The guidance hypothesis (Schmidt, 1991) predicts that the guiding properties of augmented feedback are beneficial for motor learning when used to reduce error, but detrimental when relied upon. Therefore, a heavily guiding form of feedback might be detrimental for learning. In addition, the guidance hypothesis predicts that practice with a high relative frequency of augmented feedback would be detrimental for learning. An experiment is described that crossed two forms of feedback with two levels of relative frequency. Subjects practiced movements to a target with either physical guidance or knowledge of results, and with either a high or faded relative frequency. The high frequency physical guidance condition resulted in the poorest retention, and both high frequency feedback conditions resulted in the least accuracy in transfer. These results provide support for the guidance hypothesis and suggest consideration of the combined effects on learning of the type and relative frequency of augmented feedback and acquisition-test conditions.