Despite much research on eyewitness confidence, we know very little about whether confidence ratings given in public might differ from those held privately. This study tested a prediction derived from self-presentation theory that eyewitnesses will give lower confidence ratings in public when there is a possibility of their account being contradicted by other witnesses as compared to when they report their confidence in private. In groups of 3 or 4 people, 96 participants watched a videotape of a simulated robbery and then answered 16 forced-choice questions about details from the videotape. In half of the experimental sessions, the participants shared their answers and confidence ratings aloud with the other participants (public condition), and in the other half, the answers and ratings were not shared (private). As predicted, confidence ratings were significantly lower in the public condition than in the private condition, but the privacy manipulation had no effect on response accuracy. These results are consistent with a self-presentation explanation, and they highlight the need to examine public confidence ratings more thoroughly.