Proportion congruency effects represent hallmark phenomena in current theorizing about cognitive control. This is based on the notion that proportion congruency determines the relative levels of attention to relevant and irrelevant information in conflict tasks. However, little empirical evidence exists that uniquely supports such an attention modulation account; moreover, a rivaling account was recently proposed that attributes the effect of proportion congruency to mere contingency learning. In the present study, the influences of shifts in list-wide (Experiment 1) or item-specific (Experiment 2) proportion congruency were investigated. As predicted by attention modulation but not by contingency learning, strong asymmetries were observed in such shifting: An increase in the proportion of congruent trials had only limited impact on the size of the congruency effect when participants were initially trained with a mostly incongruent list, but the impact was substantial for an equivalent increase of incongruent trials when participants were initially trained with a mostly congruent list. This asymmetrical list shifting effect directly supports attention modulation by proportion congruency manipulations and as such provides a novel tool for exploring cognitive control. Implications of our findings for existing theories of cognitive control are discussed.