Learning to Learn: Real-Time Features and a Connectionist Model

Abstract

Two experiments used classical conditioning to examine transfer of response features specific to the warning interval between a conditioned stimulus (CS) and an unconditioned stimulus (US). Rabbits were given initial training with a stimulus (CSA) in one modality (tone) at a designated interval (e.g., 200 ms). In a second stage, the conditioned response (CR) to CSA was extinguished. Finally, training was shifted to a new stimulus (CSB) in another modality (light) at a new interval (e.g., 400 ms). Compared to rest controls, there was an enhanced rate of CR acquisition to CSB and a tendency for early CRs to CSB to reflect the CSA-US interval. However, the extinction procedure and the change in CS-US interval together reduced these two aspects of transfer across stimulus modalities. The findings are discussed in terms of their implications for theories of cross-modal transfer. A layered network with real-time features is offered.

Savings in classical conditioning in the rabbit as a function of extended extinction

“…The modest evidence for facilitation of cross-modal acquisition in the present experiment stands in stark contrast to repeated previous demonstrations in the rabbit NM preparation that transfer across modalities survives extinction of responding to the original CS (Kehoe et al, 1995;Kehoe et al, 1984;Macrae & Kehoe, 1999). In the previous studies, however, many more initial acquisition trials were used than in the present experiment.…”

“…Second, in contrast to the modest cross-modal savings, concurrent recovery of CRs to the original CS was robust even in the face of the most extended extinction. As was found in previous demonstrations of concurrent recovery, the level of responding to the original CS far exceeded any cross-modal stimulus generalization seen in the rest control group (Kehoe et al, 1995;Macrae & Kehoe, 1999).…”

“…The second form of savings occurs in cross-modal acquisition (CMA): When a new CS from a different modality is introduced after extinction, CR acquisition to the new CS is more rapid than it is in a naive control group (Kehoe, Morrow, & Holt, 1984;Macrae & Kehoe, 1999). The third form of savings is concurrent recovery, which occurs during CMA: When the original CS is tested, CRs to the original CS recover up to 75% of their original likelihood, despite the fact that the original CS has not been reinforced since initial acquisition training (Kehoe, Horne, & Macrae, 1995;Macrae & Kehoe, 1999).…”

Temporal discrimination using different feature-target intervals in classical conditioning of the rabbit’s nictitating membrane response

“…To obtain the temporal features of conditional discriminations, neural network models have borrowed from real-time models of conditioning (Desmond, 1990;Desmond & Moore, 1988;Grossberg & Schmajuk, 1989;Kehoe, Horne, & Macrae, 1995;Sutton & Barto, 1990). According to many real-time models, a CS is encoded as a cascade of hypothetical stimulus elements.…”

Stimulus specificity of concurrent recovery in the rabbit nictitating membrane response

“…During CSB-US training, CSA is tested occasionally without reinforcement. As responding is acquired to CSB, responding to CSA recovers up to 75% of its original strength, despite its not having been reinforced since initial acquisition training (Kehoe, Horne, & Macrae, 1995;Macrae & Kehoe, 1999;.…”