It has been shown that between-trial effects in Stroop-like interference tasks are caused by differences in the amount of cognitive control. Trials following an incongruent trial show less interference, an effect suggested to result from the increased control caused by the previous trial (the Gratton effect). In this study we show that cognitive control not only results in a different amount of interference, but also in a different locus of the interference. That is, the stage of the task that shows the most interference changes as a function of the preceding trial. Using computational cognitive modeling we explain these effects by a difference in the amount of processing of the irrelevant dimension of the stimulus.