The recognition of the acquisition pairs was tested before and after the test trials of two chaining paradigms, a response-equivalence paradigm, a stimulus-equivalence paradigm, and a general facilitation paradigm, X-Y, B-Z, A-C. The second-stage pairs were recognized on almost all tests. The first-stage pairs were recognized less often and the frequency of recall was highest for the equivalence paradigms, less for the chaining paradigms, and lowest for the general facilitation paradigm. The test-pair terms were recognized most often after the test trials and the B terms (mediators) were recognized more often for the equivalence paradigms than for the other paradigms. A second experiment showed all the acquisition-stage pairs as a single list, the test trials as a separate list, and tested the recall of the acquisition pairs before and after the test-trial list. The pairs with a common term tended to be recalled together by the mediation group both before and after the test-trial list.Mediational theory assumes that mediation takes place during either the second or third stage of three-stage paradigms such as B-C, B-A, A-C. The second-stage hypothesis (e.g., Horton & Kjeldergaard, 1961;Jenkins, 1963) assumes that presentation of the B trigram of the second-stage pair tends to elicit the term with which it was associated during the first stage, C, so that an association may occur between A and C during the second stage which will transfer to the third stage. The third-stage hypothesis posits that presentation of A as the test-stage stimulus tends to elicit the term with which it was previously associated, B, which in turn tends to elicit another former associate, C.Others have claimed that mediation is an artifact due to differential unlearning and/ or interference. For example, Handler and Earhard (1964) suggested that the first-stage pairs of mediation sets were unlearned as the second-stage pairs were acquired. Then the first-stage pairs would not be available to in-