Amnesia for approach-avoidance learning was induced in mice by injecting the protein synthesis inhibitor anisomycin (ANI) immediately, 1, or 2 hours, but not 3 hours after training. A robust amnesia could be demonstrated if ANI was administered 3 hours after training, immediately following a 60 second exposure to the training apparatus or to a structurally similar environment. The temporal gradient of effectiveness of amnesia production by ANI was significantly steeper following reactivation treatment than it was following initial training. In addition, while amnesia produced by the conventional procedure remained stable for 6 days, the amnesia induced following reactivation treatment spontaneously recovered 4 days after training. These findings are discussed in terms of their relevance to interpretations of retrograde amnesia studies.