Show Summary Details

Page of

PRINTED FROM OXFORD HANDBOOKS ONLINE ( © Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

date: 05 April 2020

Free Will: A Neurophilosophical Viewpoint

Abstract and Keywords

Determinism—the idea that we have no power to do otherwise than we do— denies free will; therefore having the capacity to do otherwise means that determinism does not hold. However, the concept of responsibility creates a motivation for making determinism and free will compatible—for arguing, in an “incompatibilist” vein, that the human brain is not subject to deterministic laws. Libertarianism—the view that human agency allows exceptions to occur in mind-brain activity—is untenable in the light of current evidence about human cognition. Compatibilism holds that an agent with no power to do otherwise can still enjoy free will. Frankfurt’s theory of a layered will is discussed, as is neuroscientific evidence in favor of Frankfurt’s model. It allows for a form of compatibilism in which an agent may be somewhat in control of what she does, depending on her environment, and thereby have a will that is somewhat free.

Keywords: compatibilism, incompatibilism, responsibility, libertarianism, wanting, liking, intention, Frankfurt

Over centuries of controversies, the issue of free will has been tailored to the needs of theology, traditional philosophy, folk psychology, ethics, and legal systems. Contemporary debates carry some of these burdens, although—under the influence of the scientific picture of the world—a predominant view now is that all existing processes and phenomena are causally determined; therefore, given the properties of the biological and physical systems considered, they cannot be different from what they are. In particular, determinism entails that the laws of nature are not up to us, nor is what went on before we were born; therefore our present acts are not up to us either. Determinism thus entails that there is no free will and that we have no power to do otherwise; reciprocally, having the capacity to do otherwise entails that determinism does not hold when one exercises this capacity. However, the matter of responsibility makes it difficult to accept that agents cannot do otherwise and creates a strong motivation for making determinism and free will compatible, or for arguing, in an “incompatibilist” vein, that the human brain is not subject to deterministic laws.

On the Power to Do Otherwise

The expression “one could have done otherwise” is tricky and deserves a brief discussion. In a common acceptation, it describes a bird’s-eye view of objectively possible alternative decisions, abstracting away from the specific agent’s beliefs and preferences and the decision context. This ideally complete representation of how actions could have been carried out is fueled by retrospective counterfactual reasonings, such as “Had I known that P, I would not have done A.” Trying to represent the range of possible actions plays a regulative and motivating role in deliberation: Agents are able to realize that their efficiency as agents depends in part on selecting one of the best possible specific goals or subgoals available in the objective action space at any point in space-time. This notion, however, does not directly apply to the issue of freedom of the will: Even if the abstract notion of an action space regulates deliberation, everyone in this debate agrees that agents are not omniscient and thus cannot have a full view of their action space: they cannot know all there is to know that would help them decide what is best to do.

In the context of the issue of free will, “could have done otherwise” is often taken to mean, rather, that the agent has the power to do what she chooses to do if she wants to do it. Such “hypothetical liberty” (David Hume, 1955, p. 104) is only denied to those who are prevented from acting as they want to or coerced to act against their desires. Most psychologists and neuroscientists have in mind this weak notion of motivated choice when they refer to “free will.” For example, Patrick Haggard’s defense of free will is targeted against the view that there is no such thing as a will, intentions being merely retrospective rationalizations of unconscious decisions to act (Haggard, 2006; Wegner, 2003a). Having free will, in this context, merely means that agents are not blind actors; their intentions have a causal role in guiding their actions, and they can recognize which actions they are performing while they perform them. This weak, hypothetical interpretation, however, does not allow one to speak of unqualified freedom of the will, for it is plausible that agents act as they do because they are determined to prefer what they do. As Haggard himself makes clear, “From a neuroscientific perspective, a conscious intention cannot be the ultimate cause of our actions, since that would imply mind-body causation” (Haggard, 2006, p. 69).

If, however, agents have no control over their choices, then it is difficult to hold them morally responsible for their acts. A stronger reading of “being able to do otherwise” has been offered to account for moral and legal responsibility: Free will now entails that free agents must be able to choose otherwise. This is the view that libertarians aim at defending: Agents are not determined by their “cognitive history”; they possess uncaused causal power. Discussions of these various claims about the nature and scope of responsibility are thus at the core of the free will debate. The table below presents an overview of the main claims on offer. There are two fundamental questions that organize the discussion about free will: (1) Are mental states causally determined? (2) Is causal determinism, as applied to the mental realm, compatible with free will? Below we present and discuss the positions defined at the intersection of these two questions, with an eye to the existence and cogency of neuroscientific evidence available in each case. To anticipate, the bottom line of this chart includes positions fully compatible with a naturalistic framework, whereas the top line must accept some form of mind-body dualism.


Incompatibilism between determinism and free will

Compatibilism between determinism and free will

Determinism does not apply to the mental

Libertarianism: free will

Free will

Determinism applies to the mental and the physical

No free will

“Subjective” or “motivational” free will


There are two ways in which deterministic causation is seen as excluding free will—that is, two ways of defending incompatibilism between determinism and free will, depending on which side of the relation is rejected. Among scientists, the most usual form of incompatibilism consists in rejecting the freedom of the will on the basis of the detailed factual knowledge available about cognitive functions. On this view, whether determinism may fail to apply to the mind-brain is no more a matter of personal opinion than the existence of unicorns (Prinz, 2006). Libertarianism, in contrast, tries to solve the puzzle of free will by arguing that human agency allows exceptions to causal determinism to occur in mind-brain activity.


Libertarianism has been a classical position for theologians and dualist philosophers such as Descartes, or the neuroscientist John C. Eccles (1994). More recently it has been espoused by neuroscientists and philosophers with monist inclinations (Libet, 1999; Searle, 2006).

A first reason to endorse libertarianism is the intuition that we are free to act as we wish unless externally restrained: This is not only a powerful experience that agents enjoy when they act voluntarily and bitterly miss if coerced to act; it is also a conviction that motivates agents to avoid specific courses of action and to select others—that is, to act in responsible ways. The agents’ belief that impulses can be resisted is thus part of a larger control system (including training in self-control as well as moral, political, and religious values and attitudes) through which they become able to defer satisfaction, comply with rules, and aim for cooperative goals. Free will undoubtedly constitutes one of the bases of our western legal, moral, and political systems in addition to being a key representation in (contemporary) human culture. In contrast, espousing determinism seems to lead to unintelligible public and legal policies and, given the present incomplete state of our knowledge, seems to constitute a socially dangerous speculation. Although this kind of argument is an obvious reason to endorse libertarianism, it can also be accommodated within an incompatibilist view on voluntary decision, as we shall see below.

A second, different argument consists in taking at face value the experience of freedom that agents have when they act: This experience offers “prima facie evidence that conscious mental processes can causally control some brain processes” (Libet, 1994). There are two main ways of articulating this view.

The first consists in positing a sui generis agentive form of causation that is “irreducible to event-causal processes within the agent” (O’Connor, 1995). Free will is taken to be “a macrodeterminative emergent property of agents” (O’Connor, 2000, p. 111). Emergent properties are called sui generis in the sense that they are not present at the lower level from which they emerge: Agents are, supposedly, not caused. They are, rather, in Chisholm’s (1982) terms, “prime movers unmoved.” Granting, then, that agents—rather than their mental states—are causal, what is the relationship of this type of causality to more ordinary types of mental causation? This alleged form of free agency must presuppose that, at some points, selves acquire the power to use their own brains as instruments of their will (shutting them off or activating them according to circumstances)—a claim that sits uncomfortably with the view that mental function is nothing other than brain activity.

The second way consists in rejecting agent causality while hypothesizing an objective causal “gap” between the causes that influence the action on the one hand and the decision and its consequences on the other (Searle, 2006). On this view, the experience of freedom leads one to form the true belief that one escapes causal necessitation at the moment of deciding to act; this impression includes the possibility of acting against one’s best reasons or for no reason at all. An obvious objection to this view is that such an experience might be a trick devised by evolution to motivate an agent to act and an efficient way of clarifying what the agent can or cannot do (feeling free when able to control a changing situation versus feeling forced when unable to control a change). Extending this impression in a metaphysical sense, to take it as referring to an actual gap in the action causal sequence, might be as illusory as taking the Müller-Lyer segments* to have objectively different lengths

Other libertarians, for this reason, do not use the experience of freedom as their main argument. Instead, they try to find a positive, scientifically respectable reason in favor of the existence of a causal gap, such as quantum indeterminacy. The idea is that, being nondeterministic, quantum mechanics might account for the dynamic emergent character of free actions: Introspection, normative conflict in a decision process, might allow a form of quantum indeterminacy to appear. Although quantum effects normally occur only at the level of the individual atom, it is speculated that they could be amplified through the interplay between chaotic states as a result of the heightened sensitivity to the smallest changes in initial conditions that such an interplay would produce (Bishop, 2002). This hypothesis, however, is at present a mere speculation: Amplification of quantum effects in the brain and neural chaotic states do not license, as far as is presently known, macrolevel forms of indeterminacy. An alternative possibility would be to extrapolate from far-from-equilibrium to agentive systems (Bishop, 2002). According to Ilya Prigogine (1997), organisms are unstable, self-organized systems existing far from thermodynamic equilibrium. Being sensitive to initial conditions, unstable systems can only be explained statistically—that is, in probabilistic terms. Instability is supposed to justify the claim that such systems defy standard deterministic explanation. Thus free will would be another name for instability. Although the view that brains are complex self-organized systems has gained wide currency, this view does not warrant the case for nondeterministic states at the macrolevel of decision making (Churchland, 2002, p. 207). Furthermore, it is not obviously coherent to identify freedom of the will with unstable, unpredictable physical states, for how can an agent exert reliable control on her movements and actions if her brain cannot predict its activity?

Scientific Incompatibilist Objections

An obvious objection from neuroscientific quarters consists in denying that there could be any intervention of an acting self over and above that of the mental processes involved in an agent’s voluntary decisions and actions (including the narratives produced to emphasize self-identity over time). Personhood is a mental construct (i.e., an outcome, rather than a precondition, for having mental dispositions and representations). From a functional viewpoint, the self is constituted of a set of reflexive abilities and representations that an individual agent acquires in the course of its development (including representations for persistence and coherence in goals and values, for responsibility for one’s actions, and for continued existence through episodes) (Koechlin & Sommerville, 2007). These human-specific abilities and representations have reflexive/normative components that foster cooperative behavior and conformity with social rules (Proust, 2003, 2013).

The reflexive components that are crucial for building up self-representations point to the existence of underlying adaptive control loops with control and monitoring dimensions. As parts of a control structure, self-representations are activated when a task involves biasing the flow of information to endogenous goals (for example, maintaining covertly the intention to perform a given action later: a prospective memory task) or retrieving self-relevant information (in contrast to information about the outer world). This type of self-representation activates the rostral, medial, and lateral prefrontal cortex, an activity also involved in mental rehearsal and simulation of behavior when connected with parietal areas (Burgess et al., 2007; Koechlin et al., 2003; Lou et al., 2004). As parts of an action-monitoring structure, self-representations depend upon basic mechanisms distinguishing self-produced actions from those generated by others, a function served by the left and right inferior parietal, precuneus, and somatosensory cortex (Ruby & Decety, 2001). This set of representations is sometimes referred to as constituting the “who system” (Georgieff & Jeannerod, 1998). A self, however, is also a normative construct, which contributes to motivate the selection of actions that fulfill self-identificatory values and goals and inhibits actions that violate them (Blair & Blair, 2009; Prinz, 2006; Proust, 2013; Rochat, 2009).

These various results lead one to reject the view that a self is an unmoved mover—that is, an entity that can, even occasionally, be the ultimate cause of actions, for there is actually no such thing as “the self.” There are, rather, many ways of focusing one’s attention on endogenous goals and associated feedback, involving emotions, contexts of action, motor or social cognition, and long-term motivations and values. Some of these “self-factories” may survive perturbations in the others, as is shown in psychopathological impairments, such as schizophrenia, or in neuropsychological impairments, such as dementia.

Neuroscientific research, then, offers tools for making more precise the role of reflexive mechanisms in action and building up from there a position that, at least prima facie, seems to suit an incompatibilist picture. The first step in the debate consisted in finding the neural signature of a decision to act when such a decision is spontaneous (i.e., purely endogenous). Kornhuber and Deecke (1965) found that such an action is associated with a readiness potential (RP) (i.e., a slow electrical change recordable on the scalp at the vertex. This RP was later found to be specific to particular movements, rather than expressing a general preparation to action, and to be produced in the supplementary motor area (SMA). Lesions in the SMA, indeed, prevent spontaneous actions from developing—a condition called akinesia.

Granting that a specific RP launches each specific action, the question was raised of its causal and temporal link with the associated conscious intention to act. Benjamin Libet addressed this question by asking subjects to perform a flexion of the wrist when they wished and to report the clock time at which they were first aware of their urge to act (Libet, 1983). The conscious intention to act was found to occur on average 350 milliseconds before the activation of the muscle but roughly 400 milliseconds after the readiness potential in SMA started to send signals to the muscle. This result, widely replicated, suggested that unconscious processes in the brain are the true initiators of volitional acts. How then could one’s will be free if one makes a decision before one becomes conscious of having made it?

Libet, however, attempted to show that, although an intention to act precedes conscious realization of having this intention, free will still manifests itself as the capacity to inhibit the developing act. Reasoning that the 50-millisecond delay from RP to muscle activation leaves time for the motor cortex to inhibit the activation of the spinal motor neurons, Libet concluded that a conscious will can, after all, operate by “stopping or vetoing the volitional process” (Libet, 1999). To qualify as “free,” this vetoing process should in turn stem from a conscious decision rather than being unconsciously initiated. Showing this, however, proves thorny: Libet proposes to distinguish the awareness of a decision to veto from the content of that awareness—a proposal that uses a version of the “gap” theory of free will: The conscious decision to veto “could still be made without direct specification for that decision by preceding unconscious processes” (Libet, 1999, p. 53). It is quite unclear, however, how a specific decision could occur without an unconscious processing of cognitive contents and motivations which would prompt the conscious decision for or against vetoing (Oakley & Haggard, 2006).

For those rejecting an ad hoc interpretation of the evidence, the main interest in Libet’s experiment is to pry apart two aspects of agency conflated by common sense and by philosophical tradition: the conscious, personal-level origination of intentions to act and the agents’ feelings of being in control. Agents do not need to consciously form intentions before acting (and thus be conscious originators of their acts) to have agentive feelings. What is it, then, that constitutes the experience of a conscious will? Most scientists agree that the experience of action does not contribute to causing behavior, but they disagree on how reliable it is. Some authors claim that such an experience frequently departs from the actual causal processes driving action; it can only provide an apparent causal role to a conscious will (Wegner, 2003a). Others maintain that the experience of the will is rooted in the causal structure of the action and proves reliable for that reason (Moore et al., 2009).

Is the Conscious Will “Mind’s Best Trick”?

On Wegner’s view, unconscious mental processes cause both our thoughts and our actions: They lead us to an impression of conscious will. In fact, however, there is no way for an agent to know whether her conscious thoughts have caused her actions. Conscious will is merely an “illusory add-on to action” (Wegner, 2003a,b). To show this, Wegner collected evidence from various fields: (1) The neurosurgeon Wilder Penfield used to stimulate the brain, before operating, with electrical probes while the patients were conscious on the operating table (under local anesthesia). He found that stimulation of the exposed brain elicited in patients movements similar to voluntary ones; patients denied, however, having carried out the corresponding actions. This suggests that the experience of will is not a cause or even a necessary concomitant of action. (2) Transcranial magnetic stimulation (TMS) can be applied to subjects’ motor cortices: It produces activity in the contralateral muscles targeted. If the subjects are instructed to move their left or right fingers when they hear a click (when a TMS occurs), they do not realize—at short response times—that their fingers are moved by the machine rather by themselves. (3) Gazzaniga’s split-brain patients, induced to perform an action through the right hemisphere (inaccessible to the verbal areas in the left hemisphere), confabulate intentions to account for their actions (Gazzaniga, 1995). Other patients, on the other hand, such as those with alien hand syndrome or schizophrenia, may deny having intentions to act as they do and often even ascribe their actions to external agents. The best explanation for these various facts, according to Wegner, is that consciousness of action is a variety of causal reasoning: A subject needs to infer that she is the agent of an action on the basis of the evidence available to her. There are three such sources of evidence: (1) the conscious thought must immediately precede the action, (2) it must be coherent with the action, and (3) there must be no alternative explanation for the action (Wegner, 2003a). If these three conditions obtain, there is no further way for an agent to recognize apparent from real causality. This inferentialist strategy, however, must address a difficulty: What the various experiments show is that an agent may, in certain biased situations, retrospectively form wrong self-attributions of agency. What remains to be shown is that there is no real difference between how a subject perceives her movements in active versus passive conditions. To this we now turn.

Consciousness of Action and Causal Structure

While inferentialists view action consciousness as an attributive matter, others, whom we might call intentional realists, hypothesize that consciousness of agency is reliably anchored in the causal structure of the control of action. Patrick Haggard and his colleagues have studied consciousness of agency through the perceived association between an action and its consequence, or effect (Haggard et al., 2002). In a Libet-style paradigm, subjects are invited to carry out key presses at will, which may or not produce an auditory stimulus. Subjects are asked to use a rotating clock hand to estimate either when they performed the action or when they heard the tone. Researchers found a strong temporal attraction, called intentional binding, between action and tone relative to baseline control conditions in which actions and tones did not co-occur. No such binding, however, was produced when the action was nonvoluntary but evoked by transcranial magnetic stimulation. Thus inferentialism seems defeated: The brain does not create in an agent the impression of acting unless an efferent motor command is sent to control motor execution. The same binding occurs for auditory, visual, or somatic effects, whether the effect was on the agent’s body or in the external world.

Further work, however, has subsequently exposed the grain of truth in inferentialism (Moore et al., 2009). Manipulating the contingency between voluntary actions and their effects by varying the probability of a tone occurring in the absence of action reveals that intentional binding is sensitive to background causal contingency. Agents’ knowledge of the background causal relations determines their subjective evaluations of the time at which they started to perform an action and observed its outcome, whether or not the agent was or was not acting. In summary, subjects seem to form, in addition to or as part of their motor model of an action, a causal model of the probabilistic contingencies between action and tone; this model modulates their conscious experience of agency at any given time.

These findings help us gain a better understanding of what it means to feel in control. This feeling is anchored in reliable information from two sources, one involving the brain’s predictions concerning the consequences of its own decisions and movements, which are subsequently experienced as the action feedback, and the other the contingencies extracted from the environment, allowing agents to discriminate the effects of their agency from what would happen anyway.

These findings also provide a way of interpreting responsibility and free will, in the weak sense indicated above, from an incompatibilist-scientific viewpoint. An agent might be judged responsible for her actions if she performs an action with the associated feeling of controlling it. As psychopathological perturbations such as schizophrenia show, an agent can act (i.e., carry out movements in a controlled way) without having the associated feeling of controlling the action. Lack of this feeling prevents her from vetoing and probably also from evaluating the consequences of her actions. Such an agent, in contrast to unimpaired agents, should not be deemed responsible for what she does. Recognizing a measure of responsibility does not amount to defending the presence of free will in the strong sense, however, because the agent does not need to be able to choose otherwise. To show this, we need to turn briefly to the brain structures that realize ordering among preferences and choice.

Dopamine Signal, Reward, and Choice

One of the issues relevant to exploring free will, left unadressed by the work reviewed above, is whether and how pleasure drives behavior. Is finding something pleasurable (i.e., attributing to it an expected reward value or “liking” it) always required for being disposed to act to get it (or “wanting” it)? These two dimensions, called experienced utility and decision utility, are in fact subserved by different neural areas that may occasionally be dissociated in rats as well as in humans (Berridge & Aldridge, 2009). In the domain of food, for example, liking involves opioid and GABA/benzodiazepine systems, subserved, inter alia, by the ventral pallidum. Wanting involves mesotelencephalic dopamine systems, which are subserved by the nucleus accumbens and amygdala. Dopamine, thus, is “a faux pleasure”: it rather attributes “incentive salience to reward stimuli, which causes them to be perceived as attractive incentives.” It provides “a common neural currency for decision utility in evaluating multiple choices” (Berridge & Aldridge, 2009, p. 518).

A traditional conception of free will would welcome the finding that wanting and liking belong to different systems (i.e., that decision utility does not always coincide with experienced or predicted hedonic value), for this dissociation should allow an agent to choose to do things that are not immediately pleasurable: Agents can choose to eat what they do not like; they can also choose an action that will bring them pain or even certain death. A rational decision is such that the decision utility must equal inclusive predicted utility (including, for example, the utilities of loved ones). The problem, however, is that a dissociation may occur without the agent having chosen to “want” to obtain what she dislikes, taken in the sense of an inclusive predicted utility. This leads to irrational decisions: In certain food (or drug) addictions, subjects cannot stop wanting to ingest food (or heroin) when presented with the opportunity, even though no pleasure is experienced or expected. This case presents a serious problem for the claim that an agent has free will in the libertarian sense that she always could have chosen to do otherwise.

Two additional findings should prove still more devastating for incompatibilist libertarians: (1) wanting (as decision utility decoupled from liking) can drive behavior with no associated subjective awareness. Evidence for this is provided in recent brain imaging: Physical effort exerted to gain various monetary rewards is modulated by pallidal activation as a function of the monetary stakes involved—an activation independent of a conscious perception of the critical stimulus (Pessiglione et al., 2006, 2007). (2) Agents, however, can ignore, vastly misrepresent, or conflate their likings and wantings as their actual reasons to act (Berridge & Aldridge, 2009; Nisbett & Wilson, 1977).

In summary, incompatibilism is the view that one cannot simultaneously claim that agents can have free will and that all mental states of agents are the output of causal mechanisms. Denying the first part of the alternative seems to threaten the existence of responsible agents as well as contradicting agents’ intuitions that they can always act as they wish and when they wish. Denying the second flatly contradicts psychological and neuroscientific evidence on how agents decide to act and become conscious of their actions. The difficulties associated with each of these disjuncts suggests that we should turn to another way of understanding free will. This new approach, compatibilism, admits that an agent is causally determined to choose what she does (and thus could not have chosen to act otherwise) but rejects the relevance of determinism to free will. The latter has nothing to do with the causal history of an act but is rooted in the structure of the will.


The two main motivations for compatibilism are (1) the observation that, although science offers a deterministic picture of the world, agents engaged in action feel free to do otherwise except when coerced, and (2) the recognition that rejecting free will has deleterious consequences on how to deal with responsibility and moral obligation. Compatibilists do not question the fact that, at every moment, an agent is not in a position to break the causal sequence that leads to a particular act. She could not have chosen to act otherwise. But this does not need to prevent her from having formed a willful action. Take John Locke’s example of an agent who wants to stay in a room for specific reasons; unbeknownst to her, the door is locked from outside—she could not have left the room had she wished to. She nevertheless acts voluntarily. The locked door can be used as a metaphor for what determinist causation prevents one from doing. The same conclusion still applies: Even though the person is unable to do otherwise because of her brain wiring, she need not have acted because she could not have done otherwise. Therefore assessing moral responsibility only needs to consider why one acted as one did and how an action sequence unfolded rather than whether one could have done otherwise.

Harry Frankfurt’s “hierarchical mesh” theory (1988) offers an influential compatibilist theory of free will based on an analysis of the layered structure of the will. In this analysis two distinctions are introduced:

  1. 1. First-order/second-order desires: First-order desires bear on certain properties or states that the agent wants to acquire; second-order desires bear on the first-order desires that the agent “identifies with.”

  2. 2. Efficient/inefficient desires: A desire is efficient in a given action if it controls it. An efficient desire is called a volition. There are first-order as well as second-order volitions.

Let us consider three drug addicts who, ex hypothesi, have no other behavioral choice than taking drugs. Each agent, however, has a different structure of the will. Addict 1 takes heroin in spite of a strong but inefficient second-order desire (to resist the urge to take it). Addict 1 is in the situation of a subject who wants some food but does not like it, and does not like to act the way she does (cf. preceding section). Addict 1 would like to resist the urge. Addict 2, in contrast, does not care about taking the drug or not: she never evaluates her first-order desires. She blindly follows her impulses. Addict 3, finally, has an efficient second-order desire to take the drug; her first-order desire is weakening but, given her identification with that desire, her second-order volition to take the drug controls her action. Which one of them has free will, and why? According to Frankfurt, a person’s will may be free even though her action is not, just in case (1) she has the ability to form efficient second-order desires and (2) the desire she acts on meshes with the second-order desire of “identifying with” it. Granting these two conditions, Addict 2 has no free will because condition 1 is not met (she only has first-order desires). Addict 1 is not free because condition 2 is not met (she does not act on the desire she prefers to have). Addict 3, in contrast, has free will because she acts on a desire that meshes with her second-order desire (concerning which first-order desire she wholeheartedly selects). In other terms, having free will is a matter of internal harmony between one’s motivations, issuing in wholehearted decisions. The source of enslavement, on the other hand, lies in motivational division.

Let us note, however, that free will, in this compatibilist perspective, does not entail that agents can form second-order desires at will or that they are always able to act on those of the first-order desires they identify with. These limitations are welcomed by a scientific determinist, who is sensitive to the fact that a subject can only control the actions that belong to her control repertoire and can only form higher-order desires if she has the associated representational and executive capacities—in particular, metacognitive and mind-reading abilities for higher self-monitoring. Such mechanisms are indeed basic to what Frankfurt calls “self-identification with a first-order desire.” Cognitive-neuroscientific research has studied the impairments of these abilities and the ways in which they constrain both agentive control and the sense of agency—the sense of having performed an action deliberately (Proust, 2013, ch. 12).

Some readers, however, might feel reluctant to attribute free will to an agent even when the agent is determined to act. The following dilemma seems to seriously threaten the coherence of compatibilism. Either determinism is true or it is false. If it is true, second-order volitions are causally determined by the agents’ former states and by the external circumstances. But then the will is only free to the extent that motivational resonance across desire levels makes it appear free to the agent. The alternative is that determinism is false: In this case, the three addicts could equally well have done otherwise. The structure of their will has nothing to do with it.

Even though one can be reluctant to attribute free will to an agent, however, one must recognize that agents who retrospectively dislike their decisions may try to change them on future occasions: They may be motivated to extend their control abilities. As Frankfurt observes, “A person who is subject to some volitional necessity is not thereby precluded from trying to eliminate it altogether by taking steps that are designed to alter the causally relevant circumstances from which the necessity derives” (Frankfurt, 2002, p. 162).

The effort to achieve greater self-control suggests, in turn, that an agent may be more or less in control of what she does, more or less dependent on her environment, and thus have a will that is more or less free. Let us see how this could make sense. Wanting to act in a certain way is contingent on having learned how to represent motives and constraints for that action. An agent who represents more goals, and more constraints on reaching them, is freer than an agent with no such ability; in other words, she is less dependent on the present environment for her decision to act. Research by Etienne Koechlin (Koechlin & Sommerville, 2007) offers insight into how—in phylogeny as well as ontogeny—the brain becomes able to resist impulsive responses to the world. In sensory control, an organism learns to associate a movement with a perceived stimulus through specific premotor cortex activations. Sensory control can be made more discriminative if contextual cues can be used to modulate the motor response: in that case, part of the prefrontal lateral caudal cortex is also activated. When, however, contextual goals need to conform to episodic rules, a third area, in the rostral lateral prefrontal cortex, kicks in. Finally, distractor-resistant, task-switching requirements are represented at the “branching” level, which activates the polar lateral prefrontal cortex in lateral Brain Area 10. Extrapolating from this model, we conclude that becoming freer does not mean just that an agent operates on the basis of endogenous, internalized goals, including a representation of the person she wants to be, but that she can do so flexibly when a task requires autonomy from current stimulation.

This cascade model of the will explains how Frankfurt’s mediating steps can be implemented. An agent who aspires to free will may acquire a higher capacity to resist impulses to respond to present environmental affordances than an agent who has no such concern. She does not thereby become free to act; however, she has access to a larger repertoire of goals and constraints. In this relative sense, one being is freer to act than another if a larger repertoire is open to her for responding to situations that require more flexible cognitive or emotional control. It is thus perfectly coherent to claim that a self-imposed requirement of becoming freer and more responsible can be favorable to an extension of one’s ability to control one’s actions while also maintaining that an individual’s will is determined by her cognitive architecture, learning history, and present motivations.


I am grateful to Dick Carter for his remarks and linguistic help. The preparation of this chapter was supported by the European Science Foundation EUROCORES Programme CNCC, with funds from CNRS and the EC Sixth Framework Programme under Contract no. ERAS-CT-2003-9804094.


Berridge, K. C., & Aldridge, J. W. 2009. Decision utility, incentive salience, and cue-triggered “wanting.” In Morsella, E., Bargh, J. A., & Gollwitzer, P. M. (Eds.), The Oxford handbook of human action (pp. 509–532). New York: Oxford University Press.Find this resource:

Bishop, R. C. 2002. Chaos, indeterminism and free will. In R. Kane (Ed.) The Oxford handbook of free will (pp. 111–124). New York: Oxford University Press.Find this resource:

Blair, J. R., & Blair, K. S. 2009. Empathy, morality, and social convention: evidence from the study of psychopathy and other psychiatric disorders. In J. Decety (Ed.), The social neuroscience of empathy (pp. 139–152). Cambridge, MA: MIT Press.Find this resource:

Burgess, P. W., Dumontheil, I., & Gilbert, S. J. 2007. The gateway hypothesis of rostral prefrontal cortex (area 10) function. Trends in Cognitive Sciences, 11(7), 290–298.Find this resource:

Chisholm, R. 1982. Human freedom and the self. In G. Watson, (ed.), Free will (pp. 24–35). New York: Oxford University Press.Find this resource:

Churchland, P. 2002. Brainwise. Cambridge, MA: MIT Press.Find this resource:

Eccles, J. 1994. How the self controls its brain. Berlin: Springer.Find this resource:

Frankfurt, H. G. 1988. The importance of what we care about. Cambridge, UK: Cambridge University Press.Find this resource:

Frankfurt, H. G. 2002. Reply to Gary Watson. In S. Buss & L. Overton (eds.), Contours of agency. Cambridge, MA: MIT Press.Find this resource:

Gazzaniga, M. S. (1995). Consciousness and the cerebral hemispheres. In M. S. Gazzaniga, (Ed.), The cognitive neurosciences (pp. 1391–1400). Cambridge, MA: MIT Press.Find this resource:

Georgieff, N., & Jeannerod, M. 1998. Beyond consciousness of external reality: A “who” system for consciousness of action and self-consciousness. Consciousness and Cognition, 7, 465–477.Find this resource:

Haggard, C., & Kalogeras, J. 2002. Voluntary action and conscious awareness. Nature, 5, 285–382.Find this resource:

Haggard, P. 2006. Conscious intention and the sense of agency. In N. Sebanz & W. Prinz (eds.), Disorders of volition (pp. 69–85). Cambridge, MA: MIT Press.Find this resource:

Hume, D. 1978. A treatise of human nature. Oxford, UK: Oxford University Press (Original work published 1739–1740).Find this resource:

Koechlin, E. & Sommerville, C. 2007. An information theoretical approach to prefrontal executive function. Trends in Cognitive Science, 11(6), 229–235.Find this resource:

Koechlin, E., Ody, C. & Kouneiher, F. 2003. The architecture of cognitive control in the human prefrontal cortex, Science, 302(5648), 1181–1185.Find this resource:

Kornhuber, H. H., & Deecke, L. 1965. Changes in the brain potential in voluntary movements and passive movements in man: Readiness potential and reafferent potentials. Pflugers Archive für Gesamte Physiologie der Menschen & Tiere, 284, 1–17.Find this resource:

Lou, H. C., Luber, B., Crupain, M., Keenan, J. P., Nowak, M., Kjaer T. W., … Lisanby, S. H. 2004. Parietal cortex and representation of the mental self. Proceedings of the National Academy of Sciences, (101)17, 6827–6832.Find this resource:

Libet, B. 1994. A testable field theory of mind-brain interaction. Journal of Consciousness Studies, 1, 119–126.Find this resource:

Libet, B. 1999. Do we have free will? Journal of Consciousness Studies, 6, 8–9, 47–57.Find this resource:

Libet, B., Gleason, C. A., Wright, E. W., & Pearl, D. K. 1983. Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential). Brain, 106(3), 623–642.Find this resource:

Moore, J. W., Lagnado, D., Deal, D. C., & Haggard, P. 2009. Feelings of control: Contingency determines experience of action. Cognition, 110(2), 279–283.Find this resource:

Nisbett, R. E., & Wilson, T. D. 1977. Telling more than we can know: Verbal reports on mental processes, Psychological Review, 84(3), 231–259.Find this resource:

Oakley, D. A., & Haggard, P. 2006. The timing of brain events: Authors’ response to Libet’s “reply.” Consciousness and Cognition, 15(3), 548–550.Find this resource:

O’Connor, T. 1995. Agent causation. In T. O’Connor (Ed.), Agents, causes and events: Essays on free will and indeterminism (pp. 173–200). Oxford, UK: Oxford University Press.Find this resource:

O’Connor, T. 2000. Persons and causes: The metaphysics of free will. New York: Oxford University Press.Find this resource:

Passingham, R. E., & Lau, H. C. 2006. Free choice and the human brain. In S. Pockett, W. P. Banks, & S. Gallagher (Eds.), Does consciousness cause behavior? (pp. 53–72). Cambridge, MA: MIT Press.Find this resource:

Pessiglione, M., Schmidt, L., Draganski, B., Kalisch, R., Lau, H., Dolan, R. J., & Frith, C. D. 2007. How the brain translates money into force: A neuroimaging study of subliminal motivation, Science, 316, 904–906.Find this resource:

Pessiglione, M., Seymour, B., Flandin, G., Dolan, R. J., & Frith, C. D. 2006. Dopamine-dependent prediction errors underpin reward-seeking behaviour in humans. Nature, 442, 1042–1045.Find this resource:

Prigogine, I. 1997. The end of certainty. Time, chaos, and the new laws of nature. New York: Free Press.Find this resource:

Prinz, W. 2006. Free will as a social institution. In S. Pockett, W. P. Banks, & S. Gallagher (Eds.), Does consciousness cause behavior? (pp. 257–276). Cambridge, MA: MIT Press.Find this resource:

Proust, J. 2003. How voluntary are minimal actions? In S. Maasen, W. Prinz, & G. Roth (eds.), Voluntary action (pp. 202–219). Oxford, UK: Oxford University Press.Find this resource:

Proust, J. 2013. The Philosophy of Metacognition. Mental agency ad Self-awareness. Oxford, UK: Oxford University Press.Find this resource:

Rochat, P. 2009. Others in mind. Cambridge, UK: Cambridge University Press.Find this resource:

Ruby, P., & Decety, J. 2001. Effect of subjective perspective taking during simulation of action: A PET investigation of agency. Nature Neuroscience, 4, 546–550.Find this resource:

Searle, J. R. 2006. Freedom and neurobiology: Reflections on free will, language, and political power. New York: Columbia University Press.Find this resource:

Wegner, D. 2003a. The illusion of conscious will. Cambridge, MA: MIT Press.Find this resource:

Wegner, D. 2003b. The mind’s best trick: How we experience conscious will. Trends in Cognitive Sciences, 7(2), 65–69.Find this resource:


(*) The Müller-Lyer segments are composed of a set of arrow-like figures, whose fins are head-shaped or fin-shaped. The segments forming the shafts, although they have the same length, are perceived to be longer when they have two tails than when they have two heads.