Veto and Vacillation: A Neural Precursor of the Decision to Withhold Action Erman Misirlisoy and Patrick Haggard Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/26/2/296/1780468/jocn_a_00479.pdf by MIT Libraries user on 17 May 2021 Abstract ■ The capacity to inhibit a planned action gives human behavior voluntary omission but not a rule-based omission. We also used its characteristic flexibility. How this mechanism operates and the regular temporal pattern of the action sequences to explore what factors influence a decision to act or not act remain relatively brain processes linked to omitting an action by time-locking EEG unexplored. We used EEG readiness potentials (RPs) to examine averagestotheinferredtimewhenanactionwouldhaveoc- preparatory activity before each action of an ongoing sequence, in curred had it not been omitted. When omissions were instructed which one action was occasionally omitted. We compared RPs be- by a rule, there was a negative-going trend in the EEG, recalling tween sequences in which omissions were instructed by a rule the rising ramp of an RP. No such component was found for (e.g., “omit every fourth action”) and sequences in which the voluntary omissions. The results are consistent with a model in participant themselves freely decided which action to omit. RP which spontaneously fluctuating activity in motor areas of the amplitude was reduced for actions that immediately preceded a brain could bias “free” decisions to act or not. ■ INTRODUCTION with anterior areas generating abstract aspects of a plan In everyday life, our initial impulses do not always produce and posterior motor areas executing them or generating optimal actions. Most people recognize the experience stimulus-driven responses. However, the mechanisms that of deciding against saying what one really feels to avoid initiate plans at the highest level areas remain unexplained. offending a friend. In such cases, the decision to inhibit One could instead think of the voluntary motor system the action is intentional, self-generated, and often very as a loop, in which each action depends on a preceding wise. However, most previous studies of action inhibition action, rather than as a linear process with an unexplained have focused on external “stop” signals (Smith, Johnstone, initiation. Frontal cortico- BG loops (Alexander & Crutcher, & Barry, 2008; Verbruggen & Logan, 2008; Falkenstein, 1990; Alexander, DeLong, & Strick, 1986) could produce Hoormann, & Hohnsbein, 1999). These tasks capture only internally generated action sequences (Boecker et al., a part of self-control and cannot capture the familiar case 1998; Tanji & Shima, 1994; Brotchie, Iansek, & Horne, of an endogenous decision to cancel an action, referred to 1991) by chaining each action to the one before. On this as “intentional inhibition” (Filevich, Kühn, & Haggard, view, there need not be any obvious hierarchical starting 2012). Recent studies have explored intentional inhibition point for voluntary behaviors. Thinking of voluntary action by asking people to prepare an action and then decide for as an iterating loop with a characteristic activation level themselves whether to execute or inhibit it. Intentional may also explain the strong relation between cognitive inhibition produced distinct fMRI activations in medial resources and action inhibition. Inhibition of action be- pFC (Kühn, Haggard, & Brass, 2009; Brass & Haggard, comes difficult and self-control may break down when a 2007) and changes in EEG power around the time of the drive to act is sustainedly present (Baumeister, Vohs, & decision (Walsh, Kühn, Brass, Wenke, & Haggard, 2010). Tice, 2007; Mischel & Ebbesen, 1970). Here we examine Although intentional inhibition may seem closer to the opposite possibility, whether action inhibition emerges human self-control than external stop signals, the brain at points where the level of activation in a repeated mechanisms involved remain unclear. Most neurocom- behavior is momentarily reduced. putational models of voluntary action have focused on We have considered how voluntary decisions to act action generation rather than inhibition. For example, or not to act arise within a continuous action sequence. several models of frontal cortex are based on hierarchies Participants were asked to omit one particular action (e.g., Kouneiher, Charron, & Koechlin, 2009; Koechlin, within a repetitive manual motor sequence, based either Ody, & Kouneiher, 2003; Brass & Von Cramon, 2002) on an endogenous voluntary choice or an external in- struction. Incorporating intentional inhibition into a regu- lar action sequence provides a background of prepotent University College London action. As a result, it may be necessary to truly inhibit an © 2013 Massachusetts Institute of Technology Published under a Creative Journal of Cognitive Neuroscience 26:2, pp. 296–304 Commons Attribution-NonCommercial 3.0 Unported (CC BY-NC 3.0) license. doi:10.1162/jocn_a_00479 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn_a_00479 by guest on 26 September 2021 action, as opposed to merely failing to initiate it (Filevich For the first nine participants, the instruction for the et al., 2012; Kühn et al., 2009). Furthermore, studying a rule-based condition was to omit every fourth keypress. regular sequence of actions helps to fix action timing, However, preliminary inspection of the data showed that allowing us to infer when inhibition of action should occur, voluntary omissions tended to occur more rarely than if it is present. this, producing a confound between condition and motor We investigated the relation between preparation of activity. Therefore, the remaining nine participants were the next element in a continuous motor sequence and instructed to omit every sixth keypress. This successfully the voluntary decision to omit an action by examining the balanced overall action and omission frequency across Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/26/2/296/1780468/jocn_a_00479.pdf by MIT Libraries user on 17 May 2021 late, lateralized component of the readiness potential conditions. (RP; Shibasaki & Hallett, 2006). Because this component immediately precedes voluntary actions (Matsuhashi & Hallett, 2008; Sirigu et al., 2004; Libet, Gleason, Wright, Procedure & Pearl, 1983), its magnitude offers a valuable neural sig- Participants first practiced synchronizing keypresses to a nature of the generation versus inhibition of voluntary 0.5-Hz auditory metronome, allowing them to learn the re- action. We described above a nonhierarchical loop model quired rhythm. In the subsequent tasks, no external pacing for voluntary chaining of sequences of motor actions. Such stimulus was given. Participants were asked to avoid count- models predict that activation states are passed around ing seconds in their timing and simply to follow the rhythm successive iterations of the loop. On this view, the genera- they had learned in the synchronization phase. In all trials, tion or inhibition of a current action may depend on the participants were first presented with an instruction to level of system activity associated with a previous action. “Press space to begin the trial.” Following the keypress, For example, a falling level of activation over successive a white fixation cross on a black background appeared loop iterations may eventually lead to an inhibition or centrally on the monitor, which participants were asked failure to generate the next action. to remain fixated on while continuing keypresses at an internally paced rate of approximately 0.5 Hz. In the practice task, participants produced a sequence METHODS of 30 keypresses at 0.5 Hz in each trial. Then the experi- Participants mental tasks began. In the rule-based omission condition, participants were instructed to omit every fourth or every Twenty-six right-handed participants (15 men, 11 women) sixth keypress (see Design). In the voluntary omission were tested. Eight participants were excluded (three had condition, participants freely chose when to omit key- excessive blink/ EOG artifacts, one could not produce a presses. Precise timing between keypresses was incenti- regular action sequence, one could not avoid tapping vized in the experimental session. Participants received their foot in addition to their finger, one made excessive feedback about their performance at the end of each trial. finger movements between actions, and two made too The mean interval between their actions had to be within few voluntary omissions for ERP analysis to be possible), 1700–2300 msec, and the standard deviation of these in- leaving 18 participants with usable data. All had normal tervals had to be below 200. Omitted actions required an or corrected-to-normal vision, and none had a history extended average interval of 3500–4500 msec between of neurological or psychiatric disorders. consecutive actions and standard deviation below 500. Fifteen pence per trial was gained for passing all criteria. Optimal performance could potentially increase participant Design income by £6. In all tasks, participants pressed the space button on a keyboard with their right index finger once every 2 sec EEG in a self-paced manner. In the rule-based omission task, participants used a rule given by the experimenter to omit Twenty-seven EEG channels (FT8, FC6, FC4, FC2, FCz, every fourth or sixth keypress in the sequence. In the volun- FC1, FC3, FC5, FT7, T8, C6, C4, C2, Cz, C1, C3, C5, T7, tary omission task, participants were instructed