Objective: This study investigated cognitive mediation of improvement in panic disorder (PD) symptomatology during and after an early intervention for panic symptoms in subthreshold and mild PD. Method: We executed a pragmatic, pre-post, two-group, multi-site, randomized trial of an early intervention for panic symptoms, based on cognitive-behavioural therapy, vs. a wait-list control group in a sample of 217 participants with subthreshold PD or mild PD. Results: First, two of the three subscales of the mediator variable Panic Appraisal Inventory (PAI-anticipation and PAI-coping) significantly mediated residual change in PD symptomatology on the PD Severity Scale-Self Report. Second, preintervention to postintervention PAI-anticipation and PAI-coping change scores significantly predicted postintervention to follow-up change in PD symptomatology after controlling for other change scores. However, the converse association was also significant. Conclusion: The results suggest that changes in cognitions may mediate changes in PD symptomatology and that the process of change is circular. © 2010 John Wiley & Sons A/S.