In three experiments, we assessed the effects of type of relation and memory test on retrieval-induced forgetting of facts. In Experiments 1 and 2, eight sets of four shared-subject sentences were presented for study. They were constructed so that half were thematically related and half were unrelated. A retrieval practice phase required participants to recall a subset of the studied sentences. In the final test, the participants were prompted to recall all the sentences (character cued in Experiment 1 and character plus stem cued in Experiment 2). The results showed that the retrieval-induced forgetting (RIF) effect was similar for thematically related and unrelated sentences, indicating that the presence of episodic relations among the sentences was sufficient to produce the effect. In Experiment 3, a recognition task was introduced and the RIF effect emerged in accuracy as well as in latency measures. The presence of this effect with item-specific cues is difficult to accommodate for noninhibitory theories of retrieval.