Over 20 years of research has shown that impairments in the identification of sequential targets—known as the attentional blink—are often eliminated when 2 or more targets are presented in direct succession. Such instances of lag-1 sparing and, more recently, extended sparing are widely attributed to the fact that directly successive visual inputs that possess common characteristics can be processed during an extended attentional window. Critically, this explanation implies that direct temporal succession and shared characteristics are sufficient to result in sparing. Here, I test this proposition directly by varying the probability of successive targets appearing during an experimental session. Across 5 experiments, I show that probability manipulations significantly affect the magnitude of sparing, with less sparing when the probability of a target appearing is relatively low. This outcome cannot be explained by resource shifts between targets, errors in perception of target order, or practice effects. Instead, the results suggest that sparing is determined not only by temporal contiguity and shared target characteristics but also by endogenous control processes that directly affect the duration of the attentional window.