The list-wide proportion congruency effect describes how the congruency effect varies depending on the frequency of incongruent trials within a block. Specifically, the congruency effect is larger in mostly congruent (MC) blocks compared to mostly incongruent (MI) blocks. Research has shown that adaptation to these blocks does not change symmetrically when transitioning between them: moving from MC to MI leads to a rapid decrease in the congruency effect while transitioning from MI to MC results in little or no increase (Abrahamse et al., 2013). We aimed to investigate this asymmetric list shift (ALS) effect in a within-participants design where all participants experienced both transitions (MC-MI and MI-MC). Throughout Experiments 1–3 using a color-word Stroop task, we identified various forms of practice – including trial-type, stimulus-specific learning, general RT decrease, and a decrease in congruency effects over time (see Schmidt, 2016) – that can either facilitate or obscure the ALS effect, depending on the order in which participants experienced the transitions. In Experiment 4, using a face-name version of the Stroop task, where we minimized practice-related confounding factors by employing more complex stimuli and used a new stimulus set in the second transition, an ALS effect was observed regardless of the order of transition and for frequency-unbiased (i.e., 50% congruent) items. The role of practice effects and implications for the (in)-flexibility of control adaptations will be discussed.