Purpose: This study examined whether the benefit of self-controlled feedback learning is induced by learner's spontaneous error estimation. Method: In the experiment I, 24 adult participants (21.6±2.5 yrs.) were randomly assigned to (1) self-controlled feedback with error estimation group or (2) self-controlled feedback without error estimation group. Self-paced straight-arm outward movement was used as experimental task. The acquisition phase consisted of 80 trials. Participants in the self-controlled feedback with error estimation group had to estimate performance after each trial, and 22 trials were administered for delayed retention test (24 hours later). In the experiment Ⅱ, 22 adult participants (22.5±2.0 yrs.) were randomly assigned to yoked groups either with or without error estimation. The task and procedure were identical to the experiment Ⅰ except that participants could not control the feedback schedule. Absolute error and variable error were adopted as dependent variables. Between-subjects design t-test was utilized for statistical analyses. Results: In experiment Ⅰ, requesting self-controlled feedback participants to estimate their performance following each movement showed no advantage in motor learning. In experiment Ⅱ, however, error estimations did facilitate motor learning of movement accuracy in yoked group (p＜.05, d=0.91). It was suggested that spontaneous error estimations took place prior to self-controlled feedbacks, which was the reason that augmented error estimations could not be beneficial to motor learning during practice. Conclusion: Spontaneous error estimation was the critical factor that self-controlled feedback facilitated motor learning.