This review of research explores characteristics associated with massive open online courses (MOOCs). Three key characteristics are revealed: varied definitions of openness, barriers to persistence, and a distinct structure that takes the form as one of two pedagogical approaches. The concept of openness shifts among different MOOCs, models, researchers, and facilitators. The high dropout rates show that the barriers to learning are a significant challenge. Research has focused on engagement, motivation, and presence to mitigate risks of learner isolation. The pedagogical structure of the connectivist MOOC model (cMOOC) incorporates a social, distributed, networked approach and significant learner autonomy that is geared towards adult lifelong learners interested in personal or professional development. This connectivist approach relates to situated and social learning theories such as social constructivism (Kop, 2011). By contrast, the design of the Stanford Artificial Intelligence (AI) model (xMOOC) uses conventional directed instruction in the context of formal postsecondary educational institutions. This traditional pedagogical approach is categorized as cognitive-behaviorist (Rodriguez, 2012). These two distinct MOOC models attract different audiences, use different learning approaches, and employ different teaching methods. The purpose of this review is to synthesize the research describing the phenomenon of MOOCs in informal and postsecondary online learning. (Abstract by authors)