Savings and recovery after extinction of a classically conditioned response in the rabbit

The experiments described in this thesis examined post-extinctionrecovery phenomena following classical conditioning in the rabbit. The firstseries of experiments examined the rate of reacquisition to the originalconditioned stimulus (CS) and to a cross-modal CS following various amountsof extinction. In both the rabbit nictitating membrane (NM) preparation andrabbit heart rate (HR) conditioning, there was a graded reduction in the rate ofreacquisition as a direct function of the number of extinction trials. In the rabbitNM preparation, there was also a graded reduction in the rate of acquisition toa cross-modal stimulus (CSB, e.g., light). However, concurrent recovery ofresponding to the original, extinguished stimulus (CSA, e.g., tone) duringtraining with a novel, cross-modal stimulus (CSB, e.g., light), appeareduniformly robust even following extensive extinction. The second series ofexperiments examined the conditions necessary for concurrent recovery tooccur in the rabbit NM preparation. The introduction of CSB-US pairingsresulted in strong recovery of responding to CSA, while the reintroduction ofthe US alone failed to result in any discernable reinstatement of responding toCSA. Concurrent recovery was specific to the trained and extinguished CSA,with virtually no generalized responding to a novel cross-modal test stimulus(CSC). However, pretraining with CSB-US pairings significantly reduced theamount of recovery to CSA during subsequent CSB–US pairings. The thirdseries of experiments revealed that concurrent recovery was specific to theextinguished stimulus (CSA) with only moderate generalization to otherfamiliar but untrained stimuli in the same modality as CSA. Together theseresults indicate that concurrent recovery is not primarily the result of theunmasking of the original CSA-US association. Rather, concurrent recoveryappears to be at least partially the result of learning-dependent generalization.However, in order for responding to CSB, to generalize to CSA, CSA must havebeen paired with the same US which is subsequently paired with CSB. Thus,the CSA-US association must be partially intact for concurrent recovery tooccur. A layered network model with multiple hidden units is able to simulaterapid reacquisition, facilitated cross-modal acquisition and concurrent recovery