Research on processes and practices around information systems is often best conducted in naturalistic setting. To conduct valid and reliable research in such settings, researchers must find ways to reliably bound the phenomenon in which they are interested. In this paper we propose that researchers use episodes—events or processes occurring over a specified period of time—to isolate that which interests them from the vast set of related human behavior. The paper discusses the nature of episodes in the literature and suggests particular research settings in which episodes can be useful. The paper describes a three stage methodology to identify episodes for systematic data collection and analysis. The paper presents an example study using episodes to study group learning process in distributed groups.