Background: Interpretation training programs, in which individuals are trained to interpret ambiguous scenarios in either a negative or benign way, have proven effective in altering anxiety-related cognitive biases in both children and adults. Aims: The current study investigated whether the effects of the interpretation training procedure in children are differentiated according to the mode of presentation of the training. Method: Ninety-four primary school children (aged 10–12 years) scoring above the mean on a social anxiety scale were randomly allocated to four groups, in which they were trained using written or spoken presentation of training materials in either the negative or benign direction. Results: For the negative training, children who heard the training material spoken aloud (spoken presentation) made more negative interpretations of ambiguous social events, compared to children who read the training material (written presentation). However, for the benign training, there was less clear evidence for a differentiation of the effects between the two modes of presentation, although children in the spoken presentation group performed better in a stressful task and showed a trend to rate their mood as more positive after the task than children in the written presentation group. Conclusions: These results not only forward our understanding of the mechanism of the genesis of cognitive bias in children, but also highlight the need for further investigation of how to optimize the effectiveness of interpretation training in children.