Instructional design theories such as the "cognitive load theory" (CLT) or the "cognitive theory of multimedia learning" (CTML) explain learning difficulties in (computer-based) learning usually as a result of design deficiencies that hinder effective schema construction. However, learners often struggle even in well-designed learning environments. In this theoretical paper, I will argue that cognitive resources-oriented theories such as the cognitive load theory might profit from extending their predominantly cognitive focus to one that additionally considers metacognitive and self-regulation demands. Empirical results on learning from multiple external representations and research on tool use are integrated to illustrate that computer-based learning environments usually pose a variety of cognitive, metacognitive and self-regulatory demands on learners which require knowledge and skills that learners often lack. Specifically, empirical findings suggest that most learners are unable to regulate their learning automatically. I thus argue that these activities consume working-memory resources as do activities that are closely related to schema construction. My article concludes with suggestions on how the concept of metacognitive load can be incorporated into resource-oriented theories such as the CLT to explain a wider variety of phenomena.