This article reviews experimental evidence and theoretical constructs that implicate serotonin (5-HT) modulation of defensive behavior within the midbrain periaqueductal gray in panic disorder (PD). Evidence with conflict tests in experimental animals indicates that 5-HT enhances anxiety, whereas results with aversive stimulation of the dorsal periaqueductal gray point to an anxiolytic role of 5-HT. To solve this contradiction, it has been suggested that the emotional states determined by the two types of animal model are different. Conflict tests would generate conditioned anxiety, whereas periaqueductal gray stimulation would produce unconditioned fear, as evoked by proximal threat. Clinically, the former would be related to generalized anxiety while the latter to PD. Thus, 5-HT is supposed to facilitate anxiety, but to inhibit panic. This hypothesis has been tested in the animal model of anxiety and panic named the elevated T-maze, in two procedures of human experimental anxiety applied to healthy volunteers or panic patients, and in CO2-induced panic attacks. Overall, the obtained results have shown that drugs that enhance 5-HT function increase different indexes of anxiety, but decrease indexes of panic. Drugs that impair 5-HT function have the opposite effects. Thus, so far the predictions derived from the above hypothesis have been fulfilled.