This study examined the effect of an explicit relational rule on sequence learning in a 3-choice serial reaction time task. Simple probabilistic contingencies between pairs of response cues were used in such a way that the sequence of cues moved predominantly in one direction (i.e. either clockwise or counterclockwise). Performance on cued and miscued responses was compared for a group given a hint about the abstract rule describing the relationship between the response cues, and a group given no information about this relationship. Experiment 1A demonstrated that XYZ and XYX subsequences showed performance differences when the location of the target on each trial was random. Experiment 1B showed that giving participants the explicit hint affected XYZ subsequences more than XYX subsequences. Implications for sequence learning and, specifically, the interaction between rule and instance learning are discussed.