Mixed Reality (MR) systems enable users to perform augmented superpowers that transcend real-world limitations. However, it remains unclear what types of action-outcome mappings can enable users to easily learn, control, and feel a sense of ownership of these augmented superpowers. Humans develop a set of sensorimotor regularities (i.e., image schemas and lawful relations between them) from recurring bodily experiences since early infancy, and use them to predict the outcome of our actions, or choose actions based on the desired outcome. We investigate whether sensorimotor regularities (SRs) can serve as effective design constraints for superpower interactions, by comparing three temporal manipulation methods in MR games: (1) mid-air button control
(2) gestures incongruent with SRs embedded in the concept of temporal manipulation
and (3) gestures congruent with these SRs. A within-subject study with 18 participants reveals that the SRs-congruent method enables significantly improved task performance, lower overall workload, and a greater sense of agency and presence compared to both an SRs-incongruent method and a mid-air button-based method. The SRs-congruent method also enabled faster mastery of the augmented superpower. No significant difference was observed in any of the above-mentioned metrics between the SRs-incongruent and mid-air button-based methods. These results empirically demonstrate multiple benefits of SRs as design constraints for superpower controls in MR, and encourage future research to explore their wider applicability in superpower interaction design.