According to the Cue-Belief Model, when we assess our own confidence, we employ self-credibility cues that reflect beliefs in our own memory faculties. We tested the influence of meta-memory feedback on self-credibility cues in the context of eyewitness testimony, both before testifying (Experiment 1) and after testifying (Experiment 2). In the present study, participants received feedback (good score vs. bad score vs. no feedback) about a fictitious scale purported to predict their memory as an eyewitness. Those who received good score feedback before testifying were more confident (but no more accurate) than those who received bad score feedback. Good score feedback also increased confidence and accuracy after testifying, but only on leading questions. These differing effects of meta-memory feedback on confidence for normal and leading questions is not explained by the Cue-Belief model. Implications for our confidence judgment process are discussed.