It is typically assumed that the conscious experience of wanting to move is not the driving force for motor planning, but the secondary consequence of the unconscious neural processes preparing the movement. A recent study by Schneider et al. [Schneider, L., Houdayer, E., Bai, O., & Hallett, M. What we think before a voluntary movement. Journal of Cognitive Neuroscience, 25, 822–829, 2013] seems consistent with this dominant view by showing that the brain can be preparing to make voluntary movements not only “prior to the conscious appreciation that this is happening” but also “while subjects are thinking about something else.” However, an alternative hypothesis exists. It is supported by several lines of evidence and suggests that the early neural signals recorded by Schneider et al. (and others) do not reflect movement preparation per se, but rather a buildup in neural activity that ultimately leads to the emergence of a conscious intention to move. According to this view, the conscious experience of wanting to move is not the consequence but the cause of movement initiation.
Every day, we make voluntary actions without really “thinking about it.” In some cases, we can even produce internally generated movements (i.e., movements that are not triggered by external imperative stimuli) in the absence of any apparent conscious intention. For instance, we can easily drive a car while talking to someone, eat popcorn while watching a movie or imitate the posture of our partners during social interactions (Chartrand & Bargh, 1999). These behaviors are generally assumed to occur automatically, independently of the main attentional focus of the participant. A recent study by Schneider et al. extends this observation by suggesting that human individuals can experience varied thoughts while initiating a voluntary movement (Schneider, Houdayer, Bai, & Hallett, 2013). These authors instructed naive participants to perform self-paced extensions of the right wrist while observing a rotating clock. The participants had to move “as spontaneously as possible.” Using real-time EEG recordings, the authors were able to predict movements with high accuracy (>80%), up to 1.5 sec in advance of their occurrence. At the time of prediction, the participants were asked about their ongoing thoughts. Although they were often thinking about their movements, they also reported thinking about a topic unrelated to the upcoming movement in almost a third of the trials. According to the authors, this result indicates that “the brain can be preparing to make voluntary movements while subjects are thinking about something else.”
Overall, the study by Schneider et al. has two major interests. First, it confirms previous reports by the same (Bai et al., 2011) and other groups (Fried, Mukamel, & Kreiman, 2011; Wessberg et al., 2000), that it is possible to predict the occurrence of a movement in advance of its initiation. Second, it generalizes from automatic to voluntary self-generated actions the finding that motor responses can be initiated while the mind is wandering or thinking about various things. This observation should open the door to future research. However, the interpretation of these results requires addressing important questions.
With respect to this point, a first issue of interest concerns the idea that the early neural signals identified by Schneider et al. “are associated with the brain's programming of a movement.” According to this widely discussed hypothesis (Hallett, 2007; Haggard, 2005), “the brain begins the preparation to move prior to the conscious appreciation that this is happening, and the sense of volition gradually develops in conscious awareness thereafter.” However, an alternative hypothesis does exist. It suggests that early neural signals are not related to movement preparation per se, but to a buildup in neural activity that ultimately leads to the emergence of a conscious intention to move (Fried et al., 2011). In other words, the early neural signals identified in advance of conscious intention might not reflect motor planning but the progressive emergence of motor intention which, in turn, triggers motor planning. Within this context, the ability of the subjects to think about nonmotor matters in trials where a movement is predicted by the model and actually occurs might reflect a threshold effect. In this case, the intention to act that leads ultimately to movement planning would not generate a sufficiently large neural response to overcome competing thoughts and emerge into consciousness (Fried et al., 2011). In agreement with this view, strong evidence exists in the literature that unconscious motor activations, related or not to an ongoing task, can occur in response to various internal states and external cues (Sumner & Husain, 2008; Dijksterhuis & Bargh, 2001; Bargh & Chartrand, 1999).
Several lines of evidence support the view that early neural signals do not reflect movement preparation but the emergence of a conscious intention. The most recent one comes from the demonstration that the same early electrophysiological responses are present when subjects make a conscious decision to move or not to move (Trevena & Miller, 2010). This finding fits well with the chronometric outcomes of several studies that have used Libet's paradigm. In his pioneering study, Libet and his coworkers asked human volunteers to fixate a single clock hand rotating on a screen (Libet, Gleason, Wright, & Pearl, 1983). The task was to press a button with the right index finger whenever the participants “felt the urge” to do so. After this movement, at a random time, the clock stopped and the participants were required to report the position of the clocks' hand at the time they first became aware of their will to move (W-Judgment). The readiness potential was collected. The W-Judgment was found to precede movement onset by around 200 msec. At the same time, the readiness potential was found to precede the W-Judgment by about 1 sec. During the last two decades, numerous studies have replicated these results. Typically, anticipation of the W-Judgment with respect to movement onset was reported to vary between 200 and 350 msec, whereas the readiness potential was found to precede the W-Judgment by about 1.5 sec (a value that is close to the predictive window used by Schneider et al.; Lau, Rogers, Haggard, & Passingham, 2004; Sirigu et al., 2004; Haggard & Eimer, 1999). Interestingly the anticipation of the W-Judgment with respect to movement onset (∼275 msec) represents roughly the delay required to initiate simple reactive movements in response to visual or auditory cues (Ballanger et al., 2006; Desmurget et al., 2004; Desmurget, Grafton, Vindras, Grea, & Turner, 2003). Such a temporal proximity is consistent with the claim that the actual kinematic details of the intended movement are not specified when conscious intention emerges. In agreement with this interpretation, recent evidence shows that conscious motor intentions, triggered by stimulating the inferior parietal lobule, are not formulated as precise motor acts, but as broad action categories (Desmurget & Sirigu, 2009, 2012). Typically, in response to electrical stimulation, the patients report that they feel a desire to talk or move a body segment. When asked whether they can describe the kind of movement they want to do, these subjects answer that they do not know and that they just want to move. A typical example is as follows: Patient: “I wanted to move my foot.” Experimenter: “Which foot?” Patient [showing his left leg]: “This one.” Experimenter: “How did you want to move it?” Patient: “I don't know, I just wanted to move it” (Desmurget et al., 2009). In an interesting experiment, Sirigu and colleagues submitted brain-damaged patients and healthy participants to Libet's paradigm (Sirigu et al., 2004). In healthy participants and cerebellar patients, the W-Judgment preceded movement onset by 250 msec and a readiness potential was recorded more than 1 sec before button press. In parietal patients, by contrast, the W-Judgment occurred only tens of milliseconds before movement onset and no early readiness potential was detected. This shows that parietal patients had lost the early subjective experience of conscious intention although they had not lost the late ability to prepare and execute the required action. The same type of dissociation is found in patients with “alien hand syndrome,” a disorder in which movements are performed outside of volitional control (Scepkowski & Cronin-Golomb, 2003). For instance, in a recent study, relevant to the present discussion, Assal and his coworkers investigated the neural bases of this disorder in a patient harboring a right posterior parietal lesion (Assal, Schwartz, & Vuilleumier, 2007). Using fMRI, these authors studied brain activity while the left arm of this patient was making uncontrolled flexion–extension of the fingers. Results indicated that the primary motor cortex was the only brain region activated during the occurrence of these movements. It was suggested that the absence of functional input from the intentional parietal regions caused this “alien” motor activation. This hypothesis fits well with the demonstration that, in resting situations, the parietal cortex inhibits the spontaneous neural activity within the motor cortex (Jaffard et al., 2008). In light of all these observations, one may wonder whether the powerful real-time algorithm developed by Schneider et al. really detects, as suggested, a neural correlate of movement preparation, or whether it measures the buildup of neural activity that leads to the emergence of a conscious intention to move. This divergence is not minor considering the ethical and philosophical implications of this kind of data for our conception of volition and free will (Smith, 2011; Haggard, 2008; Hallett, 2007). Also, at a more pragmatic level, the possibility that the early neural signals decoded by Schneider et al. reflects the emergence of syncretic motor intentions may question the usefulness of these signals for managing the “efferent limb of a brain computer interface” through the identification of a “variety of intentional actions.” More specific signals, arising later during motor preparation, might be required to achieve this goal.
Functionally, the “planning then intention” hypothesis proposed by Schneider et al. (Schneider et al., 2013) in agreement with Libet's view (Libet et al., 1983) and the alternative “intention then planning” assumption discussed above entail a fairly strong seriality. However, the human brain is certainly able to perform multiple processes in parallel (van der Meer, Kurth-Nelson, & Redish, 2012; Burnod et al., 1999; Mesulam, 1998). Therefore, it could be tempting to speculate that motor planning and the emergence of conscious intention are generated concurrently and, as a consequence, that none of the serial models above is valid. According to this view, the act of planning would promote/sustain an intention to act or, alternatively, the emerging feeling (not yet conscious) of motor intention would prompt/nurture movement preparation. Although appealing, this “intention and planning” hypothesis also raises questions. In particular, if both processes evolve in parallel, the early electrophysiological signal recorded prior to movement should reflect a compound of both activities. But, if such is the case, the absence of early electrophysiological response in patients with posterior parietal lesions becomes difficult to explain (Sirigu et al., 2004). Also, the parallel assumption makes it hard to account for the long delay that exists between actual movement onset and the time of W-Judgment (∼275 msec on average); a delay that is exactly in the range required for planning a goal-directed movement (see above). Finally, all these models, whether serial or parallel, could be questioned based on the assumption that the type of paradigm used by Schneider et al. does not really measures “intentionality.” Indeed, in this kind of task the intention to move is driven by an externally imposed instruction. It is inherent to the task and specified a priori by the experimenter. Here, for instance, the participants were “asked to perform a self-paced, voluntary movement task of right wrist extension (…) as spontaneously as possible, within the constraints of the experiment.” The authors recognize that these movements are not “fully spontaneous,” but they argue that they “have a strong sense of volition.” This is true. However, volition does not concern the type of movement to be performed. It only applies to the time of movement occurrence. Therefore, it could be argued that the movement is already planned before the onset of any trial (i.e., loaded in the sensorimotor network) and that the experimental task amounts, in fact, to a classical go/no-go paradigm. According to this view, early neural activity would not reflect a process of motor planning, but rather a release of inhibition when the subject decides that it is time to elicit the movement. This view fits well with the hypothesis that the mesial precentral areas mediate the W-Judgment task in Libet's paradigm (Fried et al., 2011; Haggard, 2008; Lau et al., 2004). Indeed, this region has been shown to control movement onset by releasing the inhibitory action it exerts on the primary motor cortex (Ball et al., 1999). With respect to this point, recent evidence has suggested that the readiness potential emerges in mesial precentral regions as a result of parietal input (Desmurget & Sirigu, 2012). In patients with parietal lesions, this input is lost, which may explain the disappearance of the early subjective experience of conscious intention (Sirigu et al., 2004).
Leaving aside the issue whether the type of paradigm used by Schneider et al. (Schneider et al., 2013) measures intentionality, it is not totally spurious to hypothesize that, in fact, different motor contexts can be associated with different preparatory processes. In novel contexts and/or environments with little or no affordance, the decision to act would precede motor planning, as suggested earlier in this comment. By contrast, in response to specific internal states (e.g., thirst) or when actions–primers are present (e.g., graspable objects, press-button, overlearned cues, etc.), the brain might automatically prepare to act, which, in turn, might favor the emergence of a conscious intention to act. This emergence might either follow (serial models) or accompany (parallel models) the process of motor planning. Of course, these considerations remain speculative at this point and further researches will be necessary to evaluate their validity.
All this being said and beyond the fact that they might be subject to alternative interpretations, the results reported by Schneider et al. make an important contribution to the literature. They eloquently confirm that noninvasive EEG can be used to reliably predict the occurrence of a movement. Also, for the first time, they provide a direct experimental proof that a neural signal predicting the occurrence of an upcoming voluntary self-generated movement can occur while the subject is not aware of preparing this movement. These findings are of interest not only for neuroscientists but also for anyone who is interested in the issue of consciousness, including clinicians, philosophers, and theologians. As emphasized by the authors, this work offers questions and methods that should help us “to delve more deeply into the nature of volition and agency.”
Reprint requests should be sent to Michel Desmurget, CNRS, Cognitive Neuroscience Center, 67 bd pinel, 69500 Bron, France, or via e-mail: firstname.lastname@example.org.