The idea that adaptation to stimulus or response conflict can operate over different time scales takes a prominent position in various theories and models of cognitive control. The mechanisms underlying temporal variations in control are nevertheless poorly understood, which is partly due to a lack of appropriate empirical measures. Inspired by reinforcement learning models, we developed a method to quantify the time scale of control behaviorally, by computing trial-by-trial effects that go beyond the preceding trial. Briefly, we extended the congruency sequence effect from 1 trial to multiple trials into the past and quantified the influence of previous trials on current-trial performance as a function of trial distance. The rate at which this influence changes across trials was taken as a measure of the time scale of control. We applied the method to a flanker task with different conflict frequencies and volatility. Results showed that the time scale of control was smaller in rare-conflict and volatile contexts, compared to frequent-conflict and neutral contexts. This is in agreement with theories differentiating transient from sustained control. The method offers new opportunities to reveal temporal differences in control modes and can easily be applied to various empirical paradigms.