Recent studies examining the role of self-controlled feedback have shown that learners ask for feedback after what they believe was a "good" rather than "poor" trial. Also, trials on which participants request feedback are often more accurate than those without feedback. The present study examined whether manipulating participants' perception of "good" performance would have differential effects on learning. All participants practiced a coincident-anticipation timing task with a self-controlled feedback schedule during practice. Specifically, they were able to ask for feedback after 3 trials in each of three 10-trial practice blocks. While one group (Self-30) was told that an error of 30 ms or less would be considered good performance, another group (Self-4) was informed that an error of 4 ms or less would be considered a good trial. A third, self-control group (Self) did not receive any information about what constituted good performance. The results showed that participants of all groups asked for feedback primarily after relatively good trials. At the end of practice, both the Self-30 and Self groups demonstrated greater perceived competence and self-efficacy than the Self-4 group. The Self-30 and Self groups also performed with greater accuracy and less variability in retention and transfer (non-dominant hand) 1 day later. The present findings indicated that the typical learning benefits of self-controlled practice can be thwarted by depriving learners of the opportunity of experiencing competence through good performance. They add to the accumulating evidence of motivational influences on motor learning.