Free Will and Neuroscience: From Explaining Freedom Away to New Ways of Operationalizing and Measuring It

Introduction—Free Will as a Problem (Not Only) for Science

The concept of free will is hard to define, but crucial to both individual and social life (Kane, 2005). Free will can be the reason why someone is not sent to jail during a trial upon appealing to insanity: the subject was not “free” when they committed the crime, not because someone was pointing a gun to their head, but because a psychiatric illness prevented them from controlling their actions. According to a long-standing philosophical tradition, if someone was not “free” when they did something, they cannot be held responsible for their deed (Glannon, 2015). And the freedom in question is both “social” freedom (linked to constraints imposed by our peers or by external factors), and the one indicated by the term free will.

Free will can be defined by three conditions (Walter, 2001). The first one is the “ability to do otherwise.” This is an intuitive concept: to be free, one has to have at least two alternatives or courses of action between which to choose. If one has an involuntary spasm of the mouth, for example, one is not in the position to choose whether to twist one’s mouth or not. The second condition is the “control over one’s choices.” The person who acts must be the same who decides what to do. To be granted free will, one must be the author of one’s choices, without the interference of people and of mechanisms outside of one’s reach. This is what we call agency, that is, being and feeling like the “owner” of one’s decisions and actions. The third condition is the “responsiveness to reasons”: a decision can’t be free if it is the effect of a random choice, but it must be rationally motivated. If I roll a dice to decide whom to marry, my choice cannot be said to be free, even though I will freely choose to say “I do”. On the contrary, if I choose to marry a specific person for their ideas and my deep love for them, then my decision will be free.

Thus defined, free will is a kind of freedom that we are willing to attribute to all human beings as a default condition. Of course there are exceptions: people suffering from mental illness and people under psychotropic substances (Levy, 2013). Nevertheless, the attribution of free will as a general trend does not imply that all decisions are always taken in full freedom, as outlined by the three conditions illustrated above: “We often act on impulse, against our interests, without being fully aware of what we are doing. But this does not imply that we are not potentially able to act freely. Ethics and law have incorporated these notions, adopting the belief that usually people are free to act or not to act in a certain way and that, as a result, they are responsible for what they do, with the exceptions mentioned above” (Lavazza and Inglese, 2015).

It is commonly experienced that the conditions of “ability to do otherwise”, “control” and “responsiveness to reasons” are very rarely at work all at once. Moreover, they would require further discussion, because there is wide disagreement on those conditions as regards their definition and scope (Kane, 2016). But for the purposes of this article, this introductory treatment should suffice. In fact, the description of free will that I have sketched here is the one that dominated the theoretical discourse on, and practical applications of, the evaluation of human actions. From a philosophical point of view, however, starting with Plato, the main problem has been that of the actual existence of freedom, beyond the appearances and the insights that guide our daily life. The main challenge to free will has been determinism: the view that everything that happens (human decisions and actions included) is the consequence of sufficient conditions for its occurrence (Berofsky, 2011). More specifically, “It is the argument that all mental phenomena and actions are also, directly or indirectly, causally produced—according to the laws of nature (such as those of physics and neurobiology)—by previous events that lie beyond the control of the agents” (Lavazza and Inglese, 2015). Determinism was first a philosophical position; then, the birth of Galilean science—founded on the existence of immutable laws that are empirically verifiable—has increased its strength, giving rise to the concept of incompatibilism, namely the idea that free will and natural determinism cannot coexist. Only one of them can be true.

Throughout the centuries, despite its conceptual progress, philosophy hasn’t been able to solve this dilemma. As a result, today there are different irreconcilable positions about human free will: determinism is not absolute and free will exists; free will does not exist for a number of reasons, first of all (but not only) determinism; free will can exist even if determinism is true (Kane, 2011). A little more than 30 years ago, neuroscience and empirical psychology came into play. Although biological processes cannot be considered strictly deterministic on the observable level of brain functioning (nerve signal transmission), new methods of investigation of the brain, more and more precise, have established that the cerebral base is a necessary condition of behavior and even of mental phenomena. On the basis of these acquisitions, neuroscience has begun to provide experimental contributions to the debate on free will.

In order to better understand the neural bases of free will, provided that there are any, in this article I’ll review and integrate findings from studies in different fields (philosophy, cognitive neuroscience, experimental and clinical psychology, neuropsychology). Unlike previous reviews on free will and neuroscience (Haggard, 2008, 2009; Passingham et al., 2010; Roskies, 2010a; Brass et al., 2013), I have no claim of being exhaustive. My goal is to highlight a paradigm shift in the analysis and interpretation of the brain determinants preceding and/or causing free or voluntary action (Haggard, 2008 takes voluntary decision to be non-stimulus driven, as much as possible). Firstly, following Libet’s experiments, a widespread interpretation of the so-called readiness potential (RP) went in the direction of a deflation of freedom (Crick, 1994; Greene and Cohen, 2004; Cashmore, 2010; Harris, 2012). Indeed, the discovery of the role of the RP has been taken as evidence of the fact that free will is an illusion, since it seems that specific brain areas activate before we are aware of the onset of the movement. However, recent studies seem to point to a different interpretation of the RP, namely that the apparent build-up of the brain activity preceding subjectively spontaneous voluntary movements (SVM) may reflect the ebb and flow of the background neuronal noise, which is triggered by many factors (Schurger et al., 2016). This interpretation seems to bridge, at least partially, the gap between the neuroscientific perspective on free will and the intuitive, commonsensical view of it (Roskies, 2010b), but many problems remain to be solved and other theoretical paths can be hypothesized. After analyzing the change of paradigm of these perspectives, I’ll propose to start from an operationalizable concept of free will (Lavazza and Inglese, 2015) to find a connection between higher order descriptions (useful for practical life) and neural bases.

Neuroscience: Purporting to Explain Free Will

Free Will as an Illusion

All these experiments seem to indicate that free will is an illusion. Yet, these relevant experiments can be interpreted in many ways. A possible view is that, in some way, determinism can be observed directly within ourselves. This interpretation might lead to the conclusion that free will is just an illusion. In fact, if one considers as a condition of free will the fact that it should be causa sui (i.e., it should be able to consciously start new causal chains), such a condition is incompatible with determinism as it is usually defined. For it, in fact, all events are linked by casual relations in the form of natural laws, which started long before we were born and which we cannot escape.

However, determinism has generally been regarded as a metaphysical claim, not refutable by empirical findings. One could properly talk of automatism in the brain, not of determinism, based on the evidence available. (In any case, endorsing indeterminism might lead to consider our behavior as the causal product of choices that every time produce different results, as if we rolled a dice. This doesn’t seem to make us any freer than if determinism were overturned; cf. Levy, 2011). Most importantly, another feature of freedom seems to be a pure illusion, namely the role of consciousness. The experiments considered thus far heavily question the claim that consciousness actually causes voluntary behavior. Neural activation starts the decisional process culminating in the movement, while consciousness “comes after”, when “things are done”. Therefore, consciousness cannot trigger our voluntary decisions. But the role of consciousness in voluntary choices is part of the definition of free will (but the very definition of consciousness is a matter of debate, cf. Chalmers, 1996).

Empirical research in psychology also shows that our mind works and makes choices without our conscious control. As proposed by psychologist Wegner (2002, 2003, 2004) and Aarts et al. (2004), we are “built” to have the impression to consciously control our actions or to have the power to freely choose, even though all that is only a cognitive illusion. Many priming experiments show that people act “mechanically” (even when their behavior might appear suited to the environment and even refined). Automatic cognitive processes, of which we aren’t always aware, originate our decisions, and they were only discovered thanks to the most advanced scientific research. Ultimately, consciousness, which should exercise control and assess the reasons for a choice, is thus allegedly causally ineffective: a mere epiphenomenon, to use the terminology of the philosophy of mind. This is what has been called Zombie Challenge, “based on an amazing wealth of findings in recent cognitive science that demonstrate the surprising ways in which our everyday behavior is controlled by automatic processes that unfold in the complete absence of consciousness” (Vierkant et al., 2013).

These experiments have triggered a huge debate and led scientists, philosophers and intellectuals to claim (or insist even more, if they already denied free will) that free will doesn’t exist (Greene and Cohen, 2004; Cashmore, 2010; Harris, 2012). It seemed as though neuroscience had produced empirical evidence against free will, so that the century-long debate on it could be considered solved. However, Libet’s experiments have been also criticized. Much criticism was directed to the philosophical interpretation of these studies (Mele, 2014) or to their theoretical assumptions (Nachev and Hacker, 2014), which are important but not relevant here. Among the forms of criticism, one has to mention the theories of action that separate the deciding from the initiating (Gollwitzer, 1999; Gollwitzer and Sheeran, 2006). In that case, free and conscious deliberating could still have a relevant casual role, long before the actual performance of the action.

Other objections, more markedly neuroscientific, were made for instance by Trevena and Miller (2010). They argued that the RP is not an intention to move, but only indicates that an attentional process is in place in the brain, since when subjects “attended to their intention rather than their movement, there was an enhancement of activity in the pre-SMA” (Lau et al., 2004). In any case, “there was no evidence of stronger electrophysiological signs before a decision to move than before a decision not to move, so these signs clearly are not specific to movement preparation”, (Trevena and Miller, 2010). Others have noted that the introspective estimates of event timing are disputable or inaccurate, and measures in general are not sufficiently exact (Dennett, 1984a,b, 2003).

More than Explaining Away

Other studies using multivariate pattern analysis with EEG confirmed that the subjectively free decisions might be made in the brain in the same way as evidence-based perceptual decisions (Bode et al., 2012a,b, 2013). Indeed, Bode et al. (2012b) wrote,

we directly decoded choice-predictive information from neural activity before stimulus presentation on pure noise trials on which no discriminative information was present. Choice behavior on these trials was shown to be primed by the recent choice history. Modelling of sequential effects in RT and accuracy confirmed that such choice priming biased the starting point of a diffusion process toward a decision boundary, as conceptualized in evidence accumulation models of perceptual decision making (Bode et al., 2012b).

In other words, the authors found that internally (and maybe stochastically) generated neural activity can bias decisions that are expected to be stimulus-responsive or (possibly) reason-responsive. In this case, as in others that I will consider below, the understanding that we begin to have of the neuronal processes in play shows us that there is a complexity of factors at work. Some of these factors seem to be genuinely random, due to the pure noise produced by the default brain activity, while other factors can be traced back to the previous history of decisions taken in similar situations or related to the present one. Therefore, there is no “mysterious” start of the action as a linear process that, from the initial command, is then executed, as in Libet’s simplified model. Rather, this outcome is the result of a multiplicity of causal elements, which are homogeneous from the viewpoint of proximal mechanisms but of different relevance from the viewpoint of interpretation in terms of intentional psychology.

Another study has shown that attempts to account for (make sense of) insufficient perceptive clues use the same neural networks as those involved in “free” decision-making (Bode et al., 2013). An fMRI-based pattern classifier can be trained to differentiate between different perceptual guesses and try to predict the outcome of non-perceptual decisions, like those made by the participants in the experiments considered so far. Specific activation patterns detected in the medial posterior parietal cortex have allowed the authors to make correct predictions on the participants’ free choices based on the previously decoded perceptual guesses decoded, and the other way round.

The task was the following: the participants were given a masked stimulus and had to say what category the stimulus belonged to. They had to freely choose among many categories. Thanks to the multivariate pattern analysis it was possible to identify the model of “free decisions” to make correct predictions in the context of perceptual judgments and identify the model of the “guess decisions”, to make correct predictions in the context of “free decisions”. It thus seems that a similar neural code for both types of decision is present. In those cases one could say that guessing is similar to making a free decision, since the brain, in the absence of sufficient external cues, has to decide internally. So perceptual decisions can be predicted from specific preceding neural activity when the brain doesn’t have enough internal elements to reach the threshold of perceptive decision.

Studies and commentaries have nevertheless drawn attention to possible confounds and bias in those experiments, namely they might be affected by previous choices with a form of auto-correlation in spontaneous decisions. In particular, Lages and Jaworska (2012) “trained a linear classifier to predict “spontaneous decisions” and “hidden intentions” from responses in preceding trials and achieved comparable prediction accuracies as reported for multivariate pattern classification based on voxel activities in frontopolar cortex”. Lages et al. (2013) have stressed a possible sequential information processing between trials that can introduce a confound, and recommended that “rather postulating a 50% chance level, prediction should be tested with a permutation test and/or separate multivariate classification analyses conditional on the previous response”.

The prediction of perceptual decisions from specific preceding neural activity is linked to what is defined “evidence accumulator model for free choice” (Bode et al., 2014). The explanation starts with the fact that predictive activation patterns preceding decisions become increasingly similar to the patterns detected when the decision is consciously experienced by the subject. This could mean that a slow build-up of decision-related activity occurs, as it happens in accumulation of decision-related evidence to a decision threshold (Ratcliff, 1978; Ratcliff and McKoon, 2008). Also, as already noted, when no external feedback is available, the previous choice is used as external feedback (Akaishi et al., 2014). The history of previous decisions has a systematic effect on subsequent choices, related to the activity in medial posterior parietal cortex/posterior and posterior cingulate cortex (Bode et al., 2011, 2013). And the systematic effect can go in the direction of repetition or of avoidance of repetition depending on the task (Mochizuki and Funahashi, 2014).

Here is an important point that deserves study from the neuroscientific point of view but also from that of a philosophical interpretation of free will. It consists in the fact that the internally generated brain activity has to do both with the stochastic noise and with the history of the subject’s choices. On the one hand, the stochastic noise comes both from the configuration that the brain has on average as a result of evolution (adaptive significance) and from individual development, resulting from random processes and environmental influences. On the other hand, the history of the choices is derived from the same process (in part stochastic) that I have just described.

In any case, if (at least some) very short-term decisions have a genesis similar to that described here, these decisions contribute to shaping the brain activity, and then, presumably, also to influencing decisions on a longer time scale that it is not yet possible to investigate experimentally. Ultimately, this could mean that there is a confluence of causal factors at the level of microdecisions. These factors add up in a way that it is hardly possible to tackle for current science. Then also the reasons motivating an action, typical of free actions, such as “I punched the stalker because it is right to punish those who behave in this way and because I wanted to set an example for all”, encoded in neural activity, can be part of the sum of neural causes.

In fact, experimental psychology has been trying to take into account long-term influences. In the so-called marshmallow experiment, researchers focused on delayed gratification (Mischel et al., 1972, 1989). A child was given a choice between one small, immediate reward and two small rewards (i. e. a larger reward) if they were able to wait some minutes while the psychologist left the room and then came back. Children who waited longer for the their rewards tended to have better life outcomes and accomplishments. Such experiments are relevant in terms of explanations and predictions, but it seems hard to trace behavioral profiles back to specific profiles of cerebral activation, once we are aware of the complexity of causal chains in the evidence accumulation model.

As Bode et al. (2014) write, in the hypothesis of an evidence accumulator that collects sensory evidence until a decision threshold is reached,

task instructions, participants’ internal motivation, and previous choices all have a strong influence on how decision tasks are performed when external information is either unavailable (as in free decisions), or unhelpful (as in perceptual guessing). In the case of free decision tasks, fluctuating intention for one or the other option may result from active competition between neural representations of both options in decision networks (or rather although not consciously monitored by the participants, the previous choice history, embodied in dynamic states of decision networks, can become the primary determinant of behavior, simply because nothing else is available (Bode et al., 2014).

However, in this way things get more complicated and at a macroscopic level of behavioral observation, this blurs but doesn’t do away with the idea of free behaviors and behaviors that could be taken as unconscious decisions, of which we become aware only when the action has been performed. What remains to be solved is the problem of the distinction between external stimuli that trigger a stimulus-response circuit, and internal self-paced intentions and decisions that trigger voluntary circuit (Haggard, 2008).

Other Neuroscientific Hypotheses on Free Will

Operazionalizing, Measuring and Verifying: From the Action to the Brain

My view is that a richer conceptualization of free will—one that is able to overcome the stall of the metaphysical debate as well as the current difficulties of neuroscience (Nachev and Hacker, 2014) and empirical psychology (Nahmias, 2014)–has to be linked to the idea of “capacity”. In fact, as claimed by Mecacci and Haselager (2015), the kind of free will investigated by neuroscientific experiments, which is self-generated and defined according to the absence of cues, “does little justice to the common sense practice of holding people responsible for their freely willed actions that consists in asking explanations and justifications from the actor” (Mecacci and Haselager, 2015).

Another important point is that there are differences in time scales between laboratory tasks (the milliseconds to seconds time range) and real life or, better, life as we measure it temporally (seconds, minutes, hours, weeks, years) regarding decisions that really concern us. Even if the underlying mechanism might be the same, the experiments described so far cannot investigate whether decisions with a longer maturation process are free and to what extent they are such. It might be possible to distinguish between proximal and distal mechanisms, but this doesn’t seem feasible lacking the tools to address decisions involving longer time scales. For this reason it might be useful to introduce other and different ways to conceptualize and operationalize (supposedly) free actions.

“By capacity, in the context of free will, we mean the availability of a repertoire of general skills that can be manifested and used without the moment by moment conscious control that is required by the second condition of free will we have previously seen” (Lavazza and Inglese, 2015). The concept of capacity is related to that of internal control, understood as the agent’s “ownership” of the mechanism that triggers the relevant behavior and the reasons-responsiveness of that mechanism (Fischer and Ravizza, 2000). And reasons-responsiveness must involve a coherent pattern of reasons-recognition. “More specifically, it must involve a pattern of actual and hypothetical recognition of reasons that is understandable by some appropriate external observer. And the pattern must be at least minimally grounded in reality” (Lavazza and Inglese, 2015). The concept of capacity used in this sense, and combined with the idea of reasons-responsiveness, also avoids the objection of determinism that has always weighed on the debate on free will. From a philosophical point of view, the approach related to capacity may fall indeed in the strand of so-called compatibilism, which defends the fact that human freedom can exist even if determinism is true of the physical world.

Cognitive abilities might be firstly operationalized as a set of neuropsychological tests, which can be used to operationalize and measure specific executive functions, as they are strongly linked to the concept of control. Executive functions, also known as control functions, are essential to organize and plan everyday behavior—which is not the instant behavior found in Libet’s experiments. Those skills are necessary to perform the greater part of our goal-oriented actions. They allow us to modulate our behavior, control its development and change it according to the environmental stimuli (the environment being both physical and social). Also, executive functions allow us to change our behavior based on it effects, with a sophisticated feedback mechanism; finally, they are also necessary for tasks of abstraction, inventiveness and judgment. Those who, for whatever reason, have a deficit in their executive functions cannot respond to their social environment appropriately, and struggle to plan their behavior or to choose between alternatives based on their judgment or interest. Sufferers of these deficits in executive functions often fail to control their instinctive responses and to modify their regular courses of action, or are unable to concentrate or persist in the pursuit of a goal (Barkley, 2012; Goldstein and Naglieri, 2014).

In general terms, the executive functions refer to the set of mental processes necessary for the development of cognitive-behavioral patterns adaptive in response to new and demanding environmental conditions. The domain of executive functions includes (Lavazza and Inglese, 2015):

• the ability of planning and evaluation of effective strategies in relation to a specific purpose related to the skills of problem-solving and cognitive flexibility.

• inhibitory control and decision-making processes that support the selection of functional response and the modification of the response (behavior) in relation to changing environmental contingencies.

• attentional control referred to the ability to inhibit interfering stimuli and to activate the relevant information.

• working memory referring to the cognitive mechanisms that can maintain online and manipulate information necessary to perform complex cognitive tasks.

• (and it can be added with regards to free will) creativity and the ability to cope with environmental changes through novel solutions.

Those of empirical psychology are higher order concepts, which act as a bridge between free will, which is something that is not in the brain but can be observed in behavior (along with its causes), and the underlying brain processes. It has been convincingly suggested that in the construction of a hierarchy of mechanisms and explanations (Craver, 2007), also to guide the exploration, one must go from inside to outside and from outside to inside. One goes from measurable skills to their brain basis, and from the tentative index of free will to the underlying (real) mechanisms.

Based on the evidence presented, I believe that a viable proposal is to construct an index related to compatible tests whose relevance can be uniformly ascertained. It would be a kind of IQ-like profile that would allow for the operationalization and quantification of a person’s cognitive skills. All the tests used (for example, Stroop Test, Wisconsin Card Sorting Test, Weigl’s Color-Form Sorting Test, Go-No Go Test) should be related to the subject’s age and education and then transformed in new standardized scores (Equivalent Scores, ES) on an ordinal scale, e. g. ranging from 0 to 4, with 0 representing scores below cut-off point and 4 representing scores equal to or better than average. Specific standardized scores exist in many countries or linguistic areas. The subjects would get for each test a raw score (or RS), given by the sum of the scores obtained in each item that makes up the test, which would then be standardized.

A synthetic index such as the one here proposed measures a certain range of cognitive and behavioral control skills that configure a certain kind of free will at the psychological-functional level. These are potential capacities measured with standardized instruments in laboratory situations, which do not consider any other factors that may restrict the freedom of a subject in specific situations, such as those that are relevant in moral scenarios and legal contexts. The same goes for moral judgment. However, an index such as the one I’m proposing here could be the first step, albeit certainly imperfect, towards more objective measures to discriminate between people who have more or less “free will” or, in other words, are more or less capable of self-control and rational choice (i.e., a reasons-responsive choice).

This hypothesis would be in line with the proposals of operationalizing free will advanced so far. According to Vohs (2010), freedom might be conceived of as the sum of executive functions and goal-directed, future-oriented behaviors, which include rational choice, planning, intelligent thought, and self-control. Free will can be then constituted by a limited stock of energy, devoted to guiding executive functioning processes. The free will index I am proposing is also consistent with Baumeister’s contribution:

Psychologists should focus on what we do best: collecting evidence about measurable variance in behaviors and inner processes and identifying consistent patterns in them. With free will, it seems most productive for psychologists to start with the well-documented observation that some acts are freer than others. As already noted, dissonance, reactance, coping with stress, and other behaviors have been shown in the laboratory to depend on variations in freedom and choice. Hence, it is only necessary to assume that there are genuine phenomena behind those subjective and objective differences in freedom. In a nutshell, we should explain what happens differently between free and unfree actions (Baumeister, 2008).

Empirical research on how human beings work has recently focused on self-control as a feature of free will. Self-control can be defined as the exertion of willpower on behavior. Self-control is thus generally regarded as the capacity to override inappropriate impulses and automatic or habitual responses and to suppress or delay immediate gratification so as to reach a long-term goal (Gailliot and Baumeister, 2007). “Being in control” includes the capacity to maintain goals, to balance long- and short-term values, to consider and evaluate the consequences of a planned action, and to resist being “carried away by emotion” (Churchland, 2006). Self-control can also be regarded as the ability of higher-order functions to modulate the activity of lower-level functions, where higher-order functions manifest themselves externally in complex behavior, adjusted according to the environmental needs, while lower-level functions are manifested in simple and stereotyped behaviors, not adjusted according to the demands of the environment (Roskies, 2010a). Everyone exhibits a different degree of self-control compared to other individuals, and for each person the degree of self-control varies over time (Baumeister et al., 2006; Casey et al., 2011; Dang et al., 2015). The variability of self-control that is manifested in behavior and can be measured with the test has its base in neuronal functioning, which in turn depends on education and habits, external circumstances and the internal neuronal noise.

However, two executive functions turn out to be central:

(i) the ability to predict the future outcomes of a given action; and (ii) the ability to suppress inappropriate, i.e., not sufficiently valuable, actions. Importantly, these two executive functions operate not only during the genesis of an action, but also during the planning of an already selected action. In fact, during the temporal gap between the time when an action has been chosen and the moment when the motor output is going to be generated, the context might have changed, altering the computed value of the action and thus requiring a radical change of the planned motor strategy (Mirabella, 2014).

It seems that the peculiarity of our freedom at the cognitive level is the ability to modulate or block courses of action that environmental stimuli automatically or unconsciously arouse in us—a reproposal in different form of Libet’s free won’t and Schultze-Kraft’s vetoing. These psychological-functional indicators must then lead to their cerebral bases. For instance, one can consider a situation in which one’s needs are satisfied (or not) and the consequent motivation to act based on the evaluation process of the need satisfaction.

This is an essential process and one that is continuously performed by our motor system. In fact, in most places where we live, if not all, we are surrounded by tools whose sight automatically activates motor schemas that would normally be employed to interact with those objects. These actions are prompted by the features of the objects, the so-called affordances (Gibson, 1979). It has been shown that even the simple observation of pictures depicting affordable objects (such as graspable objects) activates a sub-region of the medial frontal cortex, the SMA, even when there is no requirement to actually act on those stimuli (Grèzes and Decety, 2002). These stimulus-driven activations are rapid, involuntary, and unconscious (Mirabella, 2014).

Environmental stimuli, in this case, can induce a subject to make specific choices through a priming process that exploits our action tendencies. Typically, individuals are able to control their behavior, but in some cases they fail to do so; for example if suffering from microlesions of the SMA, people have a tendency to invariably implement a certain type of action, even if the environment, both physical and social, does not require it (Sumner et al., 2007). In fact, “the suppression of a triggered action might be seen not as an active process, but rather as an automatic consequence of the evaluative procedure” (Mirabella, 2014). One could then say that those who have the ability to better monitor, control and direct their own behavior are “freerer” than those who do not have this capability. Individuals affected by disorders of the executive functions are not able to grasp and process environmental stimuli to direct their behavior. For example, these people may not be able to stop the utilization behavior, an automatic mechanism that tends to make us interact with all the objects that are in our perceptual sphere.

Churchland (2006) and Suhler and Churchland (2009) proposed a hypothesis concerning the neural basis for control, which can bridge the gap between higher-order concepts and brain mechanisms. As she wrote,

Perhaps we can identify various parameters of the normal profile of being in control, which would include specific connectivity patterns between amygdala, orbitofrontal cortex, and insula, between anterior cingulate gyrus and prefrontal cortex, and so forth. Other parameters would identify, for each of the six non specific systems [identified via the neurotransmitter secreted at the axon terminals: serotonin, dopamine, norepinephrine, epinephrine, histamine and acetylcholine], the normal distribution of axon terminals and the normal patterns of neurotransmitter release, uptake, and co-localization with other neurotransmitters such as glutamate. Levels of various hormones would specify another set of parameters. Yet other parameters contrast the immature with adult pattern of synaptic density and axon myelinization. At the current stage of neuroscience, we can identify the normal range for these parameters only roughly, not precisely (Churchland, 2006).

This hypothesis would allow for specific brain correlates of a free will index based on the executive functions-guided self-control and even, hypothetically, a direct brain measure of being in control For example, a recent study (Bartelle et al., 2016) highlights the possibility of having MRI imaging of dopamine release thanks to a engineered protein that binds to the neurotransmitter and works as a MRI-visible probe. As the authors put it, “one could imagine a future in which molecular fMRI is used to determine brain-wide neurochemicals maps corresponding to a universe of stimuli and behavioral programs”. Even though one should always consider that there isn’t perfect correspondence between higher-order concepts and putative neural correlates.

In particular, one must consider that what matters in interpersonal relations and in law, to give two examples of practical relevance of free will, is freedom as actually performed: that is, freedom as it can be observed and with some approximation, also measured through a series of psychological tests. This does not mean that the same level of freedom manifested in behavior matches the same level of activation of the related brain areas. However, one can investigate the brain causes of “freedom deficit” compared with the average shown by relevant samples of the population, and so come to a progressive refinement of the research on the neural bases of free will.

Another example is given by the investigation of the role of the cholinergic interneurons in behavioral flexibility (Aoki et al., 2015). This class of neurons seem to be connected to survive in an ever-changing world, which requires behaving flexibly. Flexibility can be assessed (and measured) at a behavioral level, but cerebral mechanisms remain largely unknown. Using conventional tests on behavioral flexibility, which require animals to shift their attention from one stimulus property (e.g., color) to another (e.g., shape), researchers probed the effects of an immunotoxin-induced lesion of cholinergic interneurons in the striatum.

A selective cholinergic ablation was made by means of injections of immunotoxin, which targeted neurons containing choline acetyltransferase in the dorsomedial or ventral striatum. A control group was instead injected with saline. “When encountering a change of behavioral rules after the set-shift, either lesion made animals stick to a previously correct but now invalid response strategy. They also showed less exploratory behavior toward finding a new rule. Most interestingly, ablation of cholinergic neurons in the dorsomedial striatum impaired a shift of set when it required attention to a previously irrelevant cue. On the other hand, ventral cholinergic lesions had an effect on a shift in which a novel stimulus was introduced as a new directional cue” (Aoki et al., 2015). Animals thus can be taken to be “less free” when striatal cholinergic interneurons don’t work properly.

This last example serves to indicate how to bridge the gap between overt behaviors (to which we tend to attribute the property of freedom) with neuronal mechanisms that are clearly identifiable and even manipulable. In fact, it is not so important to look at the conscious aspect of a single proximal mechanism, but rather to consider the manifest behavioral effect that the considered mechanism helps to produce. This way there would be a paradigm shift with respect to the neuroscience research on free will, which seems to have long been too closely linked to the falsification of the theoretical assumption that an action is free only if it has a beginning that is fully controlled by a conscious process. The proposal, I am making here has only the ambition to be a potentially helpful contribution to theoretical debate and empirical research, although its limits are very clear. First, it focuses on a specific part of what is intuitively called “free will”, relating it to the idea of “capacity”. Second, it proposes to measure free will at a psychological level by means of a unitary index that inevitably misses many nuances of the notion and the relative capacity. Furthermore, the search for the neural correlated of such capacities implies not only the identification of causal mechanisms, but also the consideration of many cerebral areas. All of this makes things harder compared to approaches à la Libet. Nevertheless, there is manifest advantage: there is a greater degree of realism and adherence to the actual behavioral manifestation of what we call “free will”.

Conclusion

Free will is an elusive but crucial concept. For many years we have known that the functioning of our brain has to do not only with the belief that we have free will but also with the existence of free will itself. Evidence of the unconscious start of movement, highlighted by the RP signal, has led to believe that we had reached an experimental proof of the non-existence of free will—which many already claimed at a theoretical level based on the argument of the incompatibility between determinism and freedom. Along with other evidence provided by experimental psychology, the branch of studies inaugurated by Libet has contributed to seeing free will as an illusion: this view seemed to be reliably supported by science, and in particular by neuroscience. Recent studies, however, seem to question this paradigm, which sees the initiation and conscious control of the action as the first requirement of free will, allegedly proving that there are no such things.

The stochastic models and the models of evidence accumulation consider decision as the crossing of a threshold of activity in specific brain regions. They do not restore the idea of conscious control but turn away from the previous paradigm. These studies cannot yet fully explain how the intention to perform an action arises in the brain, but they better account for the complexity of the process. In particular, they recognize the role of the spontaneous activity of the brain, of external cues and other factors—including those that might be called “will” and “reasons” (which, however, do not currently have precisely identified neural correlates)—in reaching the critical threshold. Studies that show how we can consciously block movements whose preparation has already begun unconsciously, then, indicate how the subject is able to exercise a form of control, whose genesis however is still unclear.

One could state that free “decision-making draws upon a rich history of accumulated information, manifested in preferences, attitudes and motivations, and is related to the current internal and external environment in which we act. Complete absence of context is impossible” (Bode et al., 2014). In this framework, I have here proposed to integrate neuroscientific research on free will by connecting higher-level concepts with their neural correlates through a psychological operationalization in terms of skills and cognitive functions that do not necessarily imply a continuous conscious control over the decision-making and action process. This may also allow one to create a quantitative index, albeit still quite rudimentary, of the degree of freedom of each subject. This freedom would be specifically defined and therefore may not perfectly coincide with the intuitive concept of free will. Starting from these functional indicators, which psychology has well clarified, one could then move on to investigate the precise neural correlates for a different and (possibly) more fundamental level of explanation in terms of brain processes that enable the executive functions.

According to Craver (2007), a mechanistic explanation is able to lead to an inter-field integration. There are two relevant aspects to this approach. The functional knowledge that can be drawn from psychological research is a tool to identify neural mechanisms; the knowledge of the brain structure can guide the construction of far more sophisticated psychological models (Bechtel and Mundale, 1999). The index of free will that I am proposing (Lavazza and Inglese, 2015)—despite surely needing further refinement—might be useful to explore the brain mechanisms that underlie what appears in behavior as “free will”, which no longer seems to be an illusion, not even for neuroscientific research.

Author Contributions

AL confirms being the sole contributor of this work and approved it for publication.

Conflict of Interest Statement

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.