Graduate teaching assistants (GTAs) in the sciences are a common feature of U.S. universities that have a prominent mission of research. During the past 2 decades, increased attention has been paid to the professional development of GTAs as instructors. As a result, universities have created training programs to assist GTAs in selecting instructional methods, curricular formats, and assessments when they serve as laboratory, lecture, or discussion group instructors. Unfortunately, few studies explore the educational and instructional environment of GTAs in these reformed settings. This study was conducted to address this specific need. As a constructivist inquiry, qualitative methods were used to collect and analyze the data to elucidate the educational and instructional environment of science GTAs at a doctoral/research university in which various training programs existed. We found that GTAs worked autonomously, that traditional practices and curricula existed in laboratories, and that instructors frequently held limited views of undergraduates' abilities and motivation. Findings in this initial study about GTAs suggest that developers of GTA training programs draw on the literature regarding science teacher education, and that reward systems be instituted that recognize faculty and staff for their participation in GTA training programs.