When the second of two targets is presented temporally close (within 500 ms) to the first target in rapid serial visual presentation, accuracy for reporting the second target is markedly diminished—an attentional blink (AB). The AB has become a well-studied phenomenon, and multiple different versions of the AB are currently in use. However, little is known about the stability of individual performance on the AB. The current study examined the reliability of two different versions of the AB task (a task-switch and no-task-switch version) within session, and over the period of 7–10 days, in order to examine performance stability. In addition to testing the reliability, we also examined the relationship between both versions of our AB tasks. Both versions of the AB were shown to be reliable within session, and over time, suggesting that performance is quite stable on this task. In addition, performance on the two different AB tasks was significantly correlated within and across sessions, suggesting that the AB phenomenon is being accurately captured by versions of the AB that include a task-switch. These findings are important given the recent interest in individual differences in performance on the AB.