Compared the effectiveness of cognitive therapy and performance-based exposure as treatments for panic attacks. Subjects were 48 panicky individuals selected without regard to agoraphobic disability, and who varied widely in that respect. Subjects were assigned randomly to either cognitive treatment, performance-based exposure treatment, a combined cognitive/performance treatment, or a no-treatment control condition. All three treatments led to marked and enduring improvements in panic, and did not differ from one another in effectiveness, whereas the control condition produced little benefit. However, on several measures of phobia and panic-related cognitions, performance exposure was significantly more effective than cognitive therapy. Degree of agoraphobic disability had a significant bearing on panic treatment effectiveness. Whereas 94% of the low agoraphobia Ss were free of panic after treatment, only 52% of the high agoraphobia Ss became panic-free. The findings suggest that when panic treatment research excludes people with serious phobias, as it has routinely done in recent years, an overly positive estimate of panic treatment effectiveness can result.