing .174 .698 .138 .030 .012 .497
Subject .675 -.047 -.184 .163 .163 .268
Hypothesis .812 -.010 -.090 .323 .043 .104
Planning .698 -.028 -.045 .404 -.048 .117
Experiment .052 .144 .072 .884 .001 .046
Analysis .303 .125 .114 .787 -.035 .175
Writing .261 .003 .065 .273 .013 .825
PhD
Junior
researcher
Lab head
Top Related