Objective: Individuals with panic disorder perceive panic attacks as unpredictable. Because predictability is fundamental to Pavlovian conditioning, failure to predict panic attacks could be due to a basic deficit in conditioning. The present study examined trace eyeblink conditioning in order to test the hypothesis that individuals with panic disorder are impaired in associative learning tasks that depend on declarative memory. Method: Delay and trace eyeblink conditioning were tested in separate experimental sessions in 19 individuals meeting DSM-IV criteria for panic disorder and 19 sex- and age-matched healthy comparison subjects. In the delay paradigm, a mild puff was delivered to the eye at the end of a 500-msec tone; in the trace paradigm, the puff was delivered after a 700-msec empty "trace" interval that followed the end of the tone. Results: Patients and comparison subjects showed similar rates of conditioned responses in the delay paradigm, but patients showed reduced rates of conditioned responses in the trace paradigm. Conclusions: These results suggest that individuals with panic disorder suffer from a deficit in declarative associative learning. Such a deficit points to impaired hippocampal function that may disrupt cognitive processing of internal and external cues predictive of a panic attack.