Reduction of costly safety behaviors after extinction with a generalization stimulus is determined by individual differences in generalization rules.

Exposure-based treatment involves repeated presentation of feared stimuli or situations in the absence of perceived threat (i.e., extinction learning). However, the stimulus or situation of fear acquisition (CS+) is highly unlikely to be replicated and presented during treatment. Thereby, stimuli that resemble the CS+ (generalization stimuli; GSs) are typically presented. Preliminary evidence suggests that depending on how one generalizes fear (i.e., different generalization rules), presenting the same GS in extinction leads to differential effectiveness of extinction learning. The current study aimed to extend this finding to safety behaviors. After differential fear and avoidance conditioning, participants exhibited discrete generalization gradients that were consistent with their reported generalization rules (Similarity vs Linear). The Linear group showed stronger safety behaviors to a selected GS compared to the Similarity group, presumably due to higher threat expectancy. After extinction learning to this GS, the Linear group exhibited stronger reduction in safety behaviors generalization compared to the Similarity group. The results show that identifying distinct generalization rules allows one to predict expectancy violation to the extinction stimulus, in addition to corroborating the idea that strongly violating threat expectancy leads to better extinction learning and its generalization. With regard to clinical implications, identifying one's generalization rule (e.g., threat beliefs) help designing exposure sessions that evoke strong expectancy violation, enhancing the reduction in the generalization of maladaptive safety behaviors.