Dissociating Task-set Selection from Task-set Inhibition in the Prefrontal Cortex Ulrich Mayr 1 , Jo ¨rn Diedrichsen 2,3 , Richard Ivry 2 , and Steven W. Keele 1 Abstract & Patients with focal lesions in the left (n = 7) and right (n = 4) prefrontal cortex were compared with controls (n = 16) in a task-switching experiment using four differ- ent, simple spatial tasks. Each of these tasks involved a left– right decision, either regarding an arrow, the word ‘‘left’’ or ‘‘right,’’ a circle position, or the direction of a moving line. We compared performance on trials that required rule switches versus rule repetitions (local switch costs) and we compared performance between blocks with bivalent stimuli (two dimensions present) and blocks with univalent stimuli (only one dimension present) to assess global switch costs. Patients with left prefrontal lesions, but not patients with right prefrontal lesions, exhibited increased costs on trials in which the relevant dimension switched (local switch costs), but also on no-switch trials with bivalent stimuli (global costs). We also assessed task-set inhibition in the form of the backward-inhibition effect [increased response times to recent- ly abandoned tasks; Mayr, U., & Keele, S. Changing internal constraints on action: The role of backward inhibition. Jour- nal of Experimental Psychology: General, 129, 4–26, 2000]. Although left frontal patients showed normal inhibition, right frontal patients showed no evidence for inhibition. These re- sults suggest a neurocognitive dissociation between task-set selection and inhibition. & INTRODUCTION The prefrontal cortex is associated with a variety of ex- ecutive functions, including the ability to flexibly change cognitive configurations (task sets) to newly relevant task demands. Theoretically, there may be two different aspects to the process of establishing a task set. First, the relevant rules must be activated (e.g., Mayr & Kliegl, 2000, 2003; Rubinstein, Evans, & Meyer, 2001). Secondly, interference from competing task sets has to be mini- mized, possibly through a process of active inhibition (e.g., Mayr & Keele, 2000). The dissociation of these two functions has proven to be notoriously difficult (e.g., Kane, Bleckley, Conway, & Engle, 2001; Cohen & Dehaene, 1998). Here, we establish this dissociation by testing patients with lesions in the left or right prefrontal cortex and control participants in a variant of the task-switching paradigm. Task-switching situations require subjects to respond to stimuli on the basis of frequently changing stimulus–response rules (e.g., ‘‘task sets’’; Meiran, 2000; Rogers & Monsell, 1995). Response-time (RT) costs that arise from demands of switching between task rules provide an indicator of task-set selection efficiency. Re- cent variants of the task-switching paradigm have been tailored to assess specific subcomponents of control, such as the inhibition of competing task rules (e.g., Mayr & Keele, 2000). In the current paradigm, subjects had to select be- tween four different, simple spatial response rules on a trial-by-trial basis (see Figure 1), pressing one of two keys on each trial depending on the value of the response-relevant dimension. The different relevant di- mensions were the word ‘‘left’’ or ‘‘right,’’ a small circle appearing on the left or the right side, two dashed lines moving either leftwards or rightwards, and arrows point- ing either to the left or to the right. In each display, up to two dimensions could be present (see Figure 1). A verbal cue presented above the stimulus ensemble signaled the response-relevant dimension for the trial. On each trial there was a 25% chance that the cued dimension remained the same as on the previous trial and a 75% chance it changed. The contrast between switch trials and repetition trials allowed us to assess local switch costs. In addition, we contrasted blocks with bivalent and univalent stimuli. Bivalent stimuli contained both a relevant and an irrelevant dimension (e.g., an arrow and a circle when the relevant dimension was ‘‘arrow’’), and therefore, required cue-based selection of the relevant set. In univalent stimuli, only the currently relevant dimension was present, and therefore, set selection could occur in a bottom-up manner. The comparison between these two block types allowed the assessment of ‘‘global’’ task-set selection costs (also referred to as ‘‘mixing’’ costs; Meiran, 2000). In order 1 University of Oregon, 2 University of California, Berkeley, 3 Johns Hopkins University D 2006 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 18:1, pp. 14–21