Performance of task sequences is assumed to rely on activation and inhibition of tasks. An empirical marker of task inhibition is the so-called n - 2 repetition cost, which is assessed by comparing performance in trial n - 2 task repetitions (i.e., ABA) with that in n - 2 task switches (i.e., CBA). Current theoretical accounts assume that inhibition acts on the level of task representations (i.e., task sets). However, another potential target of task inhibition could be the representation of the task cue. To decide between these two alternatives, the authors used a 2:1 cue-to-task mapping design. They found significant n - 2 task repetition costs both with n - 2 cue repetitions and n - 2 cue switches. These costs were about equal (Experiment 1), and this data pattern was found for both short and long cuing intervals (Experiment 2). Together, the data suggest that task inhibition acts on task sets and not on cue representations.