In three experiments, subjects who had rated 35 stimulus words on a seven-point affective scale (like/dislike or pleasant/unpleasant) showed better recall and recognition after incidental learning that subjects who had not so rated words did after intentional learning. Clustering by rating categories was not consistently related to, while the number of rating categories was significantly correlated with, level of recall. It is argued that the critical variable is not the intention to learn but the effectiveness of the operations to learn, and the facilitation with affective ratings is interpreted in terms of Bartlett's 'schemata.'