Reckoning the moment of reckoning in spontaneous voluntary movement

COMMENTARY COMMENTARY Reckoning the moment of reckoning in spontaneous voluntary movement Sebo Uithola,b and Aaron Schurgerc,d,e,1 In everyday life, movements are sometimes triggered by external sensory stimuli, like when the traffic light ahead turns from green to red and you move your foot from the gas pedal to the brake pedal in response. But many of our voluntary movements are spontaneous, meaning that they are not tied to any recent stimulus in the current sensory environment. This capacity affords us what Gold and Shadlen refer to as “freedom from immediacy” (1). Interest in so-called “self-initiated” movements has grown considerably since the discovery of the “Bereitschaftspotential,” or readiness potential (RP), by Kornhuber and Deecke in 1965 (2), a slow buildup of neural activity in motor areas leading up to movement onset. This same slow buildup has since been observed at the single-neuron level in both humans and other animals (3–7). The RP was presumed to reflect the covert inner trigger for self-initiated movements: the process of “planning and preparation for movement” (8). One question that naturally arises is: When, if at all, along the time course of the RP does the brain make the final commitment to initiate movement? Is there a point of no return after which the sequence of action potentials becomes “ballistic” and movement, although not yet happening, can no longer be aborted? This is the question that Schultze-Kraft et al. (9) ask through a clever experiment involving a direct brain–computer interface (BCI). On-line detection of the RP allowed them to present a stop signal when the probability of an impending movement was high. This process afforded the authors a unique perspective on the inhibition of voluntary, uncued actions. In their EEG experiment, Schultze-Kraft et al. (9) asked subjects to press a button with their foot spontaneously, at a moment of their own choosing. A BCI was trained to predict upcoming actions in real time based on the RP. This approach allowed the researchers to present stop-signals at different times with respect to the onset of a self-initiated movement. Their BCI, although far from a perfect predictor of movement onset, allowed the researchers to deliver Fig. 1. The abstracted averaged results of Schultze-Kraft et al. (9). Time t = 0 is the onset of the EMG signal. The decision to act is made between −1,000 and +50 ms, but cannot be associated with a particular neural event at a smaller time scale. Note that the rather long delay between EMG onset and button press is because of the fact that for a button the foot had to be lifted and moved to the button before pressing down. stop signals much more efficiently than a random distribution would. They also made very clever use of the classifier’s false alarms, which they rightly point out may in fact have been a mix of true false alarms and “early cancellations”: trials on which the subject was able to abort the movement very early on. The authors found that most stop-signals presented after 200 ms before electromyographic (EMG) onset could not prevent the start of EMG activity; and most stop signs after 50-ms post-EMG could no longer prevent a button press. It is important to keep in mind that the button press in this study was performed with a foot pedal, and this required (at least) two movements: (i ) lifting the foot so as to position it over the pedal and (ii ) pressing downward on the pedal. This relatively long time between EMG onset and button press allowed the authors to differentiate two points of no return, one for the initial EMG activity associated with lifting the foot, and another for pressing the pedal. The results of Schultze-Kraft et al. (9) show that the movement goal—the button press—can still be withheld even when the first sparks of EMG activity have a Bernstein Center for Computational Neuroscience, Charité Universitätsmedizin Berlin, 10119 Berlin, Germany; bDonders Institute, Radboud University, 6500 GL Nijmegen, The Netherlands; cCognitive Neuroimaging Unit, NeuroSpin Research Center, INSERM, Gif-Sur-Yvette 91191, France; dCenter for Neuroprosthetics, Ecole Polytechique Fédérale de Lausanne, CH-1202 Geneva, Switzerland; and eLaboratory of Cognitive Neuroscience, Brain Mind Institute, Department of Life Sciences, École Polytechnique Fédérale de Lausanne, CH-1202 Geneva, Switzerland Author contributions: S.U. and A.S. wrote the paper. The authors declare no conflict of interest. See companion article 10.1073/pnas.1513569112. 1 To whom correspondence should be addressed. Email: aaron.schurger@gmail.com. www.pnas.org/cgi/doi/10.1073/pnas.1523226113 PNAS Early Edition | 1 of 3 appeared, in line with previous research looking for a point of no return (10). In fact, most of the aborted button presses occurred when the stop signal arrived just after EMG activity had begun. This finding suggests that the processes leading up to movement are never truly “ballistic,” but rather in constant interaction with perceptual, cognitive, and other processes. The processes might be self-sustaining to some extent, meaning that when nothing interferes, they will continue to generate the activity patterns needed to control movement, but they remain open to external influences. So what is known about the neural processes underlying a decision to perform a voluntary movement? Shenoy, Sahani, and Churchland (11) argue that voluntary movements are initiated by bringing the motor system into a region of state space from which the movement in question will ensue by virtue of the connectivity and current state of the system. It is conceivable that once activity is in this “optimal subspace,” the motor system is abruptly drawn into the local attractor, with little chance of an override, consistent with a relatively late commitment to move (12). This view is supported by recent evidence showing that action preparation can also be read out from muscle tonus (13), suggesting that during movement preparation cortical, spinal, and peripheral systems are all interacting in triggering the upcoming movement. In addition, recent physiological findings in muscle spindles show that the activation of these receptor cells in the belly of a muscle does not correlate with current, but rather with upcoming muscle length (14), indicating not only feedback to the motor cortices about planned actions, but also constant input from the cortical motor system to the muscle spindles. So in a restricted task like this (i.e., only one button), it is imaginable that the exact timing of the movement is not exclusively decided in the motor cortex, but in interaction with the skeletomotor system (12), and relatively late in the game with respect to the onset of EMG activity. The study by Schultze-Kraft et al. (9) clearly shows that a neural commitment to act cannot be identified with the onset of the RP, over 1,000 ms before EMG onset in this study, because EMG can still be prevented up to 200 ms before EMG onset. Additionally, the peak of the RP cannot be associated with such a commitment either. The RP reported here peaks at 50 ms after EMG, so by the time the RP is at its peak the movement is already in progress (Fig. 1). Schultze-Kraft et al. (9) have found two sequential points of no return: one with respect to preventing an EMG signal, situated around 200 ms before EMG, and one related to preventing the button press, roughly 50 ms after EMG onset. The presence of dual points of no return is likely specific to their task, which required first lifting the foot (EMG onset) to then depress the button. Stop signals that were presented after 50 ms post-EMG onset were too late to catch up with the unfolding movement, and muscle tension or a button press could no longer be prevented. Note that in both cases the point of no return was about 200 ms before the relevant event. The fact that this interval is consistent with the “race model” (10), according to which the processes that cause the movement and the processes that inhibit it race against each other. If the movementcausing processes win the race, then there is a movement despite the stop signal, whereas if the inhibitory processes win, then there is no movement. The race model has traditionally been used in the context of stimulus-triggered movement, in which case we know exactly when the response process began: with the presentation of the stimulus. However, there is no a priori reason why this model could not also account for spontaneous actions. Note that from the fact that a point of no return was found at −200 ms, one cannot conclude that the decision is made at that time. A point of no return at −200 ms is as compatible with a decision to move at −200 ms as it is with a decision to move at −2 days. It is, however, in line with Libet’s famous work on conscious intentions (15). Libet et al., and many replications since (but see also ref. 16), found that, when asked to report the time of their conscious decision to move, subjects reliably indicate ∼200 ms before movement onset as the decision time. Libet (17) argued that the decision to initiate movement must be preconscious, because the readiness potential appears to begin well before subjective estimates of the time of the conscious decision to move (by 300 ms or more). He tried to soften the consequences for proponents of conscious volition by suggesting that we can veto an impending movement during the final 200 ms before movement onset, which is when the decision to move goes from being preconscious to being conscious. However, according to Schultze-Kraft et al. (9), the final 200 ms before movement onset is precisely when one can no longer veto the movement. Therefore, their data argue against the idea of the last 200 ms being a window of opportunity for acts of “free won’t,” and instead suggest that the subjectively reported time of the urge to move, although imprecise, is an accurate estimate of the time of the neural commitment to move. Acknowledgments S.U. was supported by Marie Sklodowska-Curie Grant 657605 of the European Union’s Framework Programme, Horizon 2020 framework; A.S. was supported by Starting Grant 640626 from the European Research Council. 1 Gold JI, Shadlen MN (2007) The neural basis of decision making. Annu Rev Neurosci 30(1):535–574. 2 Kornhuber HH, Deecke L (1965) [Changes in the brain potential in voluntary movements and passive movements in man: Readiness potential and reafferent potentials]. Pflugers Arch Gesamte Physiol Menschen Tiere 284(1):1–17. German. 3 Romo R, Schultz W (1987) Neuronal activity preceding self-initiated or externally timed arm movements in area 6 of monkey cortex. Exp Brain Res 67(3): 656–662. 4 Lee IH, Assad JA (2003) Putaminal activity for simple reactions or self-timed movements. J Neurophysiol 89(5):2528–2537. 5 Seki T, Gemba H, Matsuzaki R, Nakao K (2005) Readiness potential and movement initiation in the rat. Jpn J Physiol 55(1):1–9. 6 Maimon G, Assad JA (2006) Parietal area 5 and the initiation of self-timed movements versus simple reactions. J Neurosci 26(9):2487–2498. 7 Kagaya K, Takahata M (2010) Readiness discharge for spontaneous initiation of walking in crayfish. J Neurosci 30(4):1348–1362. 8 Kornhuber HH, Deecke L (1990) Readiness for movement—The Bereitschaftspotential Story. Current Contents Life Sciences 33(4):14. 9 Schultze-Kraft M, et al. (2016) The point of no return in vetoing self-initiated movements. Proc Natl Acad Sci USA, 10.1073/pnas.1513569112. 10 De Jong R, Coles MG, Logan GD, Gratton G (1990) In search of the point of no return: The control of response processes. J Exp Psychol Hum Percept Perform 16(1):164–182. 11 Shenoy KV, Sahani M, Churchland MM (2013) Cortical control of arm movements: A dynamical systems perspective. Annu Rev Neurosci 36(1):337–359. 12 Schurger AA, Uithol S (2015) Nowhere and everywhere: The causal origin of voluntary action. Rev Phil Psychol 6(4):761–778. 13 Selen LPJ, Shadlen MN, Wolpert DM (2012) Deliberation in the motor system: Reflex gains track evolving evidence leading to a decision. J Neurosci 32(7): 2276–2286. 14 Dimitriou M, Edin BB (2010) Human muscle spindles act as forward sensory models. Curr Biol 20(19):1763–1767. 2 of 3 | www.pnas.org/cgi/doi/10.1073/pnas.1523226113 Uithol and Schurger 15 Libet B, Gleason CA, Wright EW, Pearl DK (1983) Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential). The unconscious initiation of a freely voluntary act. Brain 106(Pt 3):623–642. 16 Verbaarschot C, Farquhar J, Haselager P (2015) Lost in time...: The search for intentions and readiness potentials. Conscious Cogn 33:300–315. 17 Libet B (1985) Unconscious cerebral initiative and the role of conscious will in voluntary action. Behav Brain Sci 8(4):529–539. Uithol and Schurger PNAS Early Edition | 3 of 3