When bivalent stimuli (i.e., stimuli with features for two different tasks) appear occasionally, performance is slower on subsequent univalent stimuli. This "bivalency effect" reflects an adjustment of cognitive control arising from the more demanding context created by bivalent stimuli. So far, it has been investigated only on task switch trials, but not on task repetition trials. Here, we used a paradigm with predictable switches and repetitions on three tasks, with bivalent stimuli occasionally occurring on one task. In three experiments, we found a substantial bivalency effect for all trials with at least one source of conflict. However, this effect was reduced for the repetition trials sharing no features with bivalent stimuli, that is, for those without conflict. This confirms that the bivalency effect reflects an adjustment of cognitive control. The news is that this adjustment of cognitive control is sensitive to the presence of conflict, but neither to its amount nor to its source. (Contains 2 figures.)