1. Intentionality

“Intentionality” is the label coined by the Scholastics in the Middle Ages to describe a basic feature of our psychology, which is that we act and believe guided by our representations of the world.Footnote ²⁰ Many but not all of our mental states are representational states in that they take “objects,” have “content,” are “directed upon propositions,” and so on. One does not just hope, fear, intend, desire, simpliciter; rather one hopes that tomorrow will be better than today, fears that it will not be, wants it to be so, or intends to make it the case that it will be so. What follows the “that” in such constructions is said to be the content of an Intentional state.

Intentionality is basic for creatures who would understand and manipulate their environment. For it is our capacity to represent (as the objects of our beliefs) the world as it is that gives us theoretical understanding, and it is our capacity to represent the world as the objects of our desires and intentions that gives us some measure of control over that world.

Despite its familiarity, and despite its rock-bottom status within folk psychology, Intentionality is a surprisingly complicated affair. For it is notoriously difficult to fix the sense and the reference of words used to describe the content of Intentional states.Footnote ²¹ This has led some to the far-fetched schemes of a kind of dualism about the mind, such as Franz Brentano's idea that the objects of Intentional states exist in a special way he termed, “Intentional Inexistence.”Footnote ²² This difficulty about the content of Intentional states has led others to a skepticism about their existence on grounds antedating and existing independently of the discoveries of neuroscience.Footnote ²³ At present, it is something of an article of faith that the best cognitive science will find a place for the Intentional states of folk psychology. Nonetheless that is a faith to which common sense, morality, and the law wholeheartedly subscribe.Footnote ²⁴

2. Three parts to the soul

Plato famously divided the soul into appetitive, rational, and executory “parts.”Footnote ²⁵ Despite its ancient lineage, this is still a fairly good functional division among the Intentional states that guide our actions and that, thus, constitute our agency.Footnote ²⁶ Such states are motivational, cognitive, or conative: we represent the world as we desire it to be, as we believe that it is, and as we intend to make it into. Beliefs, desires, and intentions are the three modalities of representional states most constitutive of us as agents who act.

3. Practical rationality

The three representational states of desire, belief, and intention come in ordered triads within our psychology. Although often explicit reference is made to just one state in the explanation of human action, such seemingly sole references are elliptical for a triad of such states. Moreover, there is an order to this triad in the form Aristotle called a practical syllogism.Footnote ²⁷ Suppose I am headed toward an umbrella store and that this action requires explanation. A full, reason-giving explanation would consist of three “premises”:

1. a. I desire not to get wet.

2. b. (Believing as I do that it will soon be raining, nonetheless): if I buy an umbrella, I will not get wet.

3. c. I intend to buy an umbrella soon.

Notice that the content of my intention matches the antecedent clause of my means/end belief, and that the content of my desire matches the consequent clause of that belief. Notice also that I have not yet fully explained my walking toward the umbrella store. To complete the rational reconstruction of my reasoning process I must add another means/end belief and a volition:

1. d. I believe that if I move my feet thus and so, I will get to the store and buy an umbrella.

2. e. I will my feet to move thus and so.

These relationships are so definitive of our understanding of each other as rational actors that we readily ascribe the full set to an actor even if only one of them is explicitly mentioned. For example, idiomatic answers to the question, “Why are you headed toward the umbrella store,” could easily be: “because I do not want to get wet;” or, “because I believe it is going to rain soon.” These highly elliptical answers are fully intelligible to us because we are so familiar with the form of a full practical syllogism.

4. Emotionality and the rationality of emotion

Both psychopathic personality types and some silicon-based intelligence systems imagined in science fiction remind us that an absence of emotions marks a being as less than fully human.Footnote ²⁸ An older view regards the necessary emotionality as a regrettable aspect of human nature; this, on the picture that emotions distract us from the rational and the moral.Footnote ²⁹ Yet a better view sees rational functions for the emotions: they can be the harbingers of moral insight (their epistemic role),Footnote ³⁰ and they can create value for projects that morality makes discretionary (their constitutive role). There is also an intelligence to the emotions that fits them for these roles, an intelligence cashed out in terms of the virtue/vice of holding certain emotions on certain occasions, of the norms of appropriateness of kinds of emotions in relation to their objects, and of the norms of proportionality in the intensity of emotion measured against its causes.Footnote ³¹ A recognizably human person is one who is passionate about certain projects, hates injustice, and grieves over objects whose loss is worth grieving about.

5. Character structure at a time and over time

Everybody has many beliefs, desires, intentions, and emotions. To be recognizably human requires more order to them beyond their being grouped into the forms of valid practical or theoretical inference. In addition, the entire constellation of an individual's Intentional states must instantiate an intelligible character structure.Footnote ³² This must be true at a given time; but it also must be true to some extent over time as well. Acts out of (one's present) character are possible; and changes of character over time are even more familiar. But fractured personalities, and temporally disjointed chameleons, are not the kind of beings we each take ourselves to be.

6. Autonomy

It is the basic sense of “autonomy” that I intend here (there are other senses of the word). Namely, folk psychology supposes that our belief/desire/intention (“BDI”) sets cause our voluntary actions.Footnote ³³ That is not a “free will” postulate, at least not in the sense of “free will” that demands the will to be uncaused. Rather, the supposition is that the will causes the behaviors that are represented in its objects.Footnote ³⁴ When I intend to kill a man by moving my finger (on the trigger of a gun), both my finger moving and his death are the causal products of my intent to kill and my intent (or volition, willing) to move my finger.

7. Consciousness₁: phenomenal experience

The other deeply mysterious feature of our mind (along with Intentionality) is that we have conscious experience. This is how we are supposed to differ from “mindless” robots and, perhaps, from other animals. It once was thought that the Intentional states of belief, desire, intention, and emotion only existed as different forms of conscious experience. As an early philosophical critic of Freud once put this idea, “when I reflect on what I mean by a wish or an emotion or a feeling, I can only find that I know and think of them simply as different forms of consciousness.”Footnote ³⁵ But this is plainly too circumscribed a view of when those Intentional states exist. Conscious experience is crucial to us being who we are, but it is not that crucial. We have many beliefs, desires, intentions, and emotions that are not conscious in this sense. Long before Freud became popular, William James noted that many of the mental states that guide skilled or habitual behavior do so without being the subject of conscious attention.Footnote ³⁶

Still, if none of such states were the subject of conscious experience, we would be very different creatures than in fact we are. There is a centrality to conscious deliberations in terms of desires, beliefs, intentions, and emotions such that the nonconscious versions of these are nonetheless tied in one way or another to the conscious versions. As even Freud acknowledged, “the property of being conscious or not is in the last resort our one beacon-light in the darkness of depth-psychology.”Footnote ³⁷

This is because there is a core identification of self with our conscious experience. Each of us experiences himself or herself as a unified self because we experience a continuity between our conscious experiences, both at a time and over time through memory, as Locke famously noted.Footnote ³⁸ We identify mental states as part of ourselves through their accessibility via our conscious experience of them. And we sense that we are in control of our bodies and, through our bodies, the larger environment, by virtue of our conscious willings issuing causally in our voluntary actions.Footnote ³⁹

8. Consciousness₂: privileged access

Paradoxically, we can be conscious of a mental state without having any datable experience in consciousness of that state. James's examples of skilled or habitual behavior serves to illustrate this. We can guide our automobile through complicated maneuvers on a curvy mountain road “with our mind on something else.”Footnote ⁴⁰ In such cases we are conscious of our behaviors and the mental states that guide them, in a nonexperiential sense of “conscious.” What we mean in such cases is dispositional: we can become aware experientially if the need arises, and we can state what we are doing and why if the need arises; this is because we have the same noninferential way of knowing these items as we do for consciously experienced states. Such immediate, noninferential, nonobservational knowledge of the contents of our mind is usually termed a kind of “privileged access.”

This attribute of our minds—our epistemic advantage in the way we each know our own mind—is often confused with two stronger claims.Footnote ⁴¹ These are the claims of incorrigibility and of transparency. We would be incorrigible in our beliefs about the contents of our own minds if we could not be wrong about them, so that, for example, if I believe I intend to help another, then necessarily I do indeed intend to help another. Transparency is the converse of the incorribility claim. Our mental states would be transparent to us if their existence necessarily generated a belief in their existence by us, so that, for example, if I intend to help another, then necessarily I know that I intend to help another.

In their criticism of folk psychology cognitive scientists and neuroscientists often attribute to folk psychology these much stronger theses. But in this they are unjustified. The folk psychology on which law and morality are built only supposes a privileged access in its weakest sense (of noninferential way of knowing).Footnote ⁴² It is not committed to the kind of omniscience about our own minds that would be true if the incorrigibility and transparency theses were true.

These eight characteristics I take to be the fundamental features of ourselves constituting us as commonsense selves, persons, moral agents, and legal subjects. Often in the scientific literature critical of folk psychology, a number of other characteristics are assigned to it, unsurprisingly making it more vulnerable to criticism. Two such characteristics are particularly salient.

9. Free will

Many cognitive scientists, joined by not a few lawyers, assume that desert-based political theories such as retributivism are hostage to the supposition that our wills are free in the sense of uncaused. The idea is that only if we are free in this sense do we have any real options, any real choice. Often this idea is conjoined with the autonomy claim (number 6 above) by saying that we need not only to be uncaused, but uncaused causers. Aquinas's prima causa is the idea. But it is important to separate these two claims. That our choices cause voluntary actions is one claim (the autonomy claim); that such choices themselves are uncaused is another, quite distinct claim.

Stephen Morse has shown quite persuasively that nothing in the criminal law as stated by its doctrines requires a finding that the accused had free will.Footnote ⁴³ If a rational agent performs a voluntary act that in fact and proximately causes a legally prohibited result, and if he does so intending, foreseeing, or unreasonably risking that result in circumstances where he had adequate capacity and opportunity not to do so, then he is responsible and punishable under the general doctrines of Anglo-American criminal law.Footnote ⁴⁴ Morse is right that formally nothing here is said about free will.

Yet those who think that free will is a supposition of both moral responsibility and of desert-based legal institutions would locate that supposition beneath and behind such legal doctrines, not in them. The idea is that no act can truly be voluntary, no intention or belief can cause such acts, no acts can be those of a rational agent, no capacity or opportunity to do otherwise can exist, unless the actor possesses free will. The supposition, in other words, is thought to be a hidden, metaphysical supposition required by morality, not an on-the-surface requirement of explicit legal doctrine. As Josh Greene and Gerald Cohen see it:

10. Cartesian dualism

The free will supposition tends to require an even more extraordinary supposition,Footnote ⁴⁶ which is that our mental states exist in an entirely different plane of existence than that in which physical states (including brain states) exist. This is the supposition of metaphysical dualism. Because Descartes proposed the best known form of this doctrine, the supposition most commonly is that of a Cartesian kind of metaphysical dualism.Footnote ⁴⁷

It is not difficult to see from where the dualist temptation arises. Supposing our wills to be free from causation requires some account of how that is possible. Cartesian dualism supplies such an account: mental states, existing in time but not in space, and having no mass and no energy, are not subject to the causal influences of items (like brain states) that are located in space as well as in time and do possess both mass and energy.

This of course is to leave the frying pan for the fire, in light of the stunning implausibility of all forms of Cartesian dualism.Footnote ⁴⁸ A long tradition in philosophy has argued that one can avoid both the fire and the frying pan by not supposing that we have free will.Footnote ⁴⁹ But again, my point here is more preliminary: it is to see that folk psychology is often stuck with a separate presupposition about who we are that, if made, would make that psychology (and anything built upon it) very vulnerable to criticism indeed.

We now need to turn to the challenges offered up by contemporary neuroscience to this view of who we are. As we shall see, neuroscientists tend to reject Cartesian dualism (point 10), free will (point 9), and all forms of privileged access (point 8). Some neuroscientists also reject the autonomy thesis (point 6) and that part of the consciousness thesis (point 7) that ascribes causal efficacy to our willings or to our consciousness of those willings. In addition, there is a sprinkling of criticisms in cognitive science that deny the Intentionality thesis and the thesis of the tripartite soul (points 1 and 2), the coherence over time of character (point 5),Footnote ⁵⁰ and various other aspects of folk psychology. What we now need to see is how all such challenges are mounted.

1. Reductionism

The greatest challenge(s) presented by neuroscience seems (judging at least from the popular literature) to stem from the anticipated reduction of mental states, and the actions they produce, to physical states of the human body. That each mental state of belief, desire, intention, and emotion is in fact a brain state, and that each voluntary human action is at base no more than a bodily movement caused in a certain way by these mechanistic brain states, seems to challenge many people's conception of themselves as responsible agents.

Why such data seems challenging I will describe in subpart B below. But here I shall describe more of what the data is that seems to ground such challenges. Let me start with the idea of reductionism itself: Reductionism comes in different flavors. An older form of reductionism was to conceive of the reducibility of one discourse to another in terms of the synonymy of the terms used in each of the two (putatively) different discourses. Thus, talk about bachelors can be reduced to talk about unmarried male persons because “bachelor” and “unmarried male person” mean the same thing. Let us call this analytical reductionism, because the connecting statement (“All bachelors are unmarried male persons”) is thought to be an analytic truth, namely, true by virtue of the meaning of the words alone.Footnote ⁵²

Less ambitious than analytic reductionism is metaphysical (or “ontological”) reductionism.Footnote ⁵³ The crucial relation here is (numerical) identity, not synonymy. Even if “Evening Star,” “Morning Star,” and “Venus” do not mean the same thing (in some sense of meaning), still one can be reduced to the other in the sense that these are all but different ways of referring to the same thing (namely, the planet Venus). Likewise, even if “water” and “H₂0” do not mean the same thing, each can be reduced to the other in that these two expressions refer to the same natural kind of thing.

Analytical and metaphysical reductions permit some form of substitution of one term for another. In the case of analytic reductions the substitution of one term for another with which it is synonymous cannot change the meaning of the sentences in which such terms appear. In the case of metaphysical reductions, the substitution of one term for another cannot change the truth value of the sentence in which such terms appear.

Thirdly, there are what are usually called nomological reductions.Footnote ⁵⁴ Here the key relationship is neither synonymy between words nor identity between the things referred to by such words; rather, it is the extensional equivalence between statements using the different words. Thus, if “x is M if and only if x is P” is true, then the extensions (the class of things of which a predicate is true) of M and P are the same. This can be true even if the properties, M, P, are not identical, and even if the words, “M” and “P” are not synonymous. Still, such extensional equivalences allow one to substitute “P” for “M,” and vice-versa, and still preserve truth. Laws in terms of M predicates, thus, may be replaced with laws in terms of P predicates. Thus the name: nomological reductionism.

Usually those who have talked of the “reduction of mind to brain” have meant a metaphysical reduction: mental states just are brain states. Such reductionists may or may not expect or hope for a nomological reduction as well—that depends on one's view of the independence of laws when the identity of mind to brain is of the “multiple realizability” kind. Some versions of this last view hold that the best laws (and thus the best explanations in terms of such laws) may be framed in mental state terms, even if mental states are “nothing but” (a complex disjunction of) brain states.Footnote ⁵⁵

On the merits, there are two identities we should consider. The first is the identity of actions with the movements of the human machine. It is a long-held thesis (originating in the philosophy of action rather than neuroscience, but no matter)Footnote ⁵⁶ that all the diverse kinds of actions we do—such as signaling our allies, deceiving a friend, checkmating an opponent in chess, etc.—are no more than movements of the human body. The thesis is that each particular action that we do (“act-token”) is identical or partially identical to some discrete bodily movement. Such movements become instances of types of actions (“act-types”) like signaling, deceiving, and checkmating, when those movements take place in certain circumstances and/or produce certain consequences in the external world.

Actions are thus at base physical things, specifically movements of the human body. To understand the nature of such movements is thus to understand human actions. Cognitive science for the past century has contributed much to our understanding of the mechanics of such movements. An organized sketch of such an understanding should start with the muscles and work upwards to higher functions.Footnote ⁵⁷ The immediate cause of voluntary movement lies in the muscles themselves. On the limbs these are arranged in flexor/extensor pairs, the flexor muscles bringing limbs closer to the body and the extensor muscles drawing limbs away. Early on the well-known psysiologist Sir Charles Sherrington charted how these pairs are subject to the principle of reciprocal innervation.Footnote ⁵⁸ This means that as a flexor muscle contracts its paired extensor muscle is inhibited from contracting, and vice versa. Such coordination prevents deadlock and allows movement to take place.

The muscles themselves are stimulated or inhibited by pairs of efferent neurons ending in the motor end plates attached to the muscles. These efferent neurons are matched by pairs of afferent neurons that transmit from muscle back to brain (rather than from brain to muscle, the direction of the efferent neurons). These afferent neurons are the feedback communication channels giving the brain information about how the muscles are performing.

What happens above the efferent neurons attached to the motor end plates depends on what kind of bodily movement is involved. The shortest story is that about the spinal reflexes, so called because the origin of these movements never gets higher than the spine itself. Some spinal reflex circuits, like the monosynaptic stretch reflex exemplified by the knee jerk (patellar) reflex, contain but a single synapse, which connects directly onto the efferent neurons that stimulate/inhibit the relevant muscle pairs. This is what makes such reactions so quick (about fifty milliseconds). Other reflexes, like the pain withdrawal reflex, use a reflex arc passing through multiple synapses. Even so, such reflexes also do not leave the spinal cord, processing afferent signals and sending out efferent signals to the flexor muscles without going higher.

The spinal cord and the reflexes working through it are part of the lower (or “common,” or “final”) motor pathway. All human movement is generated by signals in this pathway, consisting anatomically of the spinal cord, brain stem, pons, and the axon fibrils extending to motor end plates on muscle fibers. Aside from the reflexes, efferent signals in the lower motor pathway are caused by signals from one of the two upper motor pathways. It has long been known that these two upper motor pathways are functionally as well as structurally differentiated. The phylogenetically oldest of these upper motor pathways, the extrapyramidal, handles motor reactions that are spontaneous, unplanned, or reactive to cues; such as laughing at a joke, reacting to a paper bag popped behind one's head, or smiling spontaneously at a pleasant face. These reactions are not spinal reflexes, because areas of the brain above the brain stem and the pons are involved. Structurally, most of the extra-pyramidal system is located in the basal ganglia.

It is the pyramidal upper motor pathway that handles most of what we think of as voluntary motor movement. This is the system used when we execute more general intentions after planning and deliberation about them. Structurally it consists of the motor strip of the primary motor cortex, which is functionally divided between the different body parts being controlled. Thus, there is an area of the strip to control the movement of thumbs, fingers, elbows, wrists, neck, brow, eyelids, lips, tongue, toes, ankles, knees, etc.

Above the primary motor cortex things get sketchier because they are less well-known. It seems well established that the supplementary motor area (SMA) and pre-supplementary motor area (pre-SMA) are activated in voluntary movement,Footnote ⁵⁹ and that this takes place before activation of the primary motor cortex.Footnote ⁶⁰ SMA activation is not a perfect indicator of voluntariness of motor movement, however. Such activation takes place for stray, unconscious finger movements,Footnote ⁶¹ and for movements engaged in due to hypnosis.Footnote ⁶² It also occurs when transcranial magnetic stimulation artificially causes movement through the pyramidal upper motor pathway.Footnote ⁶³

Things are much more speculative for the reasoning, goals, and general or distal intentions that in commonsense psychology cause volitionally caused bodily movement. Areas within the parietal cortex and the frontopolar cortex have been provisionally linked to these higher functions.Footnote ⁶⁴

The picture of human action that emerges from all of this is a picture of a complicated well-organized machine, the human body. The neuroscientific picture of the mental states of more general intention, belief, desire, and emotion that cause human action, is similarly mechanical and machine-like. This is the second identity we need to consider, that of mental states like intentions being identical with certain brain events. Locating the areas of the brain that oxygenated blood flows reveal to be activated when we form a certain intention is a start in identifying such intentions with some brain states.Footnote ⁶⁵ But it is only a start, and a modest one at that. For left untouched by such area mappings is the basic question of what the mechanisms are by which such areas of the brain produce such mental states. How does a decision (choice, intention) about what to eat for lunch arise out of the mechanical equipment that is the brain—a complicated set of two-valued switches going off? How does the pattern of firing of such two-valued switches in the brain give rise to mental phenomenon, with its Intentionality,Footnote ⁶⁶ its privileged access, and its subjective experience?Footnote ⁶⁷

Contemporary neuroscience cannot answer such questions at present. But there is nothing that suggests that, in principle, it will not be able to do so in the not so distant future. Indeed, my colleague Mike Gazzaniga guesses that 98 to 99 percent of neuroscientists believe that this possibility will become a reality.Footnote ⁶⁸ I thus attribute to neuroscience reductionist “data” that is at present only a speculative possibility.Footnote ⁶⁹ I do this because it is this possibility that gives rise to the broader challenges that many people feel neuroscience presents to our conception of ourselves. This possibility is nicely illustrated by this imagined bit of neuroscience by Greene and Cohen:

Greene and Cohen hypothesize that many people would find such a reduction of decisions to a mechanical process of neuronal firings (the content of the decision being dictated by sheer numbers of firings of one kind of neuron over another) to challenge their sense of themselves as responsible agents. In this bit of speculative sociology about popular belief, I think they are right, for reasons I shall explore in subpart 2 below.

2. The Data of the Libet and Libet-Inspired Experiments

If we leave the globally reductionist ambitions of neuroscience for more concrete data actually collected, there are two sets of data that seem particularly challenging to folk psychology and to the moral and legal institutions built on that psychology.Footnote ⁷¹ The first of these stems from the experimental work of Benjamin Libet.Footnote ⁷² Libet's initial paper has been touted as “one of the most philosophically challenging papers in modern scientific psychology.”Footnote ⁷³ That paper, we are also told, reported a “groundbreaking experiment” that “forced upon the scientific and philosophical community” fundamental questions of free will, conscious efficacy, and the like.Footnote ⁷⁴

Libet built on work by two German neuroscientists in the early to mid-60's, who discovered that there was a slow negative shift in electrical potential in the brain of subjects who performed spontaneous (unplanned and not externally stimulated) voluntary movements of their fingers.Footnote ⁷⁵ Using the technology of the day, electroencephalograms (EEG), Kornhuber and Deecke discovered that approximately one second before such movements took place this slow negative shift began, a shift they termed the “Bereitschaftspotential,” and that has since been called in English the “Readiness Potential” (RP). The RP was detected by electrodes attached at the vertex of the scalp, and it evidences events taking place in the Supplementary Motor Area of the brain. Such SMA activity involved in voluntary motor movement has been shown by subsequent functional magnetic resonance imaging (“fMRI”) and other techniques to precede the activation of the primary motor cortex (as well as the activation of motor pathways, motor plates, and muscles).Footnote ⁷⁶

Libet's innovation was to conjoin his longstanding interest in consciousness to this data about the brain area involved in voluntary motor movement. In Libet's best known experiments his subjects were told to flex their right wrists or the fingers of their right hands whenever they wished. EEG readings were taken from the subject's scalps over the relevant portions of their brain, the SMA. Such readings detected the negative shift, or RP, beginning about 550 milliseconds prior to the time at which their muscles began to move so as to flex their wrists or move their fingers. These subjects were also told to watch a spot revolving on a spatial clock and to register when their initial awareness of intending to move their wrist/fingers began. They reported the beginnings of awareness of their decision to move their finger or wrist on average 350 milliseconds after the shift in readiness potential began. Such beginning of awareness preceded the beginning of actual movement by 200 milliseconds.

We can represent these results along a simple schema that I will use throughout the remainder of this article. Let “t₁” be the time of the onset of the shift in readiness potential; “t₂,” the time of initial awareness of an intention to move, and “t₃,” the time of the beginning of the bodily movement in question. The basic schema is represented below:

The total time from t₁ to t₃ is 550 milliseconds, t₂ being located 350 milliseconds after t₁ and 200 milliseconds before t₃. Libet added two more times of some interest here. One is Libet's corrected time of the onset of awareness; finding from other studies an error of about 50 milliseconds in subjects' estimation of the onset of awareness of sensations generally, Libet corrected the temporal location of t₂ to be 50 milliseconds closer to t₃, and 50 milliseconds further from t₁, than the subjects reported. Secondly, Libet notes that (what I shall call) the “Rubicon Point”—the point beyond which the subject cannot stop the movement already decided upon—occurs 50 milliseconds prior to actual muscle movement initiation (t₃) and 100 milliseconds after the corrected time for onset of awareness of the decision to move (t₂).

The values above for t₁, t₂, and t₃ are given for what Libet called “Type II RP's.” These are readiness potentials and awareness onsets measured when subjects both were instructed to move whenever they wished and when they reported (in post movement debriefings) no experience of any advanced pre-planning to move. Shifts in readiness potential measured when there was such experience of pre-planning, or where subjects were not free to move spontaneously (i.e., where they moved in response to a pre-set signal), Libet termed “Type I RP's.” Type I RP's had considerably earlier onsets than did Type II RP's (about 1050 milliseconds prior to movement).

In what follows I shall focus on Type II RP's, as did Libet. For Libet thought “that the RP component that starts at about—550 milliseconds, the one that predominates in Type II RP's, … is the one uniquely associated with an exclusively endogenous volitional process.”Footnote ⁷⁷ It is this data that thus forms the core of the neuroscientific challenge(s) to the presuppositions about intentions made by the law and morality in their responsibility assessments.

Libet's results have been replicated in subsequent experiments.Footnote ⁷⁸ Considerable critical commentary and research attention have been focused on the accuracy of persons' perceptions of when awareness of their decision to move begins.Footnote ⁷⁹ But however one construes the degree of inaccuracy here, or their skewing effects by various influences, Libet's main result seems secure: brain processes initiating voluntary motor movement began before reported awareness of any decision to move.

Libet's result has been extended to a variety of other circumstances. Haggard and Elm found the RP pattern to exist when subjects were given the freedom to use either their left or their right hand, thus involving more choice than simply whether to move a given finger or a given wrist.Footnote ⁸⁰ In such cases a lateralized RP (“LRP”) was discovered later than the onset of RP, indicating when one hand's movement rather than another has been specifically selected. Movements taking place when greater freedom of movement is open to subjects have also shown the RP pattern, as have experiments giving greater freedom as to when a movement takes place.Footnote ⁸¹ Movements done under hypnosisFootnote ⁸² and movements induced by transcranial magnetic stimulation,Footnote ⁸³ have also been studied, showing activation of the same brain areas.

The time before awareness of decision by the subject has also been considerably extended. Lau has found distinctively movement-initiating brain activity (in the pre-SMA) two to three seconds prior to motor movement, much earlier that the .5 to 1.5 seconds typical for the RP pattern in the SMA.Footnote ⁸⁴ John-Dylan Hayne's even more recent studyFootnote ⁸⁵ isolated activity in areas of the frontopolar cortex and the parietal cortex, which activities were predictive of which finger the subject would move, left or right index finger, up to seven to ten seconds prior to movement (and up to six to nine seconds prior to awareness of the decision to move one finger or the other). One of Haynes's other studies found a similar pattern of brain activity when the choice was about adding or subtracting—a purely mental operation involving no motor movement by the subject.Footnote ⁸⁶ This suggests his measurements are of a more distal intention or decision, as opposed to the motor-related and immediate “act now” decision studied by Libet.

3. The data of the Wegner and Wegner—Inspired Experiments

The third major type of specific psychological literature taken to be challenging of our actions of responsible agency is that associated with the work of Daniel Wegner. Wegner's work should be seen as part of a larger set of data dealing with the degree to which we are accurate in judging the causes of our own behavior. Nisbett and Wilson much earlier explored a fact widely recognized by good novelists long before Freud, namely, that even our most sincere judgments of what moves us to action (that is, our motives) are often mistaken.Footnote ⁸⁷ John Bargh's more recent work has shown in convincing detail how many unconscious influences there are on what we (consciously) decide and do,Footnote ⁸⁸ influences of which we are not aware (and which we disavow when others bring them to our attention). For example, in reaction time experiments, subjects are found to make faster approach movements of their arm (such as pulling a lever toward oneself) if they have, however unconsciously, positively evaluated the object named on the screen, and they are found to make faster avoidance movements (such as pushing the lever away) if they have, however unconsciously, negatively evaluated the object on the screen.Footnote ⁸⁹ Bargh extends this to explain everyday behavior. He explains some of voting behavior, for example, in terms of immediate visceral reactions to one second showings of the facial appearance of candidates for election.Footnote ⁹⁰

Despite these antecedents, the best known of this literature is Daniel Wegner's misleadingly titled book, The Illusion of Conscious Will. Footnote ⁹¹ This sounds like a book producing evidence that the will is caused by unwilled brain events, and/or that the will does not in fact cause the bodily movements it seems to cause. Yet the data produced are not of this sort and in fact support quite different theses.Footnote ⁹² The book's theses are epistemic: we do not have direct access to the causing of bodily movements by our consciously willing them, and as a result, we are (perhaps sometimes, often, or always) wrong in our judgments of whether our willings have caused the bodily movements they seem to cause. As Wegner puts his thesis, “The illusion of conscious will is the belief that we are intrinsically informed of how our minds cause our actions by the fact that we have an experience of causation that occurs in our minds.”Footnote ⁹³ Or more crisply: “The experience of conscious will … is not direct evidence of a causal relation between thought and action.”Footnote ⁹⁴ Rather than our knowledge of our own minds being a direct, non-inferential privileged access, Wegner holds that “conscious will is an experience that arises from the interpretation of cues to cognitive causality” and that this is “an interpretive process that is fundamentally separate from the mechanistic process of real mental causation.”Footnote ⁹⁵

One implication of the fact that our knowledge is inferential and “not a direct readout of some psychological force that causes action from inside the head,”Footnote ⁹⁶ is that we are fallible knowers of when we acted. Wegner assembles experiments and historical anecdotes both ways: both where we think we have not acted but in reality we have; and where we think we have acted but in reality we have not. He, thus, denies the transparency and incorrigibility claims of strong privileged access.

Wegner calls instances of the first kind “automatisms.” These include: the alien hand syndrome (think Dr. Strangelove), where “a person experiences one hand as operating with a mind of its own”;Footnote ⁹⁷ movements during hypnosis, where sometimes subjects report that “they felt no sense of moving their arm voluntarily but rather experienced the downward movement as something happening to them”;Footnote ⁹⁸ the table turning done in nineteenth-century séances where a group sitting at a table for an extended period experiences the table as rotating on its own when, in fact, by minute movements people in the group are collectively moving it;Footnote ⁹⁹ the modern descendant of this, the Ouija Board, where a group may produce an intelligible message with each member of the group experiencing a “sense of involuntariness” that can be “quite stunning”;Footnote ¹⁰⁰ automatic writing, which despite a lack of consciousness about the content being produced or despite a lack of experience of willing the movements involved, can produce intelligent prose;Footnote ¹⁰¹ the Chevreul pendulum which moves when the subject wishes not to move it and which can even move in directions the subject wants not to move it;Footnote ¹⁰² movements of dowsing rods, which can be experienced as involuntary even though clearly responsive to knowledge of the moving subject.Footnote ¹⁰³

Wegner also spends considerable time detailing the old (William James, e.g.) and new (John Bargh, e.g.) experiments on ideomotor actions. These are movements guided by thoughts without any apparent mediation by will between thought and action. Absent-minded munching during social conversations, tilting in response to imagining falling over, unconscious mimicry of others' acts, are examples.Footnote ¹⁰⁴ Wegner's conclusion about all such behaviors is that “thinking about acting can produce movement quite without the feeling of doing,”Footnote ¹⁰⁵ and that “these actions seem to roll off in a way that skips intention.”Footnote ¹⁰⁶

In a way, Wegner's focus on automatisms is curious. If one wanted to show that sometimes we act voluntarily without knowing that we do so, these would not be the sort of examples that would come to mind. For although Wegner dubs them “unconscious actions,”Footnote ¹⁰⁷ few of them would be considered the kind of voluntary movements with which we identify human action. Freud's older examples of unconscious actions—certain dreams, parapraxes, and neurotic symptoms—seem to make the point better, because Freud's distinctive claim here was that the subject could recapture (in principle, at least) the act-ish phenomenal feel that was lacking when the actor acted.Footnote ¹⁰⁸

Yet Wegner's target here is not so much the transparency thesis—if we are acting, we know that we are. Rather, what he wants to infer from examples of automatism is that there is no causal connection between what really causes our behavior, on the one hand, and our experience of causing our behavior, on the other. As Wegner puts it, “automatism and ideomotor action may be windows on true mental causation as it occurs without apparent mental causation.”Footnote ¹⁰⁹ The suggestion is not like Freud's; Wegner is not out to expand the category of behaviors we should consider actions, as was Freud. Rather, he is out to establish that we in general have no privileged access to the real workings that produce our voluntary behavior, because the source of that access—our experience of will—has no connection to those real workings.

The trick for Wegner is moving from the somewhat recherché phenomena of automatisms to his desired truth of general psychology. The move for him is from the observed fact that in the automatisms, thought controls action without the mediation of will or intention. The general truth which he wishes to infer from this is that this is no less true for ordinary voluntary actions.Footnote ¹¹⁰ These too, he urges, are produced by the “empirical will” unmediated by the workings of the conscious will that we experience, no matter how much our experience of that conscious will inclines us to believe otherwise.

The other kind of phenomena that are the focus of Wegner's attentions are those where there is an experience of willing the behavior but that behavior does not in fact occur. Wegner calls these cases of the “illusion of control.”Footnote ¹¹¹ Examples from everyday life include “body English” executed on erring golf shots, the sense in gambling that we had some control of dice, coins, or roulette balls,Footnote ¹¹² perceived “sports jinxes” where one thinks one's mere presence at a sporting event affects the outcome of that event, and the like.Footnote ¹¹³ More persuasive, perhaps, are the data of more scientific studies and experiments. One of these is Wegner's own, the “I spy” study. Here the subject infers that she has selected items on a screen showing a group of items taken from the familiar child's game, “I Spy”; in reality, a confederate moved the cursor. The illusion of control in the subject was created by priming the thought of the item selected just before the item was selected on the screen. The illusion was measured by the subject's self-report of intending and allowing the movements.Footnote ¹¹⁴

Other “illusion of control” cases include dramatic instances of confabulation, where the subject makes up intentions he could not possibly have had. These occur in both split-brain patients and in those making movements in response to post-hypnotic suggestions. “Split-brain” subjects are those whose connecting tissue between the two hemispheres of their brain has been severed or damaged. In experimental conditions with such persons in which the input and output functions of each hemisphere are separated (so the left hand literally does not know what the right hand is doing), subjects nonetheless confabulate intentions to explain behavior to which really they must be clueless.Footnote ¹¹⁵ Similar confabulation has long been studied by subjects who move their limbs in response to post-hypnotic suggestions: people invent intentions they couldn't possibly have had in order to explain why they did what they did.Footnote ¹¹⁶

About all such cases of the “illusion of control,” Wegner concludes that “we are not infallible sources of knowledge about our own actions and can be duped into thinking we did things when events conspire to make us feel responsible.”Footnote ¹¹⁷

1. The eliminativist challenge

Start with the (at present largely imagined) data of reductionism. One challenge directly spawned by this “data” is what I shall call the eliminativist challenge. This challenge is so named because the reduction of minds to brains, and of actions to physical movements, is taken by some to eliminate the self (or the “I”) of conscious experience and agency. Consider this lament by Francis Crick:

The challenge of reductionism is that we are said to lose our mental life and our agency—indeed, our very selves—if we are “nothing but” machines. Holding us responsible, or attributing to us subjective experience, would make no more sense than it would to do so for other, more obvious machines, namely, robots, automobiles, or lunar landers. As Richard Dawkins once said, punishing us (if reductionism is true) would make as much sense as John Cleese's famous scene in Faulty Towers, where Cleese punishes his automobile for its poor performance by beating it with a sledge hammer.Footnote ¹¹⁹

Despite its popularity, inferring this challenge from the reductionist ambitions of neuroscience is puzzling. One can grasp just how puzzling this is by comparing the present eliminativist challenge to the metaphysical position known as “eliminative materialism” in philosophy.Footnote ¹²⁰ The eliminative materialist also believes that we should jettison talk of minds, selves, agency, and all that is built upon these (such as moral blameworthiness); yet she derives this skeptical conclusion from a nonreductionist premise. It is precisely because we cannot reduce intentions to brain states, minds to brains, that we should dispense with these ghosts. It is because mental states (given their Intentionality) cannot be reduced to brain states that they can have no place in (what will turn out to be) the best explanation of human behavior, an explanation, in terms of brain states. In Greene and Cohen's imagined string of neural firings determining a decision of salad over soup for lunch, for example, there is no room for persons, selves, or moral agents, on nonreductionist premises. We—our agency—would have to be something extra, a ghostly commander leading the blue neurons (for salad with blue cheese dressing?) to victory over their red (for tomato soup?) competitors. And this is silly—it would make the soup versus salad decision at the neural level like the battle scenes in Kurosawa's film, Kagemusha, with all the roles in the scenes filled before selves enter the stage.

If, however, we reject the premise of philosophy's eliminative materialism—if in other words we embrace reductionism—this will have the potential to get us on the stage. Then, we, our intentions, agency, and the like, will be identical to some swatch of what is happening on that stage at the neural level; so if it is on stage, then so are “we.”

Thus the puzzle: why is reductionism a challenge to all we value? How can it threaten us with loss of our selves and our responsibility when it is nonreductionism that seems to threaten us with that kind of loss? We have faced this kind of puzzle before. Recall that a generation or more ago, Fred Skinner propounded two bases for moving “beyond freedom and dignity,” beyond the “autonomous man,” beyond the inner life of the mind.Footnote ¹²¹ One came to be called logical behaviorism, and it was reductionist: mental states like intentions just were logical constructs out of behavior, not inner causes of that behavior. The other came to be called methodological behaviorism, and it was nonreductionist: since mental states were not publicly observable in replicable experiments as was behavior itself, there was no place for these distinctively non-behavioristic, inner, private episodes in a properly scientific psychology.

A younger Dan Dennett once excoriated Skinner, declaring it “unfathomable” how Skinner could be so sloppy as to vacillate on such a watershed issue.Footnote ¹²² Yet if we look carefully at the kind of reduction Skinner's logical behaviorism was proposing, perhaps we can see how his reductionist logical behaviorism was as skeptical about mental states as was his nonreductionist, methodological behaviorism. This will not mean that one would not have to choose between these two kinds of skepticism—after all, the admission ticket to rational discussion of anything is not knowingly embracing a contradiction (like affirming and denying reductionism). Yet perhaps we can see how both rejection and acceptance of reductionism has (inconsistently, to be sure) made some people skeptical about minds, selves, and responsibility.

The skepticism of methodological behaviorism was easy to see. Explanations framed in terms of those methodologically inferior, private, nonobservable, nonreplicable things Skinner thought mental states to be, would inevitably lose out to explanations framed in terms of scientifically respectable entities like environmental stimuli and behavioral responses. Today's eliminative materialism is the direct descendent of methodological behaviorism, except that what makes mental state explanations inferior now includes Intentionality and what is better is brain state (rather than behaviorist) explanations. The skepticism of both is easily intelligible, however, and is of the same kind: physicalist explanations of behavior (of either kind) will be complete, with no need or room for mental state explanations; the latter explanations will lose out in any explanatory competition; and the entities and qualities such mental state explanations posit will disappear from our commitments “when our thoughts turn seriously ontological.”Footnote ¹²³

The skepticism of logical behaviorism is more difficult to see. Logical behaviorism is an instance of what I call a skeptical reduction.Footnote ¹²⁴ Saul Kripke's example of this is Hume on causation.Footnote ¹²⁵ Hume famously identified the causal relation with the regular concurrence of events. For this reduction Humeans are rightly regarded as skeptics about causation, because that reduction leaves out what many thought distinctive of that relation, what Hume called the “cement” of the universe.

Logical behaviorism was a skeptical reduction in the same way, and so, for many people, is the neuroscientific reduction of mental states to brain states. For seemingly left behind by both is all that we prize in the life of the mind, all that (we thought) makes us different than robots and other “mere machines.” Such reductions, whether behaviorist or neuroscientific, make many people skeptical because such reductions classify us differently than we previously thought. We thought we were special, and reductionism seemingly shows that we are not. We too are just machines.

2. The lack-of-free-will challenges

Apart from the threatened dissolution of self, a second and distinct challenge arises from reductionism, albeit as an indirect inference. This is the familiar challenge that we lack free will. This challenge, of course, has been with us for a very long time, well antedating the rise of neuroscience. In religious form the challenge was presented to Calvinists by their own doctrine of “predestination.” According to this doctrine, God had already picked those who would enjoy the Christian Heaven before they were born—in which case, the challenge was, why make the effort to do good works? It was the rise of materialism with Hobbes that gave this challenge its distinctively modern form, a form in which mechanical processes replaced God as the cause of all we do.

Despite the fact that this challenge has been with us for a very long time, reductionist neuroscience presents the lack of free will challenge to responsible agency in a particularly dramatic and hard-to-ignore way. This is partly due to the immediacy of the causes presented by neuroscience: unlike genetic and environmental influences, brain states are much closer (spatiotemporally) to the behaviors they cause. Hence, it looks like there is less room for extraneous factors, or even chance, to intervene. “Brain states don't influence behavior; they fully cause it,” is the inference sometimes drawn from this spatiotemporal closeness. This sense of sufficient causation is enhanced by another feature of neuroscientific explanations. This is that they are complete. All environmental and generic influences have to do their causal work through the human brain (the behaviorists' mistake was to think otherwise); Greene and Cohen nicely call this the “bottleneck” nature of neuroscientific explanations.Footnote ¹²⁶ The result is, again, an enhanced sense of inevitability, here arising from neuroscientific explanations that by their nature are sufficient (during their “bottleneck” time-slice) to produce behavior.

The enhanced challenge presented by neuroscience is also partly due to the greater knowledge we are coming to possess about the causes of behavior that neuroscience posits. It is one thing to speculate in the abstract about there being causes of human choice; it is another to be told that certain patterns of firing by certain neurons in certain areas of the brain fully determine your choice for salad at lunch.

The concreteness of the causes that are the subject of neuroscientific explanation joins the immediacy, the completeness, and the better-known nature of such causes, in explaining why neuroscientific causation seems more challenging to responsible agency than is determinism generally. For neuroscience is not committed to a general determinism, one that denies that there are any uncaused causers in the universe; it can be agnostic about whether God, the Big Bang, energy level shifts by electrons, or some other form of Aquinas's prima causa, has ever existed. More importantly, neuroscience can be indifferent regarding whether determinism in general is true, either at the micro level of modern physics or at the macro level of everyday life. For the challenge presented by brain-state causation is independent of whether universal causation (determinism) is true. Even if the micro level (of energy shifts in electrons, e.g.) is irreducibly probabilistic, making the macro level the same but with much higher probabilities, that won't blunt the challenge to responsible agency presented by brain state explanations.

To see that this is so, let us isolate that challenge. There are, in fact, two versions of this lack-of-free-will challenge to responsible agency which need to be dealt with separately. The first, more radical version denies that choice is even possible if choice is caused by factors themselves unchosen. To show that “choice” is caused, on this version, is to show that choice does not exist. I call this the analytical version of the lack of free will challenge. For the plausibility of the challenge seemingly presupposes that it is analytic of the word, “choice,” that the referent of the word be uncaused. Choices on this view differ markedly from most natural phenomena with which we are familiar. Take lakes for example. Imagine a claim that something could be a lake only if it was uncaused (no glaciation could explain Lake Michigan, for example). Since this is wildly implausible about lakes, choices must be very different from lakes in order for the analytic version of the lack of free will challenge to be very plausible.

The second, more moderate version, admits that there can be choices even if they are caused by factors themselves unchosen. Yet, this version asserts, there cannot be responsible choices in such a case. Being morally responsible and legally punishable for our choices is incompatible with causation of choice, even if choice itself is not. The nerve of this second thought is that if choice is caused by factors themselves unchosen, then we have no (real) choice, no (real) options to choose between, because in such a case we could not have acted other than we did. While our choices can still cause the behaviors chosen, the choices themselves constitute no (real) control over those behaviors because there is no control over the unchosen factors causing choice.

Neither of these challenges depend on determinism being true. Even if there were an irreducible randomness in the determination of human choice (because there were such randomness in the micro-physical processes constituting such choice), that would hardly give us more control, more options, more real choice. Such randomness is not more subject to our control than is a full causal determination. The probabilities of us choosing tomato soup for lunch would be completely fixed by factors over which we have no control, even if actually choosing tomato soup would not be completely fixed. Such randomness gives us no window of opportunity in which to effect our control. It is as dumb, as impervious to our efforts, as is full causal determination. Such randomness cannot frame a choice set giving us real options to choose between. It is nothing like the framing of real options done, for example, by the existence of both salad and soup for lunch.Footnote ¹²⁷

The upshot is that we can put the grand issues connected to the truth (or even the meaning) of determinism to the side. Neuroscience's concrete version of the lack of free will challenge(s) should be put as: in whatever sense and to whatever extent macro-sized natural events and natural processes are caused, so too are both human actions, and the choices that cause them, caused by brain states.

It remains to say how this challenge, like the eliminativist challenge, arises from the reductionist nature of neuroscience. How is it that for many neuroscientists, “if the self reduces to the brain … then the individual doesn't have real contra-causal free will … ?”Footnote ¹²⁸ The answer lies in a straightforward inference. According to reductionism, mental states like intention are identical to certain brain states, and human actions are identical to those motions of the human body that are caused in the right way by such intentions/brain states. There is nothing in such identity claims mandating that intentions themselves be caused; in principle they could be uncaused-causers even though they are brain states. Yet if intentions are themselves brain states that cause other physical states, it is a reasonable (if separate) thesis that they must themselves be caused by earlier brain states. Only if intentions were not brain states but existed in some non-physical way would it be plausible to think that they could be uncaused. Getting rid of Cartesian dualism, as almost all neuroscientists do, forecloses this possibility. For a respectable neuroscience there can be no ghost in the machine that is the brain, no more than there can be such ghosts in robots, automobiles, or lunar landers.

It would be a mistake to think that the lack of free will challenge stems only from the global (and still speculative) data of reductionism. For some neuroscientists believe that such lack has been experimentally demonstrated by the second set of data I described above, that developed in the Libet and post-Libet experiments. Susan Pockett, for example, prefaces her volume collecting recent literature on this issue by urging that “one of the philosophical outcomes of Libet's experiments is that they might be taken as suggesting we do not have conscious free will.”Footnote ¹²⁹ The well known London psychologist, Patrick Haggard, for one, so takes the Libet experiments:

The idea is that the readiness potential evidences certain brain states (mostly in the supplementary motor area), and that these SMA states cause intentions (choice, willings) to occur. The question of whether there is a global determinism can be put aside, for here, it is thought, we can see that the will is caused by factors themselves unwilled, and is thus not free. It is for this reason that for many neuroscientists, “Libet's experiments have stood the test of time and become the scientific spine of the anti-free will movement.”Footnote ¹³¹

Libet shares with much of popular literature the thought that having (contra-causally) free will is essential to being morally responsible and blameworthy. Libet's thought is that if brain states precede and cause one's choices, then such choices cannot ground a true responsibility. In order for choice to give one control (and, thus, responsibility), it would have to be undetermined by any brain events that are outside the control of the chooser. If there are such unchosen brain events causing us to will what we do, “the individual would not consciously control his actions; he would only become aware of an unconsciously initiated choice. He would have no direct conscious control over the nature of any preceding unconscious processes…. We do not hold people responsible for actions performed unconsciously, without the possibility of conscious control.”Footnote ¹³²

True enough, Libet himself drew back from consigning our blaming practices to the dust-heap. Yet he did this only by positing a veto function that the will can possess.Footnote ¹³³ According to Libet, the will does not initiate actions, but is the mere puppet of the brain events that cause it; yet the will has about a 100 millisecond window (the difference between corrected awareness time and the Rubicon Point) in which to exercise a veto over the movement already initiated by the SMA. By exercising such a veto, the will can prevent a movement already initiated. Remarkably, Libet held that such preventative choices are uncaused.Footnote ¹³⁴ Unlike the initiation of movements, blocking the movements is an effort of pure will. Here, Libet thought, is where free will resides (or as some wags put it, where “free won't” resides). Needless to say, many neuroscientists who otherwise admire Libet's work do not follow him here.Footnote ¹³⁵ For them, the exercise or nonexercise of a veto function is as caused by earlier brain events that precede awareness as is the exercise of the function that initiates motor movements. For them, thus, the implication of Libet's experiments is that there is no free will, full stop, with no caveats for last minute, “free” vetos. For them, thus, the Libet experiment shows that there is no responsibility or blameworthiness.

3. The challenge of a merely epiphenomenal will

It is common to elide the challenge of there being no free will with a quite distinct challenge, one that regards the will as epiphenomenal with behavior rather than being the cause of that behavior. These are in fact quite distinct challenges. The first is based on the will being caused; the second is based on the will doing no causing but is silent about whether the will is itself caused.

One can picture the difference in the two skeptical claims this way. The free will skeptic pictures the relation between brain states, willings, and bodily movements (on which he bases his skepticism) as a simple causal chain:

Figure 1.

Willings as Caused Causes of Bodily Movements

On this picture, willings do cause the bodily movements that are their objects. But the willings are themselves caused by certain brain events, and it is this aetiological feature that the free will skeptic says robs willings of the freedom allegedly needed for responsibility. Contrast this with a second picture:

Figure 2.

Willings as Epiphenomenal with Bodily Movements (simple version)

Willings are caused on this picture too, but what here grounds the skepticism about responsibility is not the existence of this causal relation but rather, the non-existence of a causal connection between willings and bodily movements. Willings and bodily movements are conceived here as being epiphenomenal of each other, co-effects of a common cause, which is why they might look like they are (even though they are not) in a causal relationship with one another.

Despite this plain difference between the two skepticisms, neuroscientists and many others have a tendency to run these claims together. Benjamin Libet, for example, often summed up what he took to be challenging from his findings this way: “If the ‘act now’ process is initiated unconsciously, then conscious free will is not doing it.”Footnote ¹³⁶ Is the challenge being expressed by Libet that the “‘act now’ process” is being caused so that there is no free will? Or is it that the will is not causing (“doing”) either that process or the bodily movement that that process causes? My interpretation is that Libet thought that both of those things were challenging to our folk psychological notions of responsible agency, but that does not make them the same.

Or consider three other psychologists' interpretation of Libet's experimental work:

Notice the clear equation of free will with the causal efficacy of the will. Notice also how that equation leads its authors to treat evidence that there are causes of conscious intention (and thus no free will) as if it were evidence that conscious intentions do not cause actions.

Or consider this description by Daniel Wegner of what it is that we suppose about ourselves that neuroscience will displace:

Which is it of which we are to be convinced by neuroscience—that we are not causers, or that we are not uncaused in our causings? Again, Wegner probably thinks both of these challenging truths are delivered up by neuroscience; and again, that is to make two claims, not one.

It is not difficult to see why neuroscientists and others often join these two claims together. The assumption behind the linking of the two claims is that we cannot be agents—choosers, “origins of our own behavior,”—if our choices are caused by factors themselves unchosen. This is the more radical, analytical version of the lack of free will claim mentioned earlier. Persons have to be free in order to be causers, even if other items (like stones, stars, lakes, and brains) do not need to be uncaused in order to be causers. If one rejects that claim, as I believe we should, then separation of the epiphenomenal challenge from the lack of free will challenge becomes not only possible but necessary.

The separation of epiphenomenal skepticism from lack-of-free-will skepticism shows that the usual evidence given in support of epiphenomenalism is not sufficient. Just because brain events both cause bodily movements and precede willings does not count as evidence against the possibility that such brain events cause bodily movements through such willings. The possibility is that the right picture is that first depicted above in figure 1, where willings are causative of bodily movements even if themselves caused by earlier brain events. The epiphenomenal skeptic thus needs to say more than that the mental events of willing are preceded by brain events that cause bodily movements. Such a skeptic may have one of two additional thoughts here.

First, he might think that the brain events at t₁ are the last events needed for the bodily movements in question. This is the idea that not only are the brain events sufficient to produce the bodily movements, but that these early brain events do not operate through any other, subsequent brain events. Once the early brain events have occurred at t₁, in other words, the bodily movements at t₃ are going to happen without need of any other events occurring, willings included. (For a complete causal chain to exist from t₁ to t₃, there must be states existing, but these are not the changes of state we think of as events.)Footnote ¹³⁹ This is depicted in figure 2 above.

Alternatively, he might think that the brain events at t₁ cause the bodily movements at t₃ via a series of other brain events. For ease of exposition, assume just one other set of brain events occurring at t₂, call them BE'. Then the picture is that depicted as:

Figure 3.

Willings as Epiphenomenal with Bodily Movements (complex version)

Willing on this picture is still epiphenomenal because BE' at t₂ is sufficient for the bodily movements at t₃, and at no point in time is there room or need for some other event, such as a “willing,” to do any causing of bodily movements.

One's choice between the two pictures depicted in figures 2 and 3 will not be settled by armchair philosophy. The issue is an empirical one, to be settled by the best neuroscience available. I nonetheless separate the two versions of this skepticism because how one answers it depends on which of these pictures one takes to be true. Notice that both of these skeptical pictures crucially depend on willing not being necessary for bodily movements to occur because the brain events at t₁, or the brain events at both t₁ and t₂, are sufficient by themselves to cause those bodily movements.

There is one construal of either of these versions of the epiphenomenal challenge that needs to be put aside on grounds of redundancy. This is the view that takes “willing” to refer exclusively to some conscious, mental experience and not to any brain state(s). Daniel Wegner for one is guilty of this move insofar as he identifies the will with what he calls the “phenomenal will.” But this interpretation is not peculiar to Wegner. It is a hidden assumption by many that makes the epiphenomenal challenge seem all but unanswerable. For if willings are to be thought of in dualistic terms, of course they are epiphenomenal to the brain states that cause bodily movements—how could such ghostly willing-experiences cause anything physical like a bodily movement? Such an interpretation of the epiphenomenal challenge makes it redundant to the physicalistic premise of the challenge of eliminativism. That premise says there are no such ghostly willing experiences, so of course they cannot be causes of anything real.

The epiphenomenal challenge only becomes interesting when it is construed to be non-redundant to the rejection of dualism. Assume the experience of willing is identical with some kind of brain state. The interesting, non-redundant epiphenomenal challenge is to assert that this experiential brain state of willing does not cause the bodily movements it seems to cause. That would be challenging indeed.

If our choices, intentions, and willings truly do not cause our voluntary actions, that challenges directly the folk psychology assumption of autonomy. Seemingly what gives us control over the world, what makes us agents who can be responsible for the results we cause, what makes us at least in part the authors of our own lives, is such autonomy, such ability to bend the world to our will. Moreso even than freedom, this seems a rock-bed requirement for us to be the kind of creatures that can be rewarded or punished by the law according to what we deserve.

4. The challenge of a merely epiphenomenal consciousness

For a long time (even antedating Freud) many have thought that one can unconsciously intend or will things.Footnote ¹⁴⁰ If so, then one response to the epiphenomenal skepticism about the will above described is to identify the brain events at t₁ as being nothing other than a willing, albeit an unconscious willing. This move preserves the causal efficacy of willings. It is this possibility that brings into focus an alternative challenge raised by the Libet experiments. That challenge is this: even if willings are the brain events at t₁ that cause bodily movements, still these are not conscious willings, and only conscious willings make one responsible, is the idea. We can only control that of which we are aware, this thought continues, so that even if the will is both free and causally efficacious, that would be without implications for responsibility because the willings that initiate movements are unconscious.

One should conceive of this as a second kind of epiphenomenal objection, pictured thusly:

Figure 4.

Consciousness as Epiphenomenal with Bodily Movements

Consciousness of willing is now pictured in the “dangler” position, an effect of the willing/brain events at t₁ but without effect itself on the movements it doesn't control. This is the third thing that is elided in Libet's statement earlier quoted, to the effect that, “If the ‘act now’ process is initiated unconsciously, then conscious free will isn't doing it.” We have seen that this is construable as both a denial that the will is free (of being caused), and as a denial that the will does any causing; now see that this statement is also construable as the claim denying that consciousness of willing does any causing.

It is easy to miss the separation of this skepticism about responsibility from the skepticisms produced by the claimed epiphenomenal status of the will. If one thinks that intentions and willings must be conscious if they are to be intentions or willings at all, then the subject can will or intend the movement of his finger only when he is conscious of that fact, which on Libet's evidence seems to be at t₂; this means that both consciousness and intending are necessarily in the same “dangler” position in the epiphenomenal chart, and so this fourth skepticism collapses into the epiphenomenal challenge already considered. For reasons I have gone into elsewhere,Footnote ¹⁴¹ I do not share the supposition that willings and intendings must be conscious. There is in that case this separable challenge to be considered.

That this second epiphenomenal thesis is a challenge to our sense of who we are as responsible agents should be clear. What is challenged is the central place occupied by consciousness in our sense of self. If consciousness is just along for the ride, enjoying the show perhaps but unable to do anything about it, that seems to strongly erode our sense of self and agency at the root of responsibility. We would be shown to be (as Freud put it about his own epiphenomenal claims about consciousness) “not even masters in our own house.”Footnote ¹⁴²

It might seem that there is yet one more epiphenomenal challenge to be considered. We can access this last possibility by speculating that consciousness itself is a brain process, just like willing. It is immensely likely that this brain process begins before phenomenal awareness is experienced by the subject—for all brain processes take real time. This requires of us an old Freudian conclusion: what is conscious can be unconscious. More exactly, the thing referred to by consciousness (i.e., the brain process underlying conscious experience) can, in its early phases, be phenomenally unconscious. With this under our belt, then suppose one identified the brain events at t₁ not just to be willings, but also to be the beginnings of consciousness of willings. Then there would be no epiphenomenal objection because consciousness of willing (in this sense) would not be in the dangler position.

This would give rise to yet a third form of the epiphenomenal objection: although conscious willings cause actions, phenomenal awareness of those willings does not. Phenomenal awareness is a mere dangler, pictured thusly:

Figure 5.

Phenomenal Awareness as Epiphenomenal with Bodily Movements

Now the charge would be that phenomenal awareness causing action is what responsibility demands, and neuroscience has shown that it does not do this causal work.

Yet notice that this challenge is illegitimate on the same redundancy ground I used before against a like use of Wegner's “phenomenal willing.” If consciousness is conceived of, not as a brain process, but as a thing in some other realm of existence, this is just dualism again. And again, there are conclusive reasons to reject any such dualism. The epiphenomenal challenge construed in this third way does no more than try to resuscitate this dead horse—at least long enough to then beat it to death again. Of course phenomenal awareness—cut off from any realization in the brain—can do no causal work in moving our bodies. Let's move on! We need to chart new challenges arising from neuroscience, not look at old ones in new verbal garb.

5. The challenge that consciousness is not veridical or privileged in its knowledge of mental states

Daniel Wegner plainly intends two challenges to arise out of his work. The first is a challenge to our infallibility in knowing the genesis of our voluntary actions. His examples are designed to show that we misjudge the voluntary status of our behaviors both ways: sometimes we misjudge that we did not perform an action when in fact we did; and sometimes we misjudge that we did perform an action when in fact we did not. The challenges here are that we are not incorrigible in our beliefs about whether or not we have acted, nor are our actions transparent to us. If folk psychology presupposed that we were infallible in either of these directions, this would be a serious challenge.

But it does not. One might read Wegner as strengthening these claims by saying that we are always wrong about whether we have exercised voluntary agency in our overt behaviors. This strong a claim would certainly challenge our sense of control, because to be typically wrong about what we are doing (or even whether we are doing anything) would leave us completely in the dark about what we are responsible for. Yet (with the caveat mentioned below) this would be too strong a claim. The truly general claim made by Wegner is different. It is that we have no direct access to causings of our bodily movements. This mode of access claim is distinct from any error claims. Like the proverbial broken clock, we could lack a reliable knowledge-producing mechanism and yet could still be correct in our readings at least some of the time.

Wegner's real challenge is this second one about privileged access. This is indeed (in contrast to Wegner's not-infallible point) a more plausibly general point: none of us, ever, has direct access to the causing of our bodily movements by our wills. Here the peripheral phenomena that he examines—automatisms, and illusion-of-control behaviors—looks more like the tip of the proverbial iceberg Wegner wants them to be. The general challenge is to the privileged access each of us is supposed to have to the contents of his own mind, his willings included. For the non-observational, non-inferential, direct knowledge we are supposed to have here, Wegner substitutes a theory-laden, inferential, interpretive mode of knowing.

This denial of privileged access is more challenging than it might seem at first blush. Such privileged access is arguably a main marker of the boundaries of self and personhood.Footnote ¹⁴³ It is what marks off a process as something a person feels, thinks, or does, from what Dan Dennett once called the “sub-personal” processes that go on in a person's brain but that are only functionally defined subroutines of various structures within his body.Footnote ¹⁴⁴ Much of the afferent-conducted feedback on our voluntary actions, for example, is not accessible to consciousness.Footnote ¹⁴⁵ These feedbacks and adjustments are subroutines a person's body has to do for him to perform well. Anyone who can read and observe can infer by theoretical inference that these subroutines must be going on in him as he stands erect, and so on. But by being unknowable in the mode of experiencing them in consciousness, these subroutines are not actions he performs for reasons. If this line between the actions we do as persons, and the subroutines our bodies do (“without us” so to speak), is eroded, that would be a challenge to our sense of selves.

Wegner's arguments and data are also often taken to suggest many of the four bolder challenges already described. That is, an “illusion of conscious will” could be taken to suggest: the disappearing self (i.e., the self that is reduced to nothing by reductionism); that there is no free will; or either or both of the epiphenomenal challenges earlier described. Indeed, Wegner himself rather flirts with and thus encourages this more challenging reading of his data. Yet the data do not support such larger readings. That the will sometimes does not cause what it seems to cause (the illusion of control cases), for example, hardly supports the metaphysical claim that wills never cause bodily movements.Footnote ¹⁴⁶ It only supports the epistemic claim that we can be mistaken in what our wills in fact cause, and (perhaps) the further epistemic claim that we have no privileged way of knowing what our wills have in fact caused.

Of course if we have other grounds to believe that the will, or consciousness of willing, is epiphenomenal, that would support Wegner's epistemic claims.Footnote ¹⁴⁷ For then, there would be no causing (by wills or by consciousness) for us to perceive via our experience of will, and we could have neither privileged access nor infallibility. But Wegner's epistemic challenges should be seen as resting on their own evidential base, not as some corollaries of the challenges earlier discussed.

6. The challenge of secondary agency

A distinct kind of challenge can be seen by anthropomorphizing three of the challenges already considered. Take the second, the lack-of-free-will challenge, for illustration. Suppose what causes the will is not just some dumb, brute physical phenomena; suppose it is “someone else,” another person-like entity residing within us but not identical to our own agency. I call these “second agent” conceptualizations. Such second agencies seem to challenge our sense of self and of responsibility more than their non-anthropomorphized counterparts do.

Daniel Wegner documents this effect in our judgments when (literally) other people's agency impinges on our own. As he recounts, “we become ill at ease when we are faced with questions of our own free will not so much because we have some aversion to causal determinism arising within us, but because of the concern that we are being pushed around by others…. All it seems to take is another's prior movements, or even the hint of a command, and we relinquish much of our own experience of will and allocate responsibility to the other.”Footnote ¹⁴⁸

Freud's “metapsychological” theories famously posited other agencies within us that were “not-me,”—the topographic “System Ucs,” replaced by Freud in 1923 by the structural “id.” Of these second agents, Freud said, “All the acts and manifestations which I notice in myself and do not know how to link up with the rest of my mental life must be judged as if they belonged to someone else and are to be explained by the mental life ascribed to that person.”Footnote ¹⁴⁹ This anthropomorphic feature of Freudian theory made it particularly challenging of our sense of will and responsibility.Footnote ¹⁵⁰ If physical causes that we can't control are already a challenge to free will, when such causes are conceived of as second agents, that seems to up the ante for many people. Being the puppet of an alien intelligence seems to many people to be more threatening than being the causal product of purely mechanical forces. (If this is right, Predestination must have been more of a challenge for Calvinists than materialism was for Hobbesians.)

When contemporary neuroscientists interpret the Libet-to-Haynes results anthropomorphically, they are inviting the greater challenges of second, “not-me” agencies controlling our decisions. In his 1985 article, Libet interpreted his findings thusly: “the brain ‘decides’ to initiate or at least to prepare to initiate the act before there is any reportable subjective awareness that such a decision has taken place.”Footnote ¹⁵¹ John-Dylan Haynes much more recently has also characterized Libet's results in similarly anthropomorphic terms: “it has been argued that the brain has already unconsciously made a decision to move even before the subject became aware of it.”Footnote ¹⁵² Haynes similarly anthropomorphizes his own more recent findings, seemingly attributing decisions to his favored brain areas: “One interpretation of this finding is that the frontopolar cortex was the first cortical stage at which the actual decision was made, whereas precuneous was involved in storage of the decision until it reached awareness.”Footnote ¹⁵³

If you take this anthropomorphic gloss seriously, neuroscientists would be replacing Freud's inner demons with another, equally alien second agent, namely, our brains. For some people this would intensify the challenge presented to free will by the Libet-to-Haynes data. It would similarly magnify the epiphenomenalist challenges earlier described, because now the common cause of (conscious) willing and behavior would not just be some dumb physical events, but “someone else” who competes with us for control of our own bodies.

One might think that the anthropomorphic description of Libet, Haynes, and others is just loose talk, that unlike Freudian theory there is nothing in neuroscience that commits it to such second-agent descriptions. While ultimately I think that this is right, still the immediacy of the causes of movement with which neuroscience deals, and their spatially proximate location (within our own head), peculiarly gives rise to ideas that “brains decide.” Suppose Haynes becomes almost as infallible a predictor of human choice as was Hume: suppose Hume's predictive success (about how if a pot of gold is left at Charing Cross station someone will pick it up)Footnote ¹⁵⁴ is at 98 percent, and Haynes achieves a like accuracy from studying the goings on in the precuneous region of the parietal cortex. Isn't the temptation to say, “the matter has already been decided,” much more natural about Hayne's prediction than about Hume's? “The decision was already made” describes Haynes situation more naturally than Hume's (where we might say that the decision had been predicted but it hadn't yet been made), because Hume's environmental causes are not in the head. In Hayne's cases the conscious subject seemingly has not made the decision until six to nine seconds later; but since (by the above inference) the decision has been made at the earlier time, someone or something must have made it. Who else but the brain? This is undoubtedly better science than were Freud's mysterious homunculi, but it challenges our sense that we are in charge as much as Freud ever did.

One can feel this challenge, of course, only if one does not have a reductionist perspective clearly in mind. If all of our mental states are brain states, so that we are (identical with) our brains, there could be no second agent challenges. On reductionist assumptions, there is only one agent, and it is our mind/brain. So this challenge can be mounted only by one rejecting reductionism. Although this rejection was true of Libet and other dualists like Eccles, it is not true of Haynes and most neuroscientists. For them this last kind of challenge has bite only in moments of forgetfulness of their true ontological commitments. For lay people it may be that there are many such moments.