This study was designed to determine whether the acquisition of second-order conditional discriminations becomes more rapid across new discriminations. Three normal grade-school children served as subjects. In general, performances improved across sets of second-order discriminations. Moreover, there was little disruption of performance when the second-order stimuli were changed from discrete forms to being compounded with the sample stimuli. Errors increased markedly when the second-order conditional discrimination shifted from one in which one secondorder conditional stimulus indicated that the original contingencies were reversed to a condition in which one second-order conditional stimulus indicated that the subject should select the same comparison stimulus regardless of which sample form was present. Errors prior to mastery decreased, however, across problems of the new type-thus reproducing the learning-set outcome with new stimuli. Harlow (1949) reported that monkeys' performances on two-choice simple discrimination problems improved markedly across new problems. In his classic study, food-deprived monkeys were trained to raise one of two objects to find a food item placed below one of the objects. After a number of training trials, the monkeys raised the object covering the food. When this occurred, another pair of objects was presented for discrimination. The procedure was repeated with several hundred different pairs of objects. After about 200 problems had been taught, the monkeys consistently selected the correct object after only one trial. In the We are grateful to Dean Williams for help with laboratory tasks, Gladys Williams and Jennifer O