We have a “sense of ownership” of our own body, which can be disrupted or lost in cases of body perception disturbances, somatoparaphrenia, out-of-body experiences, or depersonalization.

Likewise, we have a “sense of ownership” of volitional actions; intentionally moving a part of one’s body feels like a choice.

We might speak of an intentional choice as a conscious choice. Volition relates to consciousness because it relates to the self. There is “somebody” who has perceptions, who is in a body, and who does things.

If you simply “find yourself” taking actions, without deliberate choice, you may be consciously aware of the actions and their results but not of the decision; the cause, the generator, is “invisible” to you. Unconscious. You can feel your heartbeat, but you have no way to perceive your heart “deciding” to beat. By contrast, the decision to lift your arm is “visible” to your conscious mind; you have a direct experience of perceiving choice. In fact, if you didn’t have that transparency, your sense of “embodiment” or “ownership” would probably be shaken; an arm you couldn’t control wouldn’t feel like your arm.

Neural Correlates of Volition

There’s a whole philosophical debate on free will, obviously, and I’m not entirely up to speed on it. But from a more practical perspective, we can certainly point to the thing humans mean by a decision/choice/volitional act.

When you move your hand, it ordinarily feels like you “decided” or “chose” to do it. It does not, by contrast, feel at all like a choice when your heart beats, when your intestines churn, when your leg jerks reflexively from a rubber hammer hitting your knee, or when a neurological patient experiences motor tremors (e.g. in Parkinson’s.) We have the internal experience of choice for some motions and not others.

Also, we can see from the outside that behaviors we’d ordinarily call volitional (brain-directed skeletal muscle movements in healthy individuals) are capable of much more flexible and complex control than non-volitional ones. Sure, your thoughts, emotions, and perceptions can affect your heart rate or digestion. But you can’t get your heart to beat out the Gettysburg address in Morse code, whereas you can easily learn to type it out with your fingers or say it with your mouth. There’s a qualitative difference.

Regardless of whether there’s some upstream cause behind volitional actions, we can point at movements that are central examples of “volition” in the colloquial sense, and central examples of “involuntary motion”, and try to understand how they differ at the neurological level from each other and from more ambiguous edge cases.

Libet Experiments and W-Time

Back in 2023 I looked at one neural signature of volition; the “readiness potential”.

The famous “Libet experiment¹” asked subjects to choose to move at any time they want, and to recall the position of a revolving spot at the moment they made the decision to move. The “readiness potential”, a slow rise in voltage detectable on the EEG, precedes the action by about 550 ms; the self-reported conscious decision to move only happened at 200-150 ms before the motion.

The Libet experiment has popularly been taken as a “proof that free will doesn’t exist”, because the brain is “preparing to move” before we are even aware of making a choice to move.

But Libet himself never interpreted his experiment as a proof of the nonexistence of free will. And in fact a later experiment he conducted points to the contrary.

When subjects were asked² to wait for a visual signal and then move their hands, a “readiness potential” detectable with the EEG arises about half a second before their hand actually moves. If they’re also asked to “veto” the movement just prior to moving, they also display the “readiness potential” about 500 ms before the 0 time, but it suddenly shifts direction at 200-150 ms.

This means that the actual difference between brain activity between people who choose to move and people who choose not to move happens, not at the onset of the readiness potential, but later, at the same time that the subjective “decision to move” happens.

…

Schurger also replicated the same phenomenon that Libet himself did — when you compare subjects asked to move spontaneously with subjects who don’t move at all, the brain activity of the two groups only diverges significantly 150 ms before movement, not the 500 ms before that would be expected if the readiness potential represented the decision to move.⁴

The “point of no return”, the 150-200 ms pre-movement when subjects say they decide to move, is also known as W-time.1

I view it as a better correlate of volition than the readiness potential. W-time is both the time when people say they make a decision, and the time when people in fact do make their final decision whether or not to move.

It may be that the readiness potential corresponds to a sort of “inner impulse” to move, a potential possible movement, but the “decision whether or not to commit” happens later, at W-time.2

W-time is affected by perceptions of movement. If you give people a video delay or audio delay that makes it appear that their hand moves several tens of milliseconds later, then their self-reported W-time moves later by the same amount. In other words, the perception of volition may be “made of” the combination of a motor signal and the sensory observation that one has in fact moved as intended.3

You might interpret this as a sign that W-time is a bad metric — people are retrospectively confabulating the time at which they “made a decision”, in ways affected by later information, rather than accurately reporting the moment at which they first had the internal sensation of choice.

However, there is some evidence that there are neurological changes at W-time, localized to particular brain regions, and thus that it’s not just a retrospectively imagined, subjective timestamp, but also an objective moment at which decision-relevant events happen in the brain.

Interestingly, patients with parietal lesions don’t report experiencing a W-time decision preceding their actual movement; they first “intend to move” almost exactly when they do move. These parietal lesion patients also didn’t show the usual slow-rising readiness potential; their EEG was flat for the ~2s before movement. 4

Likewise, Parkinson’s patients show delayed W-times; they also don’t notice a decision to move until shortly before they do.5

In a study on epileptic patients with implanted electrodes, there’s a population of neurons in the anterior cingulate cortex and supplementary motor area (both in the frontal lobe) whose firing rates change significantly at W-time. By contrast, temporal neurons didn’t exhibit this kind of significant change.6

The supplementary motor area seems especially important in urges to move. Stimulating it electrically (in epileptic patients) induces what patients report as urges to move, even if they don’t actually move at all. 7

Neurons in the primary motor cortex, M1, also show a spike in activity coinciding with the timing of self-reported intentions to move, in the case of a tetraplegic individual equipped with a brain-machine interface that measures M1 activity and translates it into signals to the nerves and muscles to generate movements. In cases where the BMI was programmed to generate involuntary movements, M1 activity only spiked after the movement; while in cases where the BMI was programmed not to send a neuromuscular signal in response to M1 activity, M1 activity spiked at precisely the same time as the self-reported (and ineffectual) choice to move.8

In ordinary voluntary motion, we perceive two events, sequential rather than simultaneous: the choice to move and then the sensory perception that we are in fact moving. It’s possible to dissociate the choice to move from actual movement; we can detect something in the brain happening before intentional movement, which doesn’t occur in involuntary movements, and which matches up temporally with the perceived moment of choosing-to-move.

Perceptions of Agency

The classic “comparator model” of agency is that we feel in control of an action when we accurately predict its sensory consequences.

That is, somewhere in our brain we’ve made an “efference copy” of every motor command, and we run a “forward model” of what we expect to perceive from it. Predicted sensory effects of self-generated actions are thus distinguishable from stimuli coming in from the environment; for instance, you can’t tickle yourself, because being tickled is surprising and using your own fingers to make tickling movements is not.

In fact, if people remotely control a robot to tickle them, the ticklish sensation is proportional to the (artificially programmed) delay between the human’s guidance and the robot’s motion. Using fMRI, it’s possible to detect neural correlates of this effect; there’s less activity in the somatosensory cortex and the right anterior cerebellar cortex from self-generated touches than from external touches.9

The cerebellum is a good candidate for the seat of the “forward model” that rapidly simulates the sensory consequences of self-generated action, and compares it to actual sensations, resulting in “surprise detection.” Then, elsewhere in the brain, activity representing sensory perceptions (such as touch perception in the somatosensory cortex) may be attenuated if those perceptions are “unsurprising”.

Introducing distortions and time lags in virtual reality contexts can lead experimental subjects to no longer think they are viewing “their own” movements.10 Analogously to the rubber-hand illusion, if you observe data consistent with you controlling your own movements predictably, they’ll feel like “your own”, whereas too many inconsistencies will make them read as “other.”

If you let subjects navigate a cursor to complete a computer task, impeding their agency (adding turbulence so the cursor doesn’t move with the subject’s motions) increases activity in the right temporoparietal junction (rTPJ), and stimulating the rTPJ with transcranial magnetic stimulation decreased the (self-reported) sense of agency.11. Remember that stimulation to the TPJ can also induce out-of-body experiences; it may be involved in “not-me, not-mine, not-my-body, not-caused-by-me” types of judgments generally, possibly due to mismatches of different types of sensory information.

Distorted Perceptions of Agency in Schizophrenia

Subjects with schizophrenia may perceive themselves as having more agency than normal subjects, even to the point of believing they “caused” responses that preceded their actions12.

Clinically, schizophrenics make errors about agency in both directions — they both believe that their own thoughts and actions are under external control (delusions of influence, thought insertion, hallucinations) and believe that their minds remotely/acausally control events in the world (delusions of control.)

In lab experiments, schizophrenics have less sensory attenuation from self-generated actions (the normal effect where e.g. auditory signals from your own voice are weaker than those from external sounds). Schizophrenics hear their own voices as sounding more like someone else’s voice; they can tickle themselves; essentially they “surprise themselves” more than non-schizophrenics do.

Separately, schizophrenics also tend to be less sensitive to discrepancies (like time delays) between their actions and sensory responses, leading to them over-attributing self-agency (i.e. too frequently believing it was their own actions that caused a response.)

These two observations work in opposite directions; to oversimplify a bit, it’s like schizophrenics perceive everything as not-self, so they have to infer what’s “their own” agency using external cues, and often (but not always) overestimate how much is “theirs.” Thus, you get both symptoms of too much and too little “sense of agency”. 13

Involuntary Movements

Various motor disorders involve involuntary movements. How do these differ neurologically from ordinary voluntary movements?

It depends on the kind of involuntary movement.

Reflexes (which are an example of involuntary movement in healthy people) do not reach the brain at all. Reflexes only pass through the peripheral nerves and spinal cord.

Problems with the basal ganglia, aka “extrapyramidal hyperkinetic movement disorders”, can result in various unintentional movements such as resting tremor (repetitive trembling), chorea (“dance-like” jerking), ballism (violent flinging), or dystonia (clenching). Unlike voluntary movements, these don’t originate with motor planning activity in the cortex; these are “bottom-up” movements where the basal ganglia don’t provide enough inhibitory input into the thalamus, so the thalamus stimulates the primary motor cortex abnormally, causing unplanned movements.

Then there are “semivoluntary” motor disorders, like Tourette’s. Tics look almost entirely like voluntary motions neurologically, and they’re preceded by resting potentials and even “urges” to tic, but they’re experienced as a loss of control from a psychological perspective. OCD compulsions, some stereotypies (as in autism), and functional movement disorders, are similarly in this “semivoluntary” bucket; there is a sense in which the patient “can’t control them”, or does not reflectively endorse them, but there is some sort of motivation to move, and the “higher” cortical motor planning regions are involved, just as they are in ordinary voluntary movement.14

In the rare “alien hand syndrome”, often a symptom of epilepsy or stroke, one hand (usually the left) seems to move “with a will of its own”, seemingly goal-oriented but contrary to the patient’s wishes. In an fMRI study, alien hand movements (but not normal voluntary movements of either the left or right hand) were associated with the precuneus and inferior frontal gyrus, as well as all the areas usually involved in motor control. 15

In alien hand syndrome, there’s typically damage to any of various cortical regions involved in motor planning and sensory integration, leading to disinhibition of movements, and/or failure to perceive one’s own hand movements, leading to the perception that the hand has “a mind of its own.”1617

It seems that it’s less that any one neural phenomenon is involved in involuntary movements, and more that everything has to be working perfectly and well-integrated for movements to be confined to the voluntary ones.

Conclusion: What Is Voluntariness?

Very roughly, the evidence looks consistent with: “we perceive action as “our own choice” when we can perceive an inner experience, which precedes action, which we infer causes action, and whose results are coherent and match our predictions.”

The “inner experience” we know as volition probably lives somewhere in the cortex, maybe frontal or parietal.

When there is no “plan to act” in the cortex preceding a motion, it is clearly not willed, like reflexes or extrapyramidal involuntary motions.

The “plan” or “decision” to act really is a measurable phenomenon that really does precede motion itself, and objectively happens pretty much at the same time we internally perceive it.

We have a monitoring system (involving the cerebellum and some parietal regions, at least) for determining whether actions are “self-generated” or “other-generated”, via whether their results are predictable or unpredictable from our own motor signals, and whether their results are coherent across different sensory modalities. This system can get confused via artificially induced distortions, TPJ stimulation, sensory and motor disorders, and schizophrenia.

I don’t totally understand the relationship between the “self-other monitoring” system and the decision-to-move. Are we tracking the final M1 motor signal that initiates motion (sending a signal that ultimately goes to the spinal cord and muscles), or a preceding cortical decision-to-move, or both?

But basically it seems that “the self” relates to “consciousness” through the fact that we consciously perceive the internal mechanisms of the self’s actions, whereas we only perceive the “outside” of external phenomena. If we can “inspect the gears” directly, if we can see/feel the decision being made, and if the results are consistent with what we predict happening because of that decision, then it’s perceived as “self-generated” or “volitional”; otherwise, not.