MOOCs (Massive Open Online Courses) were first introduced to the wider public in 2008, with the first language MOOCs appearing in 2012. Following the initial hype, a number of problems with the way MOOCs had been conceived and implemented have emerged from research and practical experiences. In this article we revisit some of the arguments for and against MOOCs, specifically for language education, and review some of the ways new forms of online learning environments are emerging, as well as new ways of using (elements of) MOOCs, for both teaching and research purposes. In particular, we focus on their potential for the collection, analysis and pedagogical application of large data sets through learning analytics and educational data mining. We argue that hybrid forms of online environments that better foreground social aspects of learning and that take better account of individual differences, have the potential to successfully support language learning on a large scale and to provide researchers and practitioners with unique insights into the language learning process.