The paired-associate learning of 52 fourth graders was related to measures of cognitive ability obtained the previous year. Subjects were administered the paired-associate task under one of three variations: at a comfortable rate with standard instructions; at a comfortable rate with a potentially effective learning strategy (visual imagery); and at a speeded rate with standard instructions. As anticipated on the basis of earlier research with children of this age, the relationship between reasoning (as reflected by Raven's Progressive Matrices) and learning was augmented when subjects were supplied with the imagery strategy and diminished when the task was speeded. Negligible correlations between rote memory (as reflected by digit span) and learning were obtained under all task variations. Interpretations of the results and speculations for future research are included. (Author)