Identification of a target is impaired when it follows a previous target within 500 ms, suggesting that our attentional system suffers from severe temporal limitations. Although control-disruption theories posit that such impairment, known as the attentional blink (AB), reflects a difficulty in matching incoming information with the current attentional set, disrupted-engagement theories propose that it reflects a delay in later processes leading to transient enhancement of potential targets. Here, we used a variant of the contingent-capture rapid serial visual presentation (RSVP) paradigm (Folk, Ester, & Troemel, 2009) to adjudicate these competing accounts. Our results show that a salient distractor that shares the target color captures attention to the same extent whether it appears within or outside the blink, thereby invalidating the notion that control over the attentional set is compromised during the blink. In addition, our results show that during the blink, not the attention-capturing object itself but the item immediately following it, is selected, indicating that the AB manifests as a delay between attentional capture and attentional engagement. We therefore conclude that attentional capture and attentional engagement can be dissociated as separate stages of attentional selection.