Scientists’ research topic choices shape individual careers, yet it is unclear how these choices connect to academic performances. We complement the conventional measure of topic switch frequencies with switch distances, and divide the scientists into four groups accordingly. We discover that ‘conservative adventurers’ who switch topics frequently but do so to ‘close’ domains have notably better future performance and can be identified at a remarkably early career stage. Though most scientists show stable behavioral patterns throughout their careers, the rare ones who drastically change to ‘conservative adventurers’ from the opposite group witness considerable future improvements. Our findings are based on three datasets covering a total of 31,780,857 papers in biomedicine, physics, and chemistry, from 1800 to 2021, and shed light on understanding and planning of scientific careers, especially for junior scientists.
The APS dataset includes information of all papers published in the Physical Review series of journals between 1893 and 2020---more than 670,000 publications. Each paper is associated with up to five 6-digit PACS codes.
The PubMed dataset, constructed from the Microsoft Academic Graph (MAG) by requiring the URLs of the papers to be from PubMed, contains 13 billion papers from 1800-2021.
The ACS dataset comprises chemical publications published in 86 American Chemical Society journals from 1879 to 2021, and is constructed by requiring MAG that the papers' URLs come from ACS.