The Rescorla-Wagner model treats context like any other conditional stimulus. Since context is always present, whenever a discrete conditional stimulus (CS) is presented on a conditioning trial, the model treats that trial as a presentation of the discrete CS plus context. More specifically, changes in the associative strengths of stimuli presented on a conditioning trial are a function of the discrepancy between the maximum associative strength supportable by the unconditioned stimulus (UCS) and the current sum of the associative strengths of all stimuli present on the trial. The Scalar Expectancy Theory (SET) advocated by Gibbon and Balsam is the most fully developed model incorporating the assumption of independence of learning, and will be the example used in our discussions of this assumption. A study by Randich and Ross (1985) tested UCS pre-exposure effect using the Conditioned Emotional Response (CER) procedure with rats. The Rescorla-Wagner model also predicts the withdrawal from CS observed in both groups.