In the first expt. rats learned "a T-maze on the basis of discrimination between 2 irrelevant drive stimuli, one resulting from food deprivation, the other from deprivation of water." Both responses were rewarded by "anxiety reduction." Because some of the subjects learned in this situation, the selective association hypothesis was not supported. "The second exptl. situation involved the learning of the same discrimination on the basis of the same drive stimuli, which were now, however, relevant in the sense that a correct response under thirst stimulation was rewarded with water, while a correct response when hungry was rewarded with food (in addition to the "anxiety reduction" of the first situation). The group which was run under the differential stimulation provided by relevant drives learned much more quickly. The results are interpreted as showing the important role of differential fractional anticipatory goal responses.