The present investigation examined patterns of change in basic features of panic and anxiety during cognitive behavioral treatment for panic disorder. Data were collected in the context of a study comparing the efficacy of cognitive therapy (CT) and relaxation training, both administered without exposure-based treatments of any kind. Thirty-six panic disorder patients completed 10 weeks of either CT or relaxation training. Weekly measures of panic frequency, state and trait anxiety, and associated fears were obtained. Data were examined using both multivariate techniques (which conceptualize change as incremental) and analysis of response slopes (wherein change is conceptualized as continuous). Results indicated that during the first half of treatment, relaxation training led to greater reductions in state and trait anxiety and agoraphobic fear, whereas CT demonstrated a slight advantage on change in panic frequency. During the second half of treatment, CT produced faster reductions in state anxiety and agoraphobic fears. Over the entire treatment course, CT conferred an advantage in rate of change in social fears. The data provide support for the utility of a more process-oriented approach to the examination of treatment effects.