Brain activity precedes intention

The existence of free will when making a simple voluntary motor movement was called into question by a study published by Libet et al. (Reference Libet, Gleason, Wright and Pearl1983). In their experiment, electroencephalography (EEG) and electromyography (EMG) were used to monitor brain activity and muscle movement in subjects instructed to make a voluntary movement of their hand. After a designated start time, subjects made a spontaneous movement whenever they felt like doing so while noting the time of the urge to move by simultaneously watching a clock. Although the urge to move preceded the movement, as one would expect, the study found that this intention was itself preceded by cerebral activity, called the ‘readiness potential’ (RP), detectable several hundred milliseconds before the conscious intention to move. This startling discovery that the time of conscious intention lagged behind the RP suggested that simple voluntary movement does not arise de novo by conscious intent or free will, but rather is already set into motion by unconscious neural activity that precedes it. Libet stopped short of concluding that free will was non-existent by proposing that conscious decision making could abort or veto the movement (Libet, Reference Libet1999), leaving room for a kind of ‘free won't’ (Obhi & Haggard, Reference Obhi and Haggard2004).

Libet's results have been extensively critiqued throughout the years, citing both methodological and interpretive problems (Gomes, Reference Gomes1998; Pocket, Reference Pockett2002; Van de Grind, Reference Van de Grind2002; Danquah et al. Reference Danquah, Farrell and O'Boyle2008; Trevana & Miller, Reference Trevena and Miller2010; Klemm Reference Klemm2010), not the least of which is the potential unreliability of measuring the precise time of conscious intentFootnote ¹ Footnote †. However, Libet remained a staunch defender of his data (Libet, Reference Libet2000, Reference Libet2002, Reference Libet2003) and the essential finding has been replicated through refined experiments by other investigators (Haggard & Eimer, Reference Haggard and Eimer1999; Trevana & Miller, Reference Trevena and Miller2002; Matsuhashi & Hallett, Reference Matsuhashi and Hallett2008).

Free will as illusion

Further research using novel versions of the Libet experiment have revealed the presence of the lateralized RP (LRP), a component of the RP that represents specific preparatory activity for movement on the contralateral side of the body (Haggard & Eimer, Reference Haggard and Eimer1999; Trevana & Miller, Reference Trevena and Miller2002). Detection of the LRP suggests that the unconscious neural activity during a simple motor task does not simply reflect general preparation or a decision of when to move, but rather what specific target will be selected long before conscious awareness of that decision. Further technological updates have confirmed that unconscious brain activity occurring before conscious intention is specific to a subject's choice of target in a motor task, to the point of being able to predict the movement prior to conscious intention. For example, Soon et al. (Reference Soon, Brass, Heinze and Haynes2008) used functional magnetic resonance imaging (fMRI) in place of EEG to monitor subjects during a motor task in which they could freely chose to press a left or right button and found that brain activity predictive of that choice (albeit with only about 60% accuracy) preceded the conscious decision by as much as 10 s (Soon et al. Reference Soon, Brass, Heinze and Haynes2008). Fried et al. (Reference Fried, Mukamel and Kreiman2011) likewise replaced EEG with implanted electrodes used to monitor the electrical activity of individual neurons within the brains of patients with epilepsy who were engaged in a variant of the Libet task (Fried et al. Reference Fried, Mukamel and Kreiman2011). They showed that neuronal firing rate and the degree of neuronal recruitment occurring before the conscious decision to move could predict whether the subject would move (with about 80% accuracy), what lateralized movement would be made (about 70% accuracy), as well as when the subsequent sense of conscious intention would occur. These results support a model in which our subjective experience of free will during a voluntary motor act including intention (conscious planning to move) and agency (the subsequent causal ownership of that movement) ‘emerges as the culmination of premotor activity starting several hundreds of milliseconds before awareness' (Fried et al. Reference Fried, Mukamel and Kreiman2011).

Other experimental strategies have demonstrated the ability to influence decision making for voluntary motor acts through undetected external stimulation. The first such strategy involves ‘backward masking’ in which a sensory stimulus (e.g. a visual image on a computer screen) is presented followed quickly by a larger one in such a way that the first, smaller stimulus escapes conscious awareness. Despite the masking of the first ‘subliminal prime’ from consciousness, researchers have demonstrated that such masked stimuli can trigger specific motor responses during forced-choice tasks (e.g. voluntarily moving either the right or left index finger) (Taylor & McCloskey, Reference Taylor and McCloskey1996; Eimer & Schlaghecken, Reference Eimer and Schlaghecken1998; Schlaghecken & Eimer, Reference Schlaghecken and Eimer2004). A second experimental strategy has featured the application of a transcranial magnetic stimulation (TMS) pulse to relevant motor areas in order to unconsciously influence forced-choice motor response. This effect was demonstrated in two studies (Ammon & Gandevia, Reference Ammon and Gandevia1990; Brasil-Neto et al. Reference Brasil-Neto, Pascual-Leone, Valls-Sole, Cohen and Hallett1992), but not replicated in a more recent experiment (Sohn et al. Reference Sohn, Kaelin-Lang and Hallett2003). As with the Libet-type experiments, such findings indicate that unconscious brain activity that precedes conscious intention can bias the outcome of simple motor acts. Although these motor responses are experienced as acts of free will, it appears that even under the relatively simple mechanics of motor activity, a variety of processes may bias choice and action before and without our conscious awareness. Such evidence has led some to label our subjective sense of free will a mere ‘introspection’ (Hallett, Reference Hallett2007), a causal inference that is little more than an ‘illusion’ (Wegner, Reference Wegner2002).

The neuroanatomy of volition

In the 1950s, Wilder Penfield applied electrical stimulation during neurosurgical procedures to different brains regions of epileptic patients that resulted in a variety of responses including involuntary vocalisms and simple motor movements (Penfield, Reference Penfield1958). Subsequent experiments with electrical brain stimulation have provoked a wide range of behaviors including speech arrest, automatisms, re-experienced memories and déjà-vu, mood and anxiety responses, sensory experiences and multimodal hallucinations, and cognitive impairment (Selimbeyoglu & Parvizi, Reference Selimbeyoglu and Parvizi2010). Such research has contributed greatly to our understanding of the functional neuroanatomy of the human brain.

Electrical stimulation applied specifically to the supplementary motor area (SMA) has been shown to produce an irrepressible urge to make a voluntary contralateral movement in the absence of any overt motor activity (Fried et al. Reference Fried, Katz, McCarthy, Sass, Williamson, Spencer and Spencer1991). In some cases, eliciting the actual motor movement could be achieved by increasing the electrical current applied to that same area. Subsequent studies have suggested that intention and agency may have precedents localizable to the pre-supplementary motor cortex (pre-SMA) (Lau et al. Reference Lau, Rogers, Haggard and Passingham2004; Moore et al. Reference Moore, Ruge, Wenke, Rothwell and Haggard2010) and that this region is the likely origin of Libet's RP (Haggard, Reference Haggard2008). In another recent experiment, electrical stimulation of sites within the inferior posterior parietal cortex produced urges to move specific body parts or to speak (Desmurget et al. Reference Desmurget, Reilly, Richard, Szathmari, Mottolese and Sirigu2009). When the intensity of that stimulation was increased, subjects reporting having actually moved, despite no EMG evidence of movement. Conversely, stimulation of the premotor cortex could produce actual movements, despite subjects firmly denying having moved. These findings suggest that distinct neuroanatomical regions govern motor intention, awareness and agency, and that these may be separable components of volition. Overall, a model of volitional action emerges in which preparatory activity occurring in the pre-SMA and SMA gives rise to an urge to move that is subsequently carried out into action by the primary motor cortex. In addition, a copy of the ‘command’ to move is also sent to the parietal cortex, which then generates a sensory representation of the predicted consequences of that movement (Haggard, Reference Haggard2009). In this way, volition in motor movement is better understood as set of processes in specific brain circuits that jointly specify information that determines our actions, such that voluntary action is best characterized as a form of neural decision making (Haggard, Reference Haggard2008).

Free will v. volition

The neuroscientific experiments presented thus far provide multiple lines of evidence that suggest that free will may be little more than a ‘naïve folk psychological intuition’ (Haynes, Reference Haynes2011). But doing away with free will does not mean that human beings are automatons in which our motivations, desires and values do not influence action. On the contrary, neuroscience clearly distinguishes between voluntary behavior and involuntary acts or reflexes. With volitional behavior, organisms have choices and make decisions about whether and when to act, what to do, and what not to do (Haggard, Reference Haggard2008). However, in a neuroscientific model, such choices are made within neural networks rather than any immaterial homunculus and often occur outside conscious awareness and before our subjective sense of intention or agency. Likewise, volition is conceptualized on a continuum, in which variable degrees of control are present across a spectrum of behaviors.

Disorders of self-control

Another approach to the question of free will in complex behaviors involves examining pathological states that represent compromised volition or self-control. Thinking about psychiatric illnesses as disorders of volition and free will could result in enhanced neuroscientific models of psychopathology as well as a broader understanding of volitional behavior across a spectrum from disorder to normality.

Addiction is a prime illustrative example. Historically, addictive behaviors have been thought of as deficits in will, suggesting that they represent free choices to be overcome by exercising greater willpower or the surrender of will to a higher power (Chappel, Reference Chappel1992; Committee on Addictions of the Group for the Advancement of Psychiatry, 2002; Gray, Reference Gray2007; Vohs & Baumeister, Reference Vohs and Baumeister2009). In contrast, the relatively novel disease model of addiction presents such behaviors as consequences of brain chemistry gone awry that result in an impairment of free will. Ironically, for many, the disease model can be difficult to accept because addiction requires the simple motor act of ingesting a psychoactive substance – an action that seems to be under control of free will (Vohs & Baumeister, Reference Vohs and Baumeister2009). And yet, the potential for free will to be illusory in simple motor acts under certain conditions has already been established.

An updated neuroscientific model of addiction abandons antiquated concepts of free will and willpower in favor of a continuum of volitional self-control and loss thereof (Lyvers, Reference Lyvers2000; Baler & Volkow, Reference Baler and Volkow2006). Cognitive control can be subdivided into quantifiable constructs including response inhibition (the ability to suppress or veto behavior) and impulsive choice (premature selections of behavior without adequately weighing consequences; Moeller, Reference Moeller, Barratt, Dougherty, Schmitz and Swann2001; Hyman, Reference Hyman2007; Perry & Carroll, Reference Perry and Carroll2008). In the laboratory, response inhibition is typically measured using the go/no-go or the stop signal tasks. In a simple go/no-go paradigm, subjects are shown visual cues on a computer screen that indicate whether to push a button (e.g. a green ‘go’ symbol means push). Those with poorer response inhibition will tend to make more frequent errors of commission, pressing the button during ‘no-go’ cues. Similarly, in a typical stop signal task, subjects press buttons in response to visual cues on a computer screen (e.g. left or right arrows), but are told not to press any buttons when they hear a beep regardless of visual cues. In this paradigm, performance is based on stop signal reaction time. Deficits in response inhibition as measured by such tasks have been demonstrated among those with current substance abuse (Perry & Carroll, Reference Perry and Carroll2008) as well as those with acute exposure to alcohol (Marczinski & Fillmore, Reference Marczinski and Fillmore2005; Easdon et al. Reference Easdon, Izenberg, Armilio, Yu and Alain2005; Perry & Carroll, Reference Perry and Carroll2008), cocaine (Fillmore & Rush, Reference Fillmore and Rush2002) and amphetamines (Fillmore et al. Reference Fillmore, Rush and Marczinski2003). Similar effects have been found among those with chronic psychostimulant dependence (Monterosso et al. Reference Monterosso, Aron, Cordova, Xu and London2005), including one study that demonstrated that greater inhibitory deficits were associated with greater lifetime drug exposure (Colzato et al. Reference Colzato, van den Wildenberg and Hommel2007). These findings indicate that drug use can result in impulsivity that may worsen with chronic dependence, supporting the idea that deficits in volitional control may be progressive during different phases (e.g. acquisition, escalation, abstinence, relapse) of addiction (Perry & Carroll, Reference Perry and Carroll2008).

A bidirectional relationship between impulsivity and addiction is supported by observations that inhibitory control deficits detected in childhood are associated with an increased risk of later substance use disorders (Ivanov et al. Reference Ivanov, Schulz, London and Newcorn2008). Such findings highlight the considerable vulnerability of adolescents to drug abuse associated with normal neurodevelopmental changes that mediate impulsivity and risk-taking behavior (Chambers et al. Reference Chambers, Taylor and Potenza2003). Note that neuroanatomical studies point to a major role of the pre-SMA and SMA in response inhibition (Chambers et al. Reference Chambers, Garavan and Bellgrove2009), the same areas that were found to be relevant to volition and agency in the experiments by Libet and others described earlier.

Impulsive choice is measured in the laboratory using the delayed-discounting paradigm in which subjects with greater impulsivity choose smaller immediate rewards over larger delayed rewards. Using this measure, subjects with current alcohol, cocaine and methamphetamine abuse have been found to make more impulsive choices than non-users, though such increased impulsivity was not seen with acute administration of those substances under experimental conditions (Perry & Carroll, Reference Perry and Carroll2008). The Iowa Gambling Task (IGT) was developed to measure behavioral choices associated with immediate reward despite significantly negative consequences observed among patients with damage to the ventromedial prefrontal cortex (VMPC). The task directs subjects to select cards from four decks with different preordained schedules of immediate and long-term reward and risk. Over the course of the task, normal subjects learn to select from the card decks that generate the most long-term reward, whereas patients with VMPC damage make choices associated with greater immediate reward, but longer-term loss (Bechara, Reference Bechara2005). Similar decision-making deficits have been demonstrated on the IGT among those with alcohol, cannabis, cocaine, methamphetamine and opiate dependence as well as drug-naïve subjects with a family history of substance abuse (Garcia-Verdejo, Reference Garcia-Verdejo2009). These findings, together with evidence of anatomical differences in the VMPC between substance abusers and normal controls, suggest an important role for the VMPC in the neuroanatomy of decision-making and addictive behavior. The ‘somatic marker theory of addiction’ proposes a neurocircuitry in which the amygdala mediates affective responses to actual substance use (or other immediate external rewards) while the VMPC triggers affective responses to thinking about or recalling previous episodes of drug use including associated negative consequences (Garcia-Verdejo, Reference Garcia-Verdejo2009). Drug choice in the face of negative consequences can result from exaggerating the incentive impact of actual drug use, thinking about drug use, or cravings as well as discounting anticipated negative consequences. Much of this neural decision making occurs unconsciously, with the insular cortex integrating the somatic signals from the amygdala and VMPC in order to produce a subjective feeling of a conscious urge to use drugs that is then implemented into action in the SMA (Naqvi & Bechara, Reference Naqvi and Bechara2008; Li et al. Reference Li, Lu, D'Argembeau, Ng and Bechara2010). Though it remains unproven (Dunn et al. Reference Dunn, Dalglieish and Lawrence2006), the somatic marker theory of addiction is supported by a considerable body of evidence and provides an appealing model for understanding the neurocircuitry of complex decision making, both pathological and normal.

Problems with response inhibition, impulsivity and reward–risk decision making have been detected in a wide variety of psychiatric disorders including bipolar disorder, attention deficit hyperactivity disorder, trichtillomania, obsessive–compulsive disorder, borderline personality disorder, conduct disorder, pathological gambling and antisocial personality disorder (Dunn et al. Reference Dunn, Dalglieish and Lawrence2006; Chamberlain & Sahakian, Reference Chamberlain and Sahakian2007; Perry & Carroll, Reference Perry and Carroll2008; Chambers et al. Reference Chambers, Garavan and Bellgrove2009; Garcia-Verdeko, Reference Garcia-Verdejo2009). This has led some to propose a new category of ‘volitional disorders’ for the Diagnostic and Statistical Manual, fifth edition (DSM-5) in which previously separate categories such as impulse control, obsessive–compulsive spectrum, and addictive disorders might be unified based on common underlying control deficits (Fontenelle et al. Reference Fontenelle, Mendlowicz and Versiani2009). Similarly, the National Institute of Mental Health has proposed cognitive control as a construct within its Research Domain Criteria that were formulated to study traits that are present among normal individuals and may be dysfunctional across a variety of different categorical DSM disorders (Insel et al. Reference Insel, Cuthbert, Garvey, Heinssen, Pine, Quinn, Sanislow and Wang2010; Sanislow et al. Reference Sanislow, Pine, Quinn, Kozak, Garvey, Heinssen, Wang and Cuthbert2010). In this way, volitional control represents a more accurate and quantifiable construct than free will that can be used to further elucidate neural mechanisms of control and understand the nature of control across a continuum of normal and pathological states.

Disorders of agency

Although our normal experience of free will may be misleading from a neuroscientific perspective, certain pathological conditions reveal just how disabling loss of agency – the subjective sense of casual ownership of behavior – can be. For example, patients with anarchic or alien hand syndrome experience a limb moving in a semi-purposeful way (e.g. unbuttoning a shirt) without any conscious intent to do so. Conversely, those exhibiting ‘utilization behavior’ experience a stimulus-driven tendency to use an object despite contextual purposelessness of the action (e.g. putting on a pair of glasses over a pair already being worn). This purposelessness suggests involuntariness, though patients rationalize their behavior as volitional. Both alien hand syndrome and utilization behavior are associated with frontal lobe injury, specifically involving lesions to the SMA and pre-SMA (Frith et al. Reference Frith, Blakemore and Wolpert2000; Blakemore et al. Reference Blakemore, Wolpert and Frith2002; Sumner & Husain, Reference Sumner and Husain2008). It is thought that these conditions reflect distinct lesions in a neural circuit that mediates the coordination of movement based upon continuous comparisons between the predicted sensory effects of a movement and its actual sensory effects (Frith et al. Reference Frith, Blakemore and Wolpert2000).

Among psychiatric disorders, certain symptoms of schizophrenia have also been modeled as disorders of agency. Sense of agency can be manipulated under experimental conditions in which normal subjects are tricked into thinking that motor movements seen on a computer screen are made by themselves when in fact they are made by others (Nielsen, Reference Nielsen1963; Wegner & Wheatley, Reference Wegner and Wheatley1999). These types of self-attribution agency errors have been found to occur significantly more frequently among patients with schizophrenia (Daprati et al. Reference Daprati, Franck, Georgieff, Proust, Pacherier, Dalery and Jeannerod1997) and prodromal psychosis (Hauser et al. Reference Hauser, Knoblich, Repp, Lautenschlager, Gallinat, Heinz and Voss2011) compared with controls, as well as among patients with schizophrenia and delusions of control compared with those without such delusions (Franck et al. Reference Franck, Farrer, Gerogieff, Marie-Cardine, Dalery and d'Amato2001). It has been proposed that such errors are rooted in deficits whereby the sensory feedback produced by self-movements does not differ, as it should, from feedback produced by the movements made by others. For example, two studies reported that patients with schizophrenia do not attenuate tactile sensations arising from self-stimulation as much as normal subjects, suggesting a possible mechanism for somatic passivity experiences (Blakemore et al. Reference Blakemore, Smith, Steel, Johnstone and Frith2000; Shergill et al. Reference Shergill, Samson, Bays, Frith and Wolpert2005). The predictive sensory effects of a volitional movement are thought to arise from an efference copy of that movement generated through a process called corollary discharge. When the predictive sensory effects of the efference copy do not match the actual sensory effects, the movement can become dissociated from agency and instead attributed to an external agent (Syznofzik et al. Reference Syznofzik, Vosgerau and Newen2008). Several studies support a model in which delusions of control arise from such mismatches caused by deficits involving internal predictions, external sensory feedback, or some difficulty with integration (Syznofzik et al. Reference Syznofzik, Their, Leube, Schlotterbeck and Linder2010; Voss et al. Reference Voss, Moore, Hauser, Gallinat, Heinz and Haggard2010). The application of this comparator model is not limited to motor acts – problems with corollary discharge may also explain thought insertion and hallucinations in which internally generated thoughts or speech are perceived as coming from an external source (Blakemore et al. Reference Blakemore, Smith, Steel, Johnstone and Frith2000; Ford & Mathalon, Reference Ford and Mathalon2005; Heinks-Maldonado et al. Reference Heinks-Maldonado, Mathalon, Houde, Gray, Faustman and Ford2007).

Additional studies of agency have focused on a normal phenomenon in which the subjectively perceived timing of a motor act (e.g. pressing a button) is delayed when followed by an effect (e.g. a tone). This intentional and temporal binding of voluntary action and effect is thought to contribute to a sense of agency. In several studies, patients with schizophrenia have demonstrated hyperbinding in which there is excessive linkage between volitional actions and external sensory events. Such studies have revealed that patients with schizophrenia experiencing delusions and hallucinations seem to have particular difficulty with internal predictions about the sensory consequences of their actions, perhaps resulting in overreliance on external cues leading to errors in agency judgments (Syznofzik et al. Reference Syznofzik, Their, Leube, Schlotterbeck and Linder2010; Voss et al. Reference Voss, Moore, Hauser, Gallinat, Heinz and Haggard2010). Modeling psychotic symptoms as disintegrative dysfunction resulting in loss of agency and the ability to distinguish self from others has intuitive appeal given the longstanding notion of schizophrenia as a schism of the mind in which core aspects of self-identity are disrupted (Jardri et al. Reference Jardri, Pins, Lafargue, Very, Ameller, Delamire and Thomas2011; Syznofzik et al. Reference Syznofzik, Vosgerau and Voss2013).