A series of experiments used rats to study the effect of a systemic or intra-amygdala infusion of the benzodiazepine, midazolam, on learning and re-learning to inhibit context conditioned fear (freezing) responses. Rats were subjected to two context-conditioning episodes followed by extinction under drug or vehicle, or to two cycles of context conditioning and extinction with the second extinction under drug or vehicle. A 20-min extinction under vehicle resulted in better long-term inhibition on a subsequent drug-free retention test than a 4-min extinction under vehicle, or a 20-min, as well as a 4-min, extinction under drug. However, a 20-min, as well as a 4-min, second extinction under drug was just as effective in producing long-term inhibition as a 20-min second extinction under vehicle and this inhibition was greater than that produced by a 4-min second extinction under vehicle. Initial extinction of 5, 10, or 20 min were equally effective in producing long-term inhibition when the second extinction under drug was 20 min; and 5-, 10-, or 20-min second extinction under drug were equally effective in producing long-term inhibition when the initial extinction was 5 min. A 4- or 20-min second extinction under an infusion of drug into the basolateral amygdala (BLA) was as effective in producing long-term inhibition as a 20-min second extinction under vehicle and was more effective than a 4-min second extinction under vehicle. The results show that midazolam impairs learning to inhibit fear responses but spares and even facilitates re-learning this inhibition.