Retrieval practice has been widely studied as an effective strategy for enhancing memory. In this review article, we discuss how its effects on learning complex problem-solving procedures are less straightforward, however, with repeated studying of worked examples sometimes more effective than problem-solving practice. This worked example benefit has been interpreted within the framework of cognitive load theory. In both memory-based tasks and problem-solving tasks, students rate retrieval as more effortful, and less effective for learning, than repeated study. Self-regulated learning decisions do not align with the evidence about the effectiveness of retrieval, as students often avoid using retrieval in memory-based tasks but frequently use it in more complex problem-solving tasks. Patterns associated with self-report survey data and self-regulated learning decisions suggest that retrieval may be used primarily as a means of checking knowledge, and the effort experienced during retrieval may drive subsequent study decisions (i.e., the choice to engage in retrieval vs. further study) to the extent that the experienced effort is interpreted as a sign that learning has been ineffective. We discuss implications of students' views about the purpose of retrieval for effective monitoring and regulation of effort during learning, and propose interventions that may improve students' optimal use of retrieval in their study decisions.