The purpose of the study was to investigate the effectiveness of the superordinate and subordinate words as mediators in mediate association learning across three different age groups in the A-B, B-C, A-C mediation x paradigm. The effects of sex difference, high and low levels of associational fluency, as well as a type of pre-training which was done by requiring subjects to create associations to the mediating B terms, were considered. The subjects used were, 50 fourth graders (22 boys and 28 girls), 54 seventh graders (26 boys and 28 girls), and 51 tenth graders (22 boys and 29 girls), for a total of 155 students. Each group of subjects was divided into high and low associational fluency groups using the cutting point. The high and low associational fluency groups were divided further into two groups each using the odd-even method on the ranks on the associational fluency scores. Prior to the mediate association learning task, one group received the pre-training procedure while the other received the no-pre-training procedure where each subject was simply required to write a short story. The learning materials consisted of superordinate and subordinate words, and high association value nonsense syllables. Two sets of A-B, B-C, A-C, mediate association learning task were constructed using superordinate words as B terms in one set and subordinate words as B terms in the other, while the same nonsense syllables were used as the A and C terms in both sets. Each set of the learning task consisted of two 12 paired-associate lists, half of which was used as the experimental pairs with the other half as the control pairs. Each list of paired-associates was presented one pair at a time for five seconds for seven trials. Each subject received both sets of the learning task and served as his own control. The multiple-choice method was used to measure the amount of mediation. The results of the experiments were as follows: Mediation was obtained with both the superordinate and subordinate words as mediators when subjects in the fourth, seventh and tenth grades were each treated as a single group, with the exception of the fourth grade group when the superordinate words were used as mediators. In general, when subjects were treated on the subgroup basis, mediation was obtained increasingly with age. The pre-training method was not superior to the no-pre-training (control) method in producing mediation. In fact, on the seventh grade level with superordinate words as mediators, the pre-training group was inferior to the control group in producing mediation. Superordinate words were not more effective mediators than subordinate words. There was a trend which indicated that subordinate words were more effective mediators than the superordinate words with the increasing age of subjects, especially among female subjects. There was no clear-cut evidence that the high associational fluency level enhanced the occurrence of mediation more than the low associational fluency level when the superordinate words and subordinate words were used as mediators. It was found that age had a definite effect on mediation. The amount of mediation produced occurred increasing with age, at least for the three grade levels compared. There was no sex difference on mediation produced superordinate words as mediators. But with the subordinate words as mediators there were sex difference effects on the amount of mediation produced in favor of female groups on the fourth and tenth grade levels. In conclusion, the results of the study were in agreement with the majority of the findings reported in the literature on mediate association experiments that mediation could be experimentally produced. The superordinate words and subordinate words proved to be about equally effective as mediators and their effectiveness increased with the increasing age of the subjects.