Three experiments tested the generality of the conclusion that associative unlearning is minimal in the A-B, A-D paradigm. In Experiment 1, single-trial study of A-D, following single-trial study of A-B, did not produce retroactive inhibition in the recognition of A-B. In Experiment 2, A-B was acquired by associative matching. The interpolated learning was by the paired-associate anticipation method and included A-D and C-B pairs. There was no evidence for losses in A-B on a recall test. In Experiment 3, A-B was learned by a paced, multiple-choice procedure, and A-D and A-Br pairs in an interpolated list were learned under the same procedure. There was a small amount of associative unlearning following A-D, and heavy associative unlearning following A-Br. The possibility was offered that little associative unlearning is detected by recognition measures in A-B, A-D because recognition decisions are not commonly based on associative information. (Author)