Inhibition in task switching is inferred from n–2 repetition costs: the observation that ABA task switching sequences are responded to slower than CBA sequences. This is thought to reflect the persisting inhibition of Task A, which slows reactivation attempts. Mayr (2002) reported an experiment testing a critical noninhibitory account of this effect, namely episodic retrieval: If the trial parameters for Task A match across an ABA sequence, responses should be facilitated because of priming from episodic retrieval; a cost would occur if trial parameters mismatch. In a rule-switching paradigm, Mayr reported no significant difference in n–2 repetition cost when the trial parameters repeated or switched across an ABA sequence, in clear contrast to the episodic retrieval account. What remains unclear is whether successful episodic retrieval modulates the n–2 repetition cost. Across 3 experiments—including a close replication of Mayr—we find clear evidence of reduced n–2 task repetition costs when episodic retrieval is controlled. We find that the effect of episodic retrieval on the n–2 task repetition cost is increased when the cue–task relationship is made more abstract, suggesting the effect is because of interference in establishing the relevant attentional set. We also demonstrate that the episodic retrieval effect is not influenced by retrieval of low-level, perceptual, elements. Together, the data suggest the n–2 task repetition cost—typically attributable to an inhibitory mechanism—also reflects episodic retrieval effects.