Memory for A-B word pairs (e.g., child-apple) was tested by a cued recall test (e.g., child-app__). Showing an A-C "relative" (e.g., child-bicycle) reduced recall, especially if it was shown recently and was highly accessible (Experiments 1 and 2). In Experiment 3, a relative facilitated recall if it was semantically similar to the target (A-B': child-cookies) but interfered if it was semantically dissimilar (A-C: child-fever). The best explanation for these results is that the relative primed features that affected the functional retrieval cue, and that interference occurred if the cue did not match the trace for the target (Martin, 1972). In other words, the interference effects are an example of cue-dependent forgetting. Neither blocking nor a discrimination process can account for these findings, although some evidence for a discrimination process has been found with other materials.