Three experiments investigated the role of target-target perceptual similarity within the attentional blink (AB). Various geometric shapes were presented in a rapid serial visual presentation task. Targets could have 2, 1, or 0 shared features. Features included shape and size. The second target was presented after five or six different lags after the first target. The task was to detect both targets on each trial. Second-target report accuracy was increased by target-target similarity. This modulation was observed more for mixed-trial design as compared with blocked design. Results are discussed in terms of increased stability of working memory representations and reduced interference for second-target processing.