Autonomic response indexes of experimental amnesia have recently been found to have higher electroconvulsive shock (ECS) intensity thresholds and steeper retrograde gradients than have traditional somatic indexes. The present studies examined the hypothesis that recovery from somatically indexed experimental amnesia depends upon the existence of autonomically available residual memory. In a between-subjects design, a 200-mA ECS was used to produce amnesia for a tone-footshock pairing as indicated by lick suppression, defection, and bradycardia. The next day, these amnesic animals received a reminder footshock outside of the training apparatus, which was found to restore memory on a test trial 24 hr later. The behavior of control groups indicated that this reminder effect was due to the restoration of specific memory rather than systemic consequences of treatment. With a within-subjects design, a second experiment obtained a reminder effect in animals individually shown to be "fully" amnesic by all three response indexes monitored. A third experiment varied the intensity of the reminder footshock and revealed that the different memory indexes examined do not have reminder-footshock thresholds inversely related to their initial resistance to amnesia. The results support a retrieval-failure view of experimental amnesia and suggest that the same fundamental physiological processes underlie both autonomically indexed memory and somatically indexed memory.