Effects of hippocampal manipulations on the classically conditioned nictitating membrane response: simulations by an attentional-associative model.

The effects of various hippocampal manipulations on the classically conditioned nictitating membrane (NM) response were simulated by a real-time attentional-associative model. The model incorporates: (a) a mechanism capable of establishing associations between conditioned (CS) and unconditioned stimuli (US) and between two CSs, (b) a mechanism that combines CS-CS and CS-US associations and builds 'computational cognitive maps', (c) an attentional rule that 'tunes in' relevant CSs and 'tunes out' irrelevant CSs, (d) performance rules that convert learning variables into topography of the rabbit NM response, and (e) rules that convert learning variables into neuronal firing. The present study explored three hypotheses regarding hippocampal function: (a) hippocampal lesions (HL) impair the 'tuning out' of irrelevant stimuli, (b) hippocampal stimulation (HS) and hippocampal long-term potentiation (LTP) facilitate the 'tuning in' of relevant stimuli, and (c) hippocampal unit activity is proportional to the instantaneous value of associative variables that participate in the 'tuning' mechanisms. Computer simulations for the HL case were carried out for conditioning under different interstimulus intervals (ISI) with different types of US, discrimination reversal, and sensory preconditioning. Computer simulations for the HS and LTP case were carried out for acquisition of delay conditioning and of classical discrimination. In addition, simulations of hippocampal unit activity during acquisition and extinction are presented. Under the 'tuning out' hypothesis for the HL case, the model proved capable of simulating a large portion of the experimental data, showing discrepancies with relevant literature only when describing HL effects on trace conditioning with shock US under short and long ISIs, trace conditioning with air puff US under long ISIs, discrimination reversal and sensory preconditioning. Under the 'tuning in' hypothesis regarding the HS and LTP case, the model proved capable of simulating acquisition of delay conditioning but not acquisition of classical discrimination. Under the assumption that neuronal activity is proportional to the instantaneous value of associative variables, the model is able to simulate hippocampal unit activity.