Abstract
Wagner’s REM theory of associative learning predicts that two equally salient stimuli reinforced as a compound CS will each acquire no more than ½ of the available associative strength, V. Using associative ratings as a proxy for V, generalization from an AB compound to A and B yielded ratings that were greater than ½. A proposed solution to this problem is to subtract the contribution of common elements before applying REM to predict generalized associative strength. Implications of including common cues affect other predictions as well.