Recent evidence suggests that video feedback helps improve the accuracy of self-ratings of performance in speech-anxious participants (e.g., Rapee & Hayman, 1996. Behaviour Research and Therapy, 34, 315-322.). Evidence also suggests that this effect is stronger for participants who have a more negatively distorted impression of their performance (self-observer discrepancy; Rodebaugh & Chambless, 2002. Cognitive Therapy and Research, 26, 629-644.). Data collected in Rapee and Hayman's study were analyzed to determine if similar results would be found in this independent sample. Evidence was found for moderating effects of self-observer discrepancy comparable to those shown in Rodebaugh and Chambless. The self-observer discrepancy itself showed relatively weak relationships with general indices of social anxiety, but a very strong relationship with initial self-rating of performance. The results suggest that self-observer discrepancy is a predictor of response to video feedback, and that clinicians may be able to estimate the discrepancy by examining self-rating in conjunction with behavior.