In two studies, a vocal Stroop task with eight different colours was employed in order to put two core assumptions of the original interpretation of the Gratton effect to the test. We verified whether top-down control processes can elicit conflict adaptation when episodic memory effects are controlled for and to what extent proactive adjustments driven by the subjects' expectancy for congruency level repetition contribute to this effect. Therefore, we presented Stroop stimuli without feature repetitions and investigated whether the induced expectancy manipulation of raising the amount of either congruency level repetitions or alternations in a training phase transferred to an unmanipulated test phase. Over the two experiments, a sequential modulation of the Stroop effect was found in the absence of stimulus feature repetitions, strongly confirming a share for top-down control processes in bringing about the Gratton effect. In the condition where congruency level repetitions were raised, a strong Gratton effect was found. When congruency level alternations outnumbered repetitions, the Gratton effect disappeared completely. However, this difference seemed mainly due to cumulative effects of local, dynamic, trial-to-trial control adjustments rather than expectancy-induced attentional shifting. Once the transition probability changed back to 50% in the test phase of each experiment, a similar Gratton effect was found in both conditions. Taken together, these results are best explained in terms of dynamic reactive control.