Analysis and modeling of neural processes underlying sensory preconditioning.

Sensory preconditioning (SPC) is a procedure to demonstrate learning to associate between relatively neutral sensory stimuli in the absence of an external reinforcing stimulus, the underlying neural mechanisms of which have remained obscure. We address basic questions about neural processes underlying SPC, including whether neurons that mediate reward or punishment signals in reinforcement learning participate in association between neutral sensory stimuli. In crickets, we have suggested that octopaminergic (OA-ergic) or dopaminergic (DA-ergic) neurons participate in memory acquisition and retrieval in appetitive or aversive conditioning, respectively. Crickets that had been trained to associate an odor (CS2) with a visual pattern (CS1) (phase 1) and then to associate CS1 with water reward or quinine punishment (phase 2) exhibited a significantly increased or decreased preference for CS2 that had never been paired with the US, demonstrating successful SPC. Injection of an OA or DA receptor antagonist at different phases of the SPC training and testing showed that OA-ergic or DA-ergic neurons do not participate in learning of CS2-CS1 association in phase 1, but that OA-ergic neurons participate in learning in phase 2 and memory retrieval after appetitive SPC training. We also obtained evidence suggesting that association between CS2 and US, which should underlie conditioned response of crickets to CS2, is formed in phase 2, contrary to the standard theory of SPC assuming that it occurs in the final test. We propose models of SPC to account for these findings, by extending our model of classical conditioning.