A test of Rescorla and Wagner's (1972) prediction of nonlinear effects in contingency learning.

According to most theories, in a simple contingency learning situation, excitatory learning occurs when the probability of the unconditioned stimulus in the presence of the conditioned stimulus (p(1)) is higher than the probability of the unconditioned stimulus in the absence of the conditioned stimulus (p(2)). In Rescorla and Wagner's (1972) model, this prediction varies, depending on the parameters used. In the following experiments, we evaluated whether the difference between p(1) and p(2) that is required to produce excitatory conditioning is the same, independent of the specific value of p(1), or whether this difference varies proportionally to p(1)'s value. To do so, an appetitive procedure of Pavlovian conditioning with rats was used. In four experiments, we compared different levels of contingency (low, medium and high) and found that the difference between p(1) and p(2) that is required to produce excitatory conditioning increases when the value of p(1) is higher. The possibility of analyzing contingency learning as a discrimination between p(1) and p(2) is also discussed.