scientistsPhoto courtesy of Rensselaer Polytechnic InstituteThe scientist who ventures away from her original research interest is far rarer than those who stay in familiar territory, according to new research in the emerging field of “the science of research science” published in Nature Human Behavior. The quantitative analysis reveals three key patterns: few scientists move away from their original research focus, those who move do so gradually, and even new interests tend to incorporate favored topics from earlier in their career.

“In science there is room for people who focus their whole life in one direction, and for those who move into different areas. And yet, not surprisingly, the larger the extent of change over their career, the smaller the percentage of people who are doing it,” said Boleslaw Szymanski, the senior author of the research and director of the Network Science and Technology (NeST) Center at Rensselaer Polytechnic Institute.

The science of research science seeks to better understand the scientific enterprise, improving the quality of results and the pace of innovation.

“There are enormous challenges to overcome for scientists to move between disciplines. With this work, we can begin to identify the barriers, and work to better manage science and the direction of innovation,” said Szymanski, the Claire and Roland Schmitt Distinguished Professor of Computer Science at Rensselaer.

To track research interests over time, Szymanski and his co-authors looked at research articles published in American Physical Society (APS) journals over 30 years, a collection that included 14,715 scientists, each of whom published more than 16 papers, for a total of about 500,000 articles. Each article included topical tags from the Physics and Astronomy Classification Scheme, which identifies 10 main topics and 67 areas within those topics. Authors tag each article with a topic and as many areas as they choose, but on average there are three tags per paper. In the analysis, Szymanski and his team chose the first three tags, creating an ordered list, or “topic tuple,” that represented the research interests of the author at the time of publication.

With this information, the team was able to track the extent to which research interests shifted over the course of a career and how quickly they shifted, and prepared a model to depict that shift.

The analysis showed that most scientists retained remarkably homogenous research interests throughout the course of their career. In choosing topics, scientists are more likely to select more recently used topics than those used long ago. And changing interests tend to happen gradually, with more elements of a new topic preserved than replaced. Looking at the frequency with which topics appear in a publication sequence, the analysis found that a small number of topics appear frequently while most subjects appear only a few times, a pattern described mathematically with a “power law distribution.”

While scientists who strayed far from their original research focus were rare, they were equally as productive as those who stayed close to home.

The team also created a series of models to test assumptions about changes in research interest, and found that the changes in research interest were best modeled by an unbiased random walk along a line of topics, with an equal chance for a scientist to move away from or back to the starting point. Also, scientists who shifted interest over time did so gradually, with more elements in a given topic tuple preserved than changed at each point.

Quantifying patterns of research-interest evolution” can be found using DOI:10.1038/s41562-017-0078. Szymanski was joined in the research by Tao Jia, a former postdoctoral research associate in the NeST Center who is now on the faculty of Southwest University in Chongqing, China, and Dashun Wang of Northwestern University.