Experimental extinction is the decline in the frequency or intensity of a conditioned behaviour resulting from repetitive performance of the behaviour in the absence of the unconditioned stimulus or reinforcer (Pavlov, 1927). Ample behavioural evidence indicates that experimental extinction does not reflect unlearning of the original trace, but rather a relearning process, in which the new association of the conditioned stimulus with the absence of the original reinforcer comes to control behaviour (Rescorla, 1996). If experimental extinction is indeed learning rather than forgetting, are the neuronal circuits that subserve learning and extinction identical? We address this question by double dissociation analysis of the role of the central (CeA) and the basolateral (BLA) nuclei of the rat's amygdala in the acquisition and extinction, respectively, of conditioned taste aversion (CTA). Whereas local blockade of protein synthesis or beta-adrenergic receptors in the CeA blocks acquisition but not extinction of CTA, a similar intervention in the BLA blocks extinction but not acquisition. Hence, the amygdalar circuit that acquires taste aversion memory differs functionally from the circuit that extinguishes it.