Medical Education 2010: 44: 765-774 Objectives Educators often speak of web-based learning (WBL) as a single entity or a cluster of similar activities with homogeneous effects. Yet a recent systematic review demonstrated large heterogeneity among results from individual studies. Our purpose is to describe the variation in configurations, instructional methods and presentation formats in WBL. Methods We systematically searched MEDLINE, EMBASE, ERIC, CINAHL and other databases (last search November 2008) for studies comparing a WBL intervention with no intervention or another educational activity. From eligible studies we abstracted information on course participants, topic, configuration and instructional methods. We summarised this information and then purposively selected and described several WBL interventions that illustrate specific technologies and design features. Results We identified 266 eligible studies. Nearly all courses (89%) used written text and most (55%) used multimedia. A total of 32% used online communication via e-mail, threaded discussion, chat or videoconferencing, and 9% implemented synchronous components. Overall, 24% blended web-based and non-computer-based instruction. Most web-based courses (77%) employed specific instructional methods, other than text alone, to enhance the learning process. The most common instructional methods (each used in nearly 50% of courses) were patient cases, self-assessment questions and feedback. We describe several studies to illustrate the range of instructional designs. Conclusions Educators and researchers cannot treat WBL as a single entity. Many different configurations and instructional methods are available for WBL instructors. Researchers should study when to use specific WBL designs and how to use them effectively. © Blackwell Publishing Ltd 2010.