Associative learning models typically reflect statistical relationships between experienced events. Causal models can go beyond this information to specify the ways in which events are related. This meta-representational aspect of causal models allows them to reflect uncertainty about relationships between events: for example, if a light initially leads to sucrose but subsequently the light is experienced without sucrose, this might first support formation of a light-causes-sucrose model and subsequently lead to uncertainty over whether the model remained accurate. Prior studies of Pavlovian conditioning in rats manipulated sucrose-magazine access during extinction to produce uncertainty about reward presence or absence. Rats were sensitive to covering of the site of reward delivery, which was interpreted as evidence for a causal-model account reflecting uncertainty. However, associative accounts-based on the direct impact of the dipper mechanism used to deliver sucrose through secondary reinforcement or contextual renewal of responding-can also explain the results. In two new experiments, manipulation of the dipper mechanism through extinction and test phases resulted in behavior consistent with these associative accounts. However, demonstration of the importance of the sucrose dipper suggests that the reward delivery mechanism should be included in a causal model. Such a revised causal model also provides an account of the impact of manipulating the sucrose dipper. While these experiments do not conclusively decide between associative and causal models as explanations of rodent behavior, they do illustrate the value of incremental experimental study and the importance of methodological detail in addressing questions of comparative cognition. (PsycInfo Database Record (c) 2024 APA, all rights reserved).
Read full abstract