The integration of ideas, methods, and data from diverse disciplines has been a transformative force in science and higher education, attracting policy interventions, program innovations, financial resources, and talented people. Much energy has been invested in producing a new generation of scientists trained to work fluidly across disciplines, sectors, and research problems, yet the success of such investments has been difficult to measure. Using the Integrative Graduate Education and Research Training (IGERT) program of the U.S. National Science Foundation as a (...) strategic research site, we conducted an experiment to determine whether and how the process and products of research of IGERT-trained scientists differ from those of scientists trained in disciplinary graduate programs. Among scientists in the early years of graduate study we found substantial and consistent differences suggesting that interdisciplinary training improved the quality and process of research, but this pattern was equally strongly reversed among students in the latter years of graduate study. Using systematic observation and other data we suggest why this might be so, then discuss the implications of these results for the design and conduct of graduate education and research. (shrink)