Fear extinction is defined as a decline in conditioned fear responses (CRs) following nonreinforced exposure to a feared conditioned stimulus (CS). Behavioral evidence indicates that extinction is a form of inhibitory learning: Extinguished fear responses reappear with the passage of time (spontaneous recovery), a shift of context (renewal), and unsignaled presentations of the unconditioned stimulus (reinstatement). However, there also is evidence to suggest that extinction is an "unlearning" process corresponding to depotentiation of potentiated synapses within the amygdala. Because depotentiation is induced more readily at short intervals following LTP induction and is not inducible at all at a sufficient delay, it may be that extinction initiated shortly following fear acquisition preferentially engages depotentiation/"unlearning," whereas extinction initiated at longer delays recruits a different mechanism. We investigated this possibility through a series of behavioral experiments examining the recoverability of conditioned fear following extinction. Consistent with an inhibitory learning mechanism of extinction, rats extinguished 24-72 h following acquisition exhibited moderate to strong reinstatement, renewal, and spontaneous recovery. In contrast, and consistent with an erasure mechanism, rats extinguished 10 min to 1 h after acquisition exhibited little or no reinstatement, renewal, or spontaneous recovery. These data support a model in which different neural mechanisms are recruited depending on the temporal delay of fear extinction. (Contains 4 figures and 5 footnotes.)