Interpretation biases are a core feature of social anxiety. Used to automatically train positive interpretation biases, the existing versions of Cognitive Bias Modification for Interpretation (CBM-I) render mixed results. The current study presents the development and testing of a one-session novel adaptation of CBM-I for social anxiety that provides an adaptive instead of a positive meaning to socially loaded situations. Informed by Rational-Emotive Behavior Therapy and latest theoretical upgrades, we argue that biases exist on a continuum from adaptive to maladaptive and defining them by valence may fuel unrealistic positive illusions. A hundred forty volunteers with high social anxiety were randomized to the CBM-I or to the sham-training condition. Baseline, post-training, and during impromptu speech assessments of self-reported anxiety were made, while adaptive interpretation bias was measured at baseline and post-training. First, near-transfer effects in favor of the CBM-I group were found. Second, no far-transfer effects were found regarding the anxiety experienced during the speech. Third, a main finding was that changes in interpretation bias fully mediated training effects on anxiety. To our knowledge, this is the first CBM-I study to evidence that experimentally induced changes in interpretation mediate changes in anxiety. The therapeutic potential of this new adaptation of CBM-I requires testing in clinical samples.
