N–2 task repetition cost is a response time and error cost returning to a task recently performed after one intervening trial (i.e., an ABA task sequence) compared with returning to a task not recently performed (i.e., a CBA task sequence). This cost is considered a robust measure of inhibitory control during task switching. The present article reports a novel sequential effect of n–2 task repetitions when trial n–3 is taken into consideration. In particular, performance is better in trials preceded by an n–2 repetition than in trials preceded by an n–2 switch. That is, performance is better in BABA sequences (where trial n–1 was an n–2 repetition) than in CABA sequences (where trial n–1 was an n–2 switch). Likewise, performance is better in BCBA (where trial n–1 was an n–2 repetition) than in ACBA or DCBA sequences (where trial n–1 was an n–2 switch). Evidence for this new n–3 effect is provided by a mini meta-analysis of a set of published data, as well as 2 new experiments applying a different paradigm. We suggest that this new effect reflects trial-by-trial modulation of cognitive control: Task conflict is higher in n–2 repetitions than in n–2 switches; therefore, cognitive control is increased in trials following n–2 repetitions, leading to improved performance. This facilitating effect of previous task conflict is discussed with respect to current theories on cognitive control.