Classic studies in the attentional blink (AB) literature indicate that despite the severe impairment in identification of the second of two briefly presented targets (T1 and T2), semantic information about T2 survives the AB. However, recent work has demonstrated that semantic processing is attenuated when i) the T1 task is perceptually demanding (Giesbrecht, et al., 2007) and ii) when there is a task-switch between T1 and T2 (Vachon, et al, 2007). In the present experiments, we examined whether a switch in attentional-set between trials affects semantic processing during the AB. In two experiments, participants identified two masked targets presented in rapid succession. In both experiments, the T1 task was to identify the direction of a central arrow while perceptual load was manipulated by the direction of irrelevant flankers (i.e., low load, [[gt]][[gt]] [[gt]] [[gt]][[gt]]; high load, [[gt]][[gt]] [[lt]][[gt]][[gt]]). In Experiment 1, the T2 task was to discriminate the gender of a name that was either the participant's name or a different name. In Experiment 2, the context-sensitive N400 event-related potential evoked by T2 was measured while participants indicated the relationship of a T2 word to a context word presented at the start of the trial. Trials were analyzed as a function of whether T1-load on consecutive trials was the same (repeat trials) or different (switch trials). Experiment 1 indicated that there was a significant AB for both name conditions on switch trials, but only an AB for someone else's name on repeat trials. Experiment 2 found that the magnitude of the N400 during the AB was suppressed on switch trials relative to repeat trials. The finding that semantic processing during the AB is influenced by switching attentional-set between trials is consistent with the notion that the locus of selection is flexible and determined by task demands.