Graduate teaching assistants (GTAs) often lead laboratory and tutorial sections in science, technology, engineering, and mathematics (STEM), especially at large, research-intensive universities. GTAs' performance as instructors can impact student learning experience as well as learning outcomes. In this study, we observed 11 chemistry GTAs and 11 physics GTAs in a research-intensive institution in the southeastern USA. We observed the GTAs over two consecutive semesters in one academic year, resulting in a total of 58 chemistry lab observations and 72 physics combined tutorial and lab observations. We used a classroom observation protocol adapted from the Laboratory Observation Protocol for Undergraduate STEM (LOPUS) to document both GTA and student behaviors. We applied cluster analysis separately to the chemistry lab observations and to the physics combined tutorial and lab observations. The goals of this study are to classify and characterize GTAs' instructional styles in reformed introductory laboratories and tutorials, to explore the relationship between GTA instructional style and student behavior, and to explore the relationship between GTA instructional style and the nature of laboratory activity.