The present study was designed to examine the effects of training in positive interpretations in clinically anxious children. A total of 87 children between 7 and 12 years of age were randomly assigned to either a positive cognitive bias modification training for interpretation (CMB-I) or a neutral training. Training included 15 sessions in a two-week period. Children with an interpretation bias prior to training in the positive training group showed a significant reduction in interpretation bias on the social threat scenarios after training, but not children in the neutral training group. No effects on interpretation biases were found for the general threat scenarios or the non-threat scenarios. Furthermore, children in the positive training did not self-report lower anxiety than children in the neutral training group. However, mothers and fathers reported a significant reduction in social anxiety in their children after positive training, but not after neutral training. This study demonstrated that clinically anxious children with a prior interpretation bias can be trained away from negative social interpretation biases and there is some evidence that this corresponds to reductions in social anxiety. This study also highlights the importance of using specific training stimuli.