A traditional and enduring image in science is that of research as an individual endeavor, with lone scientists pushing the boundaries of knowledge in projects of their own. These days, however, scientific projects have become more and more of a team effort. For one measure, the fraction of science and engineering papers written in teams has steadily increased between 1955 and 2000, from about 50% to 80%, with the average number of co-authors nearly doubling during the same period.
Sociologists John P. Walsh and You-Na Lee of the School of Public Policy at the Georgia Institute of Technology in Atlanta wondered what this increase in team size meant for the way that scientific projects are conducted. They found that, today, it is common for research project teams to use practices such as division of labor, task standardization, and hierarchies within teams that make them akin to small factories for knowledge production—and that the size of the team predicts the extent to which such factory-like practices are implemented.
Increasingly … there may be less demand for integrated scientists and more demand for highly-specialized subscientists who can participate in group research as … efficient member[s] of the team.
These results have implications for the training of young scientists and for their academic career prospects, Walsh and Lee argue. Consider that many graduate students and postdocs hope to pursue a career as an academic principal investigator (PI), a role that requires a great diversity of skills. If during their training they are part of a team in which work is highly divided and each participant is a specialist rather than a generalist, then “we are failing to train them for the job they are trying to have (PI). And the jobs we are training them for”—essentially research specialists—“are marginalized” in the current academic research structure, Walsh and Lee write in an email to Science Careers.
Walsh and Lee analyzed the organization of work within research projects across a broad range of fields and institutions. Drawing from the Institute for Scientific Information Web of Science database, they conducted a survey among U.S. scientists who had published a research article between 2001 and 2006, inquiring about the project, the size of the team, and how the work was organized. Altogether, they obtained 2327 responses, of which about half—1223—were from authors based in a university or hospital whose project involved at least one other team member. The authors subsequently focused on those teams, publishing the results online last month in the journal Research Policy.
The survey revealed that, on average, each research team had seven members, but the paper’s list of co-authors excluded three of the contributors, including postdocs, Ph.D. candidates, undergraduate students, and technicians. “[J]ust under half the members of the teams were non-authors, suggesting that authorship credit, while critical for the reputation system of science, may not be accounting for many members of the team,” Walsh and Lee write in an email to Science Careers.
To measure division of labor, Walsh and Lee asked respondents to describe the extent to which “[t]he project involved a strict division of labor [within the lab] with each person responsible for a specific part of the research.” Around three-quarters of the respondents reported some degree of such labor division. Walsh and Lee also found frequent evidence of task “standardization,” which they defined as the presence of supervisory practices encouraging team members to adhere to standard protocols and schedules. The practice of checking graduate students’ notebooks was mentioned in 46% of the projects, with another 66% holding lab meetings. In addition, 61% of the respondents said there was some degree of hierarchical reporting within the team, and almost all of them reported that the project was done with some form of management. Despite these hierarchies, some degree of “decentralization” could be found in more than half of the projects, as measured by the ability of graduate students to make changes of their own to research protocols.
These characteristics—division of labor, standardization of tasks, hierarchy within teams, and decentralization of the decision-process—are all telltale signs of the “bureaucratic” structure typically adopted by firms and government agencies, as defined by organization theory in the early 1970s, Walsh and Lee explain in the paper.
The authors next analyzed the effect of team size on the way the projects are organized and found that the larger the team, the more it functions along bureaucratic lines. They found, for example, that for a chemistry project in a private university increasing from three to seven team members, there is a 7% increase in the probability of being run with internal division of labor, a 13% increase in the probability of standardizing through regular notebook checks, a 12% increase in the probability of employing hierarchical reporting, and a 11% increase in the probability of showing decentralized decision-making. “Our results suggest that the increasingly common large research groups in science have more bureaucratized work organization,” the authors conclude in the study.
Walsh and Lee argue in their paper that the growth of team science and the accompanying trend toward bureaucratization, compounded by the increasing emphasis on lab productivity and the concentration of research resources around major funding initiatives or expensive equipment, may have far-reaching consequences for the training and employment of young scientists.
Traditionally, scientists have been trained using a craft model, under which aspiring scientists learn to become fully fledged scientists through years of apprenticeship under a master craftsman and further exposure to all aspects of research during their postdoc years, Walsh and Lee write in their paper. But, as science shifts toward bureaucratization, trainees may be pushed into premature specialization. Young scientists “might spend much of graduate school optimizing computer code for a large physics experiment, or extracting samples in a biology lab, or doing the statistical analyses on other people’s data,” Walsh and Lee write in their email. Such training best prepares young scientists to fill specialist positions, such as technician, permanent postdoc, and staff scientist—not necessarily to perform all the tasks needed to conduct studies of their own and eventually become a PI.
A shift toward bureaucratization may impact the job market, too. “Increasingly… there may be less demand for integrated scientists and more demand for highly-specialized sub-scientists who can participate in group research as an efficient member of the team,” Walsh and Lee write in their email. The problem, they continue, is that such specialist jobs tend to be few and undervalued. “[T]he systems for allocating career slots, reputations, prestige and other resources do not yet easily accommodate these roles.” Still, it is the preferred career destination for some scientists, Walsh and Lee say, and according to a recent Nature poll, many researchers see the creation of staff scientist positions as the best way to ease the employment bottleneck postdocs currently face. For this to happen, “[w]e need a means for incorporating these types of specialists into the universities and into the funding stream so they are not so vulnerable or so marginalized, so they can make careers out of their specialist contributions to a team, or a series of teams,” Walsh and Lee write.
“It has been pointed out before that sheer size can lead to labs that need lots of postdocs to do the work, without a career path for them. This paper shows how the way work is organized can make this problem worse,” Sara Kiesler, a professor of computer science and human-computer interaction at Carnegie Mellon University in Pittsburgh, Pennsylvania—who formerly supervised Walsh as a postdoc, but was not involved with the study—writes in an email to Science Careers.
One alternative option to creating staff scientist positions, Walsh and Lee argue in their email, would be to resist the current pressures toward the bureaucratization of science. “[W]e can try to retain a more traditional, craft-based model, and we can then modify the funding, labor market and evaluation systems to match,” they write. For example, the calls from part of the scientific community for the National Science Foundation and National Institutes of Health to offer more “fellowships rather than project-based research assistant and post-doc funding are attempts to maintain a more master-apprentice type system.” Other necessary changes would include reducing the pressure on PIs to be exceptionally productive and to get the next publication out, and the next grant in, just to keep their science and large teams going. Relieving this stress would give “more time for trainees to learn, and PIs to teach, the broad set of skills needed to be a fully-integrated scientist.”
Meanwhile, young scientists can take some control over shaping their own training by setting priorities in their career development and choosing their lab accordingly, Walsh and Lee say. “PIs vary in their reputations for [favoring] a more craft model emphasizing training experiences versus a more bureaucratic model emphasizing productivity. Thus, PhD students and post-docs might want to choose labs based on whether they want broader training, or they want faster publication.”
The trouble, though, is that both of these attributes are important for obtaining and excelling in a PI role, especially given the current dearth of job openings for PI positions. Although trainees may be tempted to prioritize immediate productivity over the effort required to develop a well-rounded knowledge base, Walsh and Lee caution that this approach is short-sighted and may not serve aspiring PIs well. “[B]ecoming trained in the full bundle of tasks in your craft is important if you want to have the ability to become a fully-integrated scientist,” they write.
The opportunity to work together with other scientists is another consideration to factor in when choosing a lab, argues Nicolas Carayol, a professor of economics at the University of Bordeaux in France, who was not involved with the study. The results suggest that, “[a]s a Ph.D. or as a postdoc choosing a project, I should take into account the way the projects are organized, so that I will be able to acquire expertise, but also so that I will not to be too isolated,” Carayol writes in an e-mail to Science Careers.
PIs, too, have a responsibility to offer their lab members an adequate training experience. They must find the right balance between breadth and specialization when training their students and postdocs, Carayol argues. Although specialization remains a key factor to trainees’ success, “[t]he paper tells me as a PI ... I should keep in mind not to specialize ... my students and postdocs [too much], to train them well.”
In their email, Walsh and Lee encourage PIs to make conscious decisions about how they go about training the young scientists in their labs. “Concerns about productivity may push labs toward more bureaucratic structures. … But, university labs have the dual function of producing science and producing scientists” who are fully trained to become future PIs. “There may be tradeoffs in these two goals, and PIs may want to think about the tradeoffs when organizing their labs.”