In 2 experiments, the authors assessed different knowledge of results (KR) schedules for observational and physical practice. In Experiment 1, participants had to learn a sequence timing task under either a bandwidth (KR being delivered when participants' performance was outside a predefined bandwidth or range) or yoked (same number of KRs provided as the bandwidth group) KR procedure. The results show that for both practice conditions the bandwidth KR schedule was more effective in promoting learning than the yoked schedule. During Experiment 2, a KR frequency was controlled (100% or 33% KR) and the data indicate that a reduced KR frequency only enhanced the learning of observers. Because a low KR frequency improves the sensory process controlling motor learning, the authors propose that action observation may be perceptual in nature.