Learning to learn and generalization in human conditioning

Abstract

Each situation we encounter is at least in part unique. Thus, the flexible use of the information acquired in previous experiences becomes a fundamental skill to maximize our chances of success. While stimulus generalization is a form of transfer that relies on the sensory similarities of the stimuli, learning to learn (LTL) denotes a facilitation of learning in the form of an increased learning rate across tasks that share a common structure. The LTL effect has been clearly shown in cognitively oriented tasks, however, it is important to understand the effect at the level of basic processes such as associative learning, since they begin to indicate boundaries for the effect. For instance, it is not clear if extinction learning transfers across different stimuli. Several experiments assessed whether LTL can be obtained in classical conditioning procedures with humans while evaluating the extent to which generalization might contribute to transfer of learning. We used a science-fiction based videogame (Nelson, Navarro, & Sanjuan, 2014) where different conditioned stimuli (flashing lights or a tone) are associated with the appearance of an attacking spaceship (unconditioned stimulus). The procedure requires the participants to give an anticipatory response (charging a weapon) which allowed us to trace the course of learning trial by-trial along several repeated acquisition and extinction tasks. Our results indicate that the LTL effect is apparent even in situations where no stimulus generalization is present. Furthermore, the effect is present both in acquisition and extinction procedures.