A lot of people show a decline in performance when they have to report a second target stimulus in a stream of distractor stimuli. Curiously, this decline only happens when the second target appears approximately 200-500ms after the first target. Recently, Choi, Chang, Shibata, Sasaki, and Watanabe (2012) have shown that a short, one-hour training can eliminate this "attentional blink". Up to now, it is still unclear why this training works. In this paper, we have evaluated a range of different training paradigms to test several hypotheses about the mechanism behind the reduction of the attentional blink. Our results show that none of these training paradigms have a large training effect when administered in isolation. The training by Choi et al. (2012) outperforms them all. The most likely explanation for this effect are temporal expectations relative to the first target.