This paper presents a personal account of developments in research on online learning over the past 30 years. Research on how to design online instruction represents an example of applying the science of learning to education. It contributes to the science of learning (as exemplified by developments in cognitive load theory, the cognitive theory of multimedia learning, and incorporating metacognitive, motivational, and affective aspects of learning), the science of instruction (as exemplified by the continuing development of research-based principles of instructional design), and the science of assessment (as exemplified by supplementing self-report surveys and retention tests with multilevel transfer tests, log file data during learning, and cognitive neuroscience measures of cognitive processing during learning). Some recurring themes are that learning is caused by instructional methods rather than instructional media, so research should focus on features that are uniquely afforded by digital learning environments; instructional practice should be grounded in rigorous and systematic research, including value-added experiments aimed at pinpointing the active ingredients in online instruction; research in online learning should identify boundary conditions under which instructional techniques are most effective; and research in online learning should test and contribute to learning theory.