ABSTRACT: Whereas movement-based STEM learning activities garner increasing interest among designers, researchers, and policy makers, much remains unknown regarding parameters of movement-based activity design affecting learning quality. One such parameter is task-initiation, namely the questions of who decides what should be accomplished with the resources—the designer or the student—and how movement-based STEM learning programs accommodate student choice during task-initiation. In this theoretical paper, we draw on an embodied design theoretical framework to lay out the issue of task-initiation by presenting and comparing two movement-based STEM programs. In both activities, students first perform a task and then model their performance as instantiating STEM concepts, but the programs differ with respect to task-initiation. In one program, the Mathematics Imagery Trainer for Proportion, students learn to perform pre-determined motor-control tasks by developing new perceptuomotor coordinations for enacting goal movements. In another program, Playground Physics, students use real playground equipment, such as a swing, and virtual playground play performances in the app to determine their own task, such as swinging as high as possible. As such, task-initiation design considerations are tightly related to designers’ overall rationales that, in turn, emanate from assumptions concerning, for example, the epistemic constitution of STEM content, the affective allure of STEM practice, manifestations of agency in STEM problem solving, and other contextual details, such as logistical, architectural, institutional, and curricular constraints governing student and teacher experience.