The congruency sequence effect, the finding of a reduced congruency effect following incongruent trials in conflict tasks, has received considerable attention in the research on cognitive control over the last two decades. This effect can reflect either the expectancy-guided, preparatory biasing of attention in anticipation of the upcoming stimulus (i.e. proactive control), or the phasic enhancement of the attentional set in response to conflict on the previous trial (i.e. reactive control). A recent study by Egner et al. in Front Psychol 1 (2010) set out to contrast these two alternatives, by exploring the congruency sequence effect across a wide range of inter-trial intervals. It was found that congruency sequence effects were subject to rapid decay over time. This decay fits well with the notion of reactive control, while at the same time speaking against the involvement of proactive regulation—which should also (and even mainly) be evident at longer intervals. In the present study, we first replicate the reduction of the congruency sequence effect with increasing inter-trial interval in a face-word Stroop task. In a second experiment, we show that congruency sequence effects are observed at longer intervals, too, when the proportion of trials with the longest inter-trial interval is increased. Our findings indicate that proactive control can prevent the congruency sequence effect from decaying rapidly.