Negative interpretation biases, defined as a tendency to interpret ambiguous social situations negatively, have been theorized to play a role in the maintenance of social anxiety. Research has shown that computer-based interpretation training tasks can modify negative interpretation biases and that this modification is associated with decreases in subjective ratings of anxiety. Negative interpretation biases have also been shown to decrease following cognitive-behavioural therapy. This study investigated the effects of interpretation training and cognitive restructuring on symptomatology, cognitive processes, behaviour, and physiological reactivity in an analogue social anxiety sample. Seventy-two participants with elevated social anxiety scores were randomized to one of 3 conditions: interpretation training (n = 24), cognitive restructuring (n = 24), and control (n = 24). Although none of the conditions showed a decrease in social anxiety symptomatology, participants in the cognitive restructuring condition evidenced a significant decrease in anxiety-related cognitive processes at the 48-hour follow-up. There were no group differences on subjective distress and self-rated performance on the speech task. However, participants in the cognitive restructuring condition were rated as having higher quality speeches by an objective rater compared to participants in the interpretation training condition. Theoretical and clinical implications are discussed.