Summary Eight mediation paradigms were evaluated by comparing test-trial responses following mediation presentation with test-trial responses following control presentation. CCCs of low association value were used as the learning materials in one experiment, and CVCs of 100% association value in a second experiment. Response-equivalence paradigms and one stimulus-equivalence paradigm yielded mediated facilitation with both types of learning materials. When CVCs were used, two chaining models also showed mediated facilitation. Forward associations and minimal temporal separation of the A and C items, as they probably existed implicitly and explicitly during the second acquisition stage, led to mediated facilitation only when CCCs were the learning materials. With more meaningful materials these two variables were considerably less potent. Two additional experiments were performed to replicate the results of Paradigm II, which yielded negative results in Exp. I and positive results in Exp. II. The data corroborated the earlier conclusions that mediated facilitation was more likely to be observed with learning materials of higher meaningfulness. Selections of C as the test-trial alternative which made the “best pair” with the stimulus were reliably above an a priori chance level for all paradigms. In addition, the eight paradigms showed significant differences in this factor (response availability) which seemed to be predictable from the temporal interval separating presentation of C in the acquisition stages and its appearance of a test-trial alternative.