In three experiments with the Eriksen flanker task (Eriksen & Eriksen, 1974), we investigated whether a regular sequence of either the targets, the congruency levels, or both could be implicitly (or explicitly) learned and help reducing the flanker congruency effect in the sequence blocks compared to a respective random test block. In Experiment 1, both sequence types were present during six sequence blocks. Subsequently, in two different random blocks, either the target sequence or the congruency level sequence was randomized independently. As a result, the flanker effect in both random blocks was only minimally larger than in the adjacent sequence blocks, indicating that sequence learning either had not happened or could not be used in order to reduce the flanker effect. In Experiment 2, we manipulated the sequence types between participants. As a result, the flanker effect increased slightly in the random blocks in both conditions. However, since Experiments 1 and 2 did not reveal substantial (or even significant) learning effects, we made the sequences explicit in Experiment 3. In this case, the flanker effect in the random block of the target sequence condition was much larger (and overall RT much longer) than in the adjacent sequence blocks, but the learning effect was, again, small in the congruency level sequence condition. The former result replicates an observation by Koch (2007), while the latter result suggests that even explicit knowledge of a regular sequence of congruency levels cannot readily be used for flexible conflict adaptation.