This study examined the relationships among certain subtypes of panic attacks (full vs. limited symptom; spontaneous vs. situational) and between these subtypes, panic disorder subtypes, and other characteristics of panic disorder, especially agoraphobia.Data were drawn from a large (n = 1,168) treatment study of panic disorder in which panic attacks were carefully subtyped and counted using a diary. Relationships between variables at baseline were examined primarily using non-parametric methods, and the course of improvement for panic subtypes among completers was plotted.The median number of spontaneous panic attacks per week at baseline was similar among patients with panic disorder without agoraphobia (PD), limited phobic avoidance (PDL), and agoraphobia (PDA). The median number of situational attacks and the median agoraphobia ratings rose progressively across diagnostic subtypes. Anticipatory anxiety, HAM-A, HAM-D, and disability scores were higher in PDA than in PD. Full and limited symptom panic attacks were positively correlated. The proportion of total attacks that were limited rose during the first two weeks of treatment, suggesting conversion of full to limited symptom attacks before complete disappearance. Spontaneous and situational attacks were correlated minimally or not at all. Agoraphobia ratings were more positively correlated with situational than with spontaneous panic attacks. Few of the correlations among measures at baseline were high.Full and limited symptom panic attacks differ primarily in severity. Spontaneous and situational attacks are relatively independent, and situational attacks are more closely related to agoraphobia. These findings are consistent with previous work suggesting that spontaneous attacks reflect a biological component, whereas situational attacks reflect a cognitive component in the psychopathology-- and possibly the pathogenesis-- of panic disorder. This provides a rationale for the use of combined pharmacotherapy and psychotherapy in the treatment of panic disorder. Future investigations of panic disorder should carefully separate panic attack subtypes.