by Hugh J. McCann, Texas A&M University

Responsibility for an action requires what Professor McCann calls an exercise of legitimate agency of the part of an agent, a necessary condition for which is libertarian freedom.  Free decisions are to be explained teleologically, not causally.  Agent causation cannot account for the existence of a free decision, but neither does event causation account for the existence of determined events.  The problem of accounting for the existence of a free decision is therefore of a piece with the problem of accounting for the existence of the world itself. All of this, like a related line of argument by Professor McCall to which you can turn, is a long way from what seems to me the continuing arguableness of determinism and the unavoidableness of the proposition that both Incompatibilism and Compatibilism about freedom are false. But we all need to remember, with Cromwell, in our own bowels if not by those of Christ, that we may be mistaken. I guess that given the proportion of false to true views in the world, we need to remember it is arguable that we are all more likely to be mistaken. -- T.H.

At the foundation of human action there lies the enigma of agency:  the phenomenon whereby, as it seems to us at least, our actions are finally to be accounted for solely in terms of our performing them.  Agency is an enigma first because it is hard to say what it is.  Exercises of agency are at best difficult to describe, and the concept resists any effort at reductive analysis.  Second, it is hard to see how events that are finally up to us are to be explained, or that their explanation can be made commensurate with our notions of an orderly universe.  On the other hand, to relinquish the idea of agency is to jeopardize the entire concept of human action, and with it our sense that we are responsible in a distinctive way for the changes we produce in the world.  It is therefore worth investigating how viable the concept of agency is.  I want to argue that while it is unlikely the objections it faces can be fully dispelled, the concept of agency is quite viable, and that its problems are in the end not much more daunting than those that face its usual competitor, the notion of event causation.  Agency, if real, results in explanatory discontinuities in the world;  but it does not introduce events that have no explanation, or whose existence is any more mysterious than the existence of things in general.


 The word "responsibility" is not always associated with agency.  To say the circuit breaker was responsible for the lights going off is just to say that it was the subject of some event which, by natural processes, resulted in the lights being extinguished.  And it is to imply that we can rectify the situation with the lights by doing something to the circuit breaker, such as flipping it back on.  Sometimes human behavior can be treated the same way.  If a muscle spasm causes my arm to strike the lamp, knocking it over and extinguishing the light, then I am responsible for the lost illumination in the same weak sense that the circuit breaker was.  And our recipe for preventing such situations from recurring is likely to be about as simple:  just keep me away from lamps.  The sense of responsibility associated with agency is far graver.  If I deliberately knock over the lamp, my doing so is an action—a manifestation of agency, for which I would normally be morally responsible.  Here, my involvement runs much deeper than merely being the subject of an event with untoward consequences, and to say I am responsible implies that I am an appropriate target for much more Draconian efforts aimed at rectifying the situation.  It implies that I can be held to account for what I did, that I am liable for the damage resulting from my behavior, that I may be blamed or punished for my action.

It might be protested that blame and punishment are not different in principle from keeping me away from lamps, that they are simply a matter of bringing to bear influences calculated to produce modifications in my future behavior.[1]  But that is not the whole story.  For one thing, punishing me is supposed to deter others also.  More important, however, is the fact that the very same treatment of an individual will count as punishment or not, depending on whether we think him responsible.  We incarcerate mentally deranged killers as well as murderers—in both cases working a hardship on the individual in question, and in both cases for the protection of society.  But only in the latter case is the hardship viewed as punishment—that is, as justified precisely because it is a hardship, appropriately visited upon the offender.  And while we may hope thereby to motivate a change in behavior, we do not require that.  Often, in fact, the agents we are most anxious to hold responsible for their deeds—mafiosi, mass murderers, and the like—are precisely those whose motives we think are least likely to be changed or even diminished.  Finally, and most important, when they work best blame and punishment do not affect behavior in just any old way.  The aim is not simply to terrorize the offender into future compliance.  We will settle, if we must, for a thief who out of fear of further punishment steals no more.  But that much can be achieved even with kleptomaniacs, whom we do not consider responsible.  And the problem in both cases is the same:  a terrorized kleptomaniac is still a kleptomaniac, and a terrorized thief is still a thief.  What we would rather achieve is a situation in which each agent acts out of a proper appreciation for the nature of his action, and the values it involves.  In the case of the thief, that requires the sort of transformation of character we call "reform."  Ideally, that is what blame and punishment are aimed at achieving.  And the problem for the objection we are considering is that a crucial step in the process of reform is that the agent take responsibility for his actions.  That brings us full circle, for it is obvious that to take responsibility for one's actions cannot be just to come to believe that suffering will inhibit their recurrence.

The true conditions for moral responsibility are somewhat complex.  One, which is probably violated in the case of the kleptomaniac, concerns knowledge:  the agent has to know what he is doing, the likely consequences of his act, and the valuational standing of the actions those consequences in turn define.[2]  Failing this, he can justly be held responsible only if he is also responsible for his ignorance.  The condition that concerns us here is, however, a deeper one:  the subject's behavior has to be a manifestation of agency.  When the motion of my arm is caused by a muscle spasm, it does not matter what I know about it.  I am not responsible, for I have performed no action.  But if I move my arm—in fact, if I only try to move it—then I have acted, and I am responsible.  The locus of responsibility lies, then, in exercises of agency.  And these are not to be found on the surface of behavior.  The motion of my arm can count, on some views of act individuation, as part of an action of mine.  But it never counts as an action in itself, at least in ordinary discourse.  Rather, it is my bringing about the motion that counts as an action, and the fact that I can undertake to do this without any physical result ensuing shows that the phenomenon of agency is an interior one.

The true center of agency lies in exercises of the will—that is, of the mental faculty of voluntary behavior.  Exercises of the will include such things as the activity of concentrating one's attention on one or another item of mental content;  the act of deciding, through which intentions are usually formed;  and the activity of volition, by which intentions for overt action are usually executed.  The distinguishing feature of these endeavors is that they possess intrinsically the characteristics we associate with responsible action:  they are, and must be, intentional exertions of the sort of control we have in mind when we think of an act as being "up to us."  This is perhaps best seen in the case of deciding, which is the most important manifestation of the will for purposes of the present discussion.  As I have argued elsewhere,[3] deciding cannot be unintentional or inadvertent:  in making a decision, we mean to decide, and to decide exactly as we do.  Neither can a decision be involuntary in the way a muscle spasm or a reflex knee-jerk can be.  Even when threats or the forces of circumstance "compel" us to decide a certain way, the compulsion is rational, not nomic.  The gunman who convinces me to turn over my wallet operates, ironically, not by the use of force but through the persuasiveness of reasons.  And if I later say of my decision, "It was involuntary," I mean only that had it been up to me, the reasons would have stacked up differently.  I do not mean that I was driven by blind emotion, or even that I could not have decided otherwise:  after all, many have.  I mean only that by my lights anyway, that would have been an irrational decision to make.

Deciding is, then, an act that possesses intrinsically the features we associate with responsible control of behavior.  Along with the other types of  exercise of agency mentioned above, it is by its own nature a purposive manifestation of voluntariness—something that is essentially action in the fullest sense.  That this character is intrinsic to decisions explains why they are not subject to problems of causal deviance.  If agency were just a relational matter—a question, say, of being caused in the right sort of way by the right sort of reason—then we should be able to find decisions that would have been exercises of agency, but for being caused by the wrong reasons or in the wrong way.  But that does not happen.  There are no decisions that fail to be exercises of agency:  none in which we do not intend to engage, none that lack the phenomenal
character that leads us to think they are our doing, something we control.  Unfortunately, however, the fact that agency is intrinsic to exercises of will also makes it difficult to understand.  I for one see no chance of analyzing it away, of reducing it to concepts more familiar and supposedly less offensive.  But neither do I see reason for regret in that.  Concepts we cannot get rid of tend to be important sources of insight;  and if they cannot be eliminated, they can still receive useful elucidation.


The concept of agency has it that the operations of my will are fully my responsibility.  My decisions and other activities of will are founded in me, not just because as aspects of mine they could not exist if I did not, but because I am active in their appearance, in a way that makes them manifestations of my rational autonomy.  This is an essentially positive idea, but it has a negative implication we need to sort out first.  If my action is autonomous it cannot be compelled, either from within or from without;  it has to be an instance of spontaneity on my part.  And for that to be so I must have had some alternative, if only the alternative of forbearance or inaction.  Applied to deciding, this means that at the moment I make a decision, it must be possible for me to decide differently—or at least to commence to do so—or to forbear deciding, so that my inactivity too would count as a demonstration of my will, of my capacity to decide what I choose when I choose.  I understand this to be an essentially libertarian conception:  to have decided freely I must, categorically, have been able to do otherwise.

There are two ways the conception can be attacked.  One is by way of an argument set forth by Harry Frankfurt, according to which responsible agency does not in fact require the possibility of doing otherwise.  It may seem to, since we usually think of any condition that would guarantee my deciding to do something as actually figuring in the etiology of the decision, so that I would be compelled to make it.  According to Frankfurt, however, this need not be so.  Suppose, for example, that I decide in more or less the normal way, with no interference or coercion, that I will vacation in Italy next summer.  Since my decision is neither uninformed nor compelled, I am surely responsible for it.  But suppose too that Jones, who is resolved that I shall decide to vacation in Italy, has implanted a contraption in my brain that would enable him to force me so to decide.  In the case at hand, he has no need to use it, since I decide that way anyhow.  But, the argument runs, had I been about to decide otherwise, Jones, who is an excellent judge of such things, would simply have pushed a button on a little transmitter he carries, sending a signal that would have compelled me to decide to go to Italy.  Here, then, I could not have done otherwise.  Still, given what actually occurred, there is every reason to hold me responsible.[4]

The plausibility of this sort of argument depends on what conception of freedom is presupposed.  Against compatibilist views it has some purchase, since if determinism is true there will be determining conditions for any decision that occurs, so that an early tipoff that I would decide against going to Italy would easily be possible, and Jones could take action.  Against libertarian views, however, the supposition is far less plausible.  It has to be remembered that doing otherwise does not, in this case, necessarily mean deciding otherwise.  I need not make any decision at all.  In a context where I am deliberating over my options for next summer, that would count as forbearing to decide (as yet), and for that I am responsible.  But it is not clear, in a nondeterministic setting, what the early tipoff of such forbearance would be.  After all, up to the point at which I actually decide to go to Italy, I am already forbearing to decide.  Only by allowing me no deliberation at all could Jones prevent that, and then I would not have the chance to decide on my own.  So it begins to look like Jones can be in the position of being able to control my decision only by actually controlling it.

There is, moreover, a general argument that tipoffs of this type are not possible in a libertarian setting.  Let t be the time when I commenced to decide to go to Italy, and consider the alternative scenario in which I would decide to go to France instead.  Now Jones cannot wait for this to occur before pushing the dreadful button, for then I would already have done otherwise:  I would have undertaken to decide to vacation in France.  True, I might not have completed the decision:  there might, in John Fischer's words, have been only a flicker of freedom.[5]  But that, I suggest, does not matter.  Moral autonomy is similar to Kant's good will:  a flicker shines like a beacon.[6]  So the tipoff to Jones must occur prior to t.  It must, moreover, be reliable:  it cannot be found only on some occasions when I am about to decide otherwise, since that leaves open the possibility that on this occasion it will not appear, and Jones will be unable to head off my decision.  So the tipoff, whatever it is, has to be a necessary condition of my deciding otherwise.  But now it turns out that in the original example, quite apart from Jones's nefarious schemes, I could not at t have decided otherwise than I did.  A necessary condition was missing—namely, the prior occurrence of whatever would have tipped Jones off.  So on a libertarian understanding of "could have done otherwise," any situation where a Frankfurt-type counterfactual intervener could have succeeded in controlling a decision is one in which there is an independent proof that the agent lacked moral freedom anyway.[7]

If this argument is correct, then the libertarian is secure against Frankfurt examples.  He can persist in maintaining that the ability to do otherwise is a necessary condition of moral responsibility, since the examples which purport to show that this is untrue in fact beg the question against a libertarian understanding of that ability.  If decisions are exercises of libertarian agency, there can be no completely reliable way for a would-be intervener to read what one will be prior to its inception.[8]

Perhaps, however, the libertarian understanding of what it means to be able to do otherwise is wrong.  A second, more traditional line of attack has it that this ability is not categorical but, as Hume said, hypothetical.[9]  To claim an agent could have done otherwise is not to say he might have done so in exactly the same circumstances, but rather that had some crucial determining condition been altered, then he would have done differently.  This, of course, is perfectly consistent with determinism, and if it is all there is to liberty then agency may not be so enigmatic after all.  To adopt this view is, of course, to raise the specter of the Frankfurt examples again, but as Frankfurt himself points out, one crucial feature of those examples is that in them, the villain does not actually influence the agent's behavior at all.[10]  So one can still claim that in such cases, action is "free" in the respect with which compatibilists tend to be most concerned:  it is a valid expression of the agent's character.  Indeed, one can even claim the agent could have done otherwise, simply by modifying the analysis:  had the crucial condition been altered, then barring alien intervention the agent would have done otherwise.

The initial plausibility of this sort of maneuver stems from the fact that conditional analyses can account for what is sometimes called freedom of action, as opposed to freedom of the will.  The difference can be seen in an example of Locke's.[11]  A man held in a locked room lacks the freedom to leave, in that no exercise of agency available to him will get him out.  Should the door be unlocked, however, he will be free to go:  if he wills the physical exertion necessary to get him out of the room, then that is where he will wind up.  So the conditional analysis succeeds.  But the problem here is not one of freedom of the will.  It concerns only the pathway from the exercise of agency to success—that is, from volition to its projected consequences in the world.  That is freedom of action.  The question of free will has rather to do with the relation between exercises of agency and their antecedents.  If the man does not know he is locked in, and decides on his own to remain, his decision will be just as voluntary as it would be had the door been unlocked.  And although Locke himself refused to accept the idea, that makes the decision an exercise of free will.  The real challenge for compatibilism is to provide a satisfactory analysis of this kind of freedom.
I do not see how the challenge can be met.  For traditional compatibilism, freedom in deciding is a conditional matter:  typically, a decision is held to be free provided it would have gone otherwise had the agent's strongest, or definitive, or predominant desire been for some other action.  But if this is supposed to capture the everyday, preanalytic notion of free will, it seems clearly to fail.  For one thing, the analysis misses the target distinction almost entirely.  Compulsives, addicts, people operating under duress—virtually everyone whose freedom to will differently we ordinarily view as compromised—would count by this criterion as free.  Surely, if determinism is true, they would have willed differently had their strongest motives been different.  Yet these are the very people whose responsibility for decisions we would question, precisely because we think their strongest motive was too influential.  Indeed, aside from Frankfurt-type cases, it is hard to think of any examples of impaired freedom of the will that we would not have to pronounce fully free by this criterion.[12]

The situation changes little if the compatibilist takes Frankfurt-type examples as showing that responsible agency does not require that the agent have been able to do otherwise.  We still need to know why responsible agency is compromised in cases of addiction, compulsion, and the like, but not with normal actions.  And the task gets no easier.  It will not do, for example, to say a decision is responsible provided it arises out of the agent's motives and character, without alien intervention.  This is surely true of addicts and compulsives, as well as normal agents.  Nor will it do to add that responsible control over one's decisions is a matter of their issuing from processes that would be responsive to good reasons for behaving differently.  The problem is that this is too strong.  Agents who behave wrongly usually are cognizant of good reasons for doing otherwise, but do not respond to them.  And if we weaken the requirement to demand only that there be some sort of reason which, had it arisen, would have been persuasive to the agent, we are back with the problem that this too is true of addicts and compulsives.  Kleptomaniacs don't steal if they know they are being watched, and even the powerfully addicted often find motives to get well.[13]

At least as telling is the way people react to compatibilist analyses of freedom and responsibility.  Laymen, in my experience, are bewildered by them, wondering how we can be held responsible if we "really had no choice."  Most students simply fail to grasp compatibilism:  they insist on taking it as a theory on which causal factors limit one's options to some extent, but allow "free will" to operate among those that remain.  And the minority who do understand are all but uniformly hostile to the view.  Above all, there is the reaction of philosophers themselves—who, if such analyses could indeed capture the ordinary notion of responsible agency, ought to flock to compatibilism, and the neat solution it provides to a profoundly difficult problem.  Yet it has produced nothing but division, garnering every reaction from enthusiastic endorsement to indignant disdain.[14]  Accurate conceptual analysis should not have these results.  One might, of course, view analyses like these as revisionist in spirit:  as efforts to lay out the best account of freedom we can have in a deterministic world.  But then the classic objections apply.  Why should I take myself to be free in my decisions, if they are determined by other events over which I finally have no control?  Why should I consider myself responsible for them in any sense more important than I am when a muscle spasm causes me to break the lamp?  This kind of conceptual revisionism seems, furthermore, to be rather weak-kneed.  If the common-sense notion of responsibility is founded upon a false conception of what agency consists in, then surely it should be set aside and a new beginning made.  Rather than hold offenders morally responsible, we should treat all so-called wrongdoing as essentially compulsive, drop the idea of blameworthiness, and design what we would conceive as therapeutic programs aimed at correcting behavior we do not like.

Neither as reports on common usage nor as proposals for conceptual revision, then, do compatibilist treatments of free will succeed.  Moreover, it is much too early for the sort of capitulation described above.  Inevitable though it may seem to some, the day when a deterministic account of the will becomes an established fact appears distant at best, just as it did three hundred years ago.  And in the meantime, there is still the sense we have when we engage in exercises of the will that what happens is up to us, a matter of our intentional control.  If the argument of this section is correct, part of what is involved in that is nomic indeterminacy:  the idea that the agent might have done otherwise in precisely the same circumstances that occurred.  In the case of deciding, this means that the agent might have decided differently, or have forborne deciding, without alteration in his information and motives.  But this negative condition cannot be all there is to responsible agency.  To see why, we need only suppose that, say, the onset of desire were unexpectedly discovered to be an undetermined phenomenon—for example, that my sudden desire to have steak for dinner was in fact not determined in its occurrence.  Were such a discovery to be made, we would not consider ourselves any more responsible for our desires than we do now.  The onset of a desire, caused or not, is still an event that befalls or happens to us, something in which we are entirely passive.  So if we are any more responsible for our decisions, that has to be owing not just to their being undetermined but to their  to their positive nature as well.


We can begin to appreciate the positive aspects of deciding by considering the sort of objection usually raised against the claim that decisions are not nomically determined.  There is in fact a cluster of such objections, built around the common theme that such an event can have no adequate explanation.  In one form, the complaint is that an undetermined decision must be a random or accidental event, over which the agent had no control, and for which he therefore cannot be responsible.  Thus Hume held that if actions do not proceed from some cause in the character of the agent, he cannot on their account become an object of punishment or vengeance.[15]  A. J. Ayer argued that if it has no causal explanation, my choosing is an accident, for which I cannot be responsible.[16]  And R. E. Hobart claimed that insofar as an act of will is uncaused, it is as if one's legs should spring up and carry him off where he does not prefer to go.[17]  These views presuppose, however, that only through nomic causation is it possible for an exercise of will to be nonaccidental, or to exhibit voluntary control.  As we have seen, the preanalytic data indicate this is false;  indeed, it would be a self-contradiction for me ever to assert that I had accidentally decided to do anything.  What that indicates is that voluntary control is not just present in acts like deciding, but is essential to them, in a way that does not reduce to a causal relation to other events.

Voluntary control consists in general in an agent's ability to direct his behavior to ends he selects.  Exercises of the will are, by their own nature, manifestations of this ability.  There are, I think, at least two important dimensions to this.  One has to do with ontological foundations.  An exercise of agency has to be spontaneous and active;  it is a creative undertaking on the agent's part, to be accounted for in terms of its intrinsic features, not via the operations of other denizens of the world.  Second, exercises of agency must be intentional;  they have to be undertaken for the sake of some objective the agent deems worthy of attainment.  As with other exercises of the will, the phenomenology of deciding indicates we view these features as essential to it.  What counts truly as an act of deciding can only be up to me;  it must be an act of mine, rather than something I undergo.  And I have to mean it:  if I decide, then I must intend to decide, and to decide exactly when and as I do.  Thus I am responsible for both the act and its content.  It is worth mentioning that neither dimension of voluntary control makes much sense without the other.  A creative act on my part not undertaken intentionally would be almost a contradiction, and an act that I mean to perform but whose origin lies elsewhere would be one whose intentionality is at best hollow and redundant.  But although the two dimensions are inseparable, they can largely be discussed independently.  It is best to begin with the matter of intentionality.[18]

To decide is to progress from having reasons to having an intention.  A course of action portrayed in my thoughts as appropriate to advancing one or another objective I value conatively is transformed by my decision into a goal or purpose of mine, something I am committed to achieving.  I might, for example, see going to Italy as necessary step toward a desired visit to Florence, and so decide to vacation in Italy next summer.  When I do this, vacationing in Italy becomes my intention, and so does visiting Florence, if it was not an intention already.  That is, the desires I conceive vacationing in Italy as a means to satisfying become further intentions, further objectives to be achieved by my going there.  What is most important, however, is that my act of deciding is in itself a step toward achieving those objectives.  It settles, as far as an advance decision can, what I shall be doing next summer, by committing me to the trip, as well as to the preparation and coordination with other activities it will require.  So when I decide to vacation in Italy, I am consciously advancing the project of so doing.  That is why it is important that deciding be an intrinsically intentional act.  Anything else would contradict its functional role, since an unintended decision would be purposeless in itself, and would result in my having an intention I never meant to have.

Because to decide to A is to advance the project of A-ing, one's reasons for deciding to A are, in the usual case, simply one's reasons for A-ing.  The desires and obligations I can satisfy by vacationing in Italy are reasons for deciding to go there as well as for going.  And to decide for the sake of those reasons is simply to mold them into the resultant intention, thereby advancing intentionally the project they represent.  Notice that there is nothing of nomic causation in this story.  What makes my desire to visit Florence the reason why I decide to vacation in Italy is not that it has some force that compels me to decide:  if that were so, we could not know what my reason was, since the question whether decisions are nomically caused is at best moot.  Rather, my desire is my reason because when I decide I intentionally make visiting Florence, and the associated goods portrayed in that desire, the guiding objective of the trip to which I am committing myself.  Quite apart from the issue of causation, therefore, my decision has an explanation.  It is a teleological one, in terms of the perceived goods which, in my deliberation, justified my decision, and which came to be intended through it.  Add to this the factor to be considered more fully in the next section—that our decisions seem to us to be under our voluntary control—and a libertarian decision appears as anything but an accident that befalls the agent.  Indeed, it looks like we would have more to fear in this respect if decisions were nomically caused.  That, it seems, would render our sense that we intentionally direct our decisions illusory, making us their victim rather than their perpetrator.

There is little chance, however, that the determinist will be satisfied with this.  For, he may argue, even if deciding has features that render it incapable of being accidental from the practical viewpoint of the agent, a satisfactory theoretical explanation of it is impossible from a libertarian perspective.  Causal explanations satisfy our need to understand because they invoke laws and conditions in terms of which events can be seen to be inevitable.  The sort of explanation promised by a libertarian view of deciding does nothing of the kind.  It proceeds in terms of the agent's thoughts—the abstracta that occupied his deliberations—and it holds that the values and objectives portrayed in those thoughts may be seen to justify the decision based on them.  But the problem is that there will also be reasons on the other side, the side the agent did not choose.  My desire to visit Florence may justify going to Italy, but if I was also considering a vacation in France then no doubt I also had reasons for going there, and they would have been held to explain my decision had I finally opted for that alternative.  But what cannot be accounted for is why I chose to go to Italy for one set of reasons rather than to France for the other.  We may be able to postpone the moment of reckoning by citing further reasons.  I may have wanted to visit Florence in order to go to San Marco, so that I might see the Fra Angelico frescoes there, to ponder their religious significance, etc.  But sooner or later my train of reasons must come to an end.  And when it does, so the argument goes, my decision will be finally unaccounted for:  there will be no explanation for my choosing to pursue those purposes, rather than the ones I could have achieved by vacationing in France.[19]

There is, I think, a legitimate concern underlying this argument, though in the end it not a decisive one.  What we need to see first, however, is that as it stands the argument is badly inadequate.  Consider first the claim that my reasons for vacationing in Italy must come to an end, at which point we will be left with no explanation for my choice.  And now suppose for a moment that the train of reasons did not come to an end:  the riches of Florence, we might imagine, are so profound that any level of appreciation for them I might have only leads on to another.  And we can imagine that my mental capacity were such as to comprise all of that, so that I could actually appreciate such bounty.  Would that refute the objection?  Clearly not, for if the riches of Florence are infinite, then surely those of France are as well.  If that were so—and for that matter even if it were not—why should prolonging explanations on the Florentine side quiet the determinist's complaint?  I still had the French option, infinite or not, and I still might have chosen it.  Furthermore, the discussion at this point takes a rather nasty turn for the determinist.  For the truth is that the very sort of explanation he favors offers the best available paradigm of failure to solve this kind of problem.  Let my decision be nomically caused, and let it belong to a world that is determined to the hilt.  Even so, there will be other possible worlds—e.g., a world minimally adjusted for my choosing to vacation in France.  And even if we conceive of these worlds as infinite in duration, no amount of explanation from within the causal sequence supposed to lead to my choosing to vacation in Italy can explain why this world exists, rather than the one in which I decide to vacation in France, or for that matter no world at all.[20]

There is reason to think, moreover, that the position of the libertarian is actually a good deal more satisfying than that of the determinist in this regard.  There are, as far as we know, no natural stopping points for causal chains.  They either proceed to infinity, or they begin arbitrarily, at least as far as nomic principles are concerned.  So if we want to know why we have one sequence of causes rather than another, we must either prolong the sequence, which only postpones failure, or seek another sort of explanation.  With sequences of reasons the situation is different.  In reason explanations, the appeal is not to law-bound inevitability but to the justifying force of perceived goods.  The point is not to detail the ontogenesis of decisions, but to demonstrate their putative value.  Given this objective, a sequence of reasons need be prolonged only until an objective the agent deemed to be of intrinsic value—that is, something he took to be valuable for its own sake—is reached.  Once that occurs, we have reached a natural stopping point.  So if the ultimate goal of my projected vacation in Italy is, say, the aesthetic enjoyment of contemplating the frescoes at San Marco, the teleological explanation of my decision need proceed no further.  If someone asks why I decided to visit Italy for the joys of San Marco, rather than France for the delights of Chartres, the answer is that I did so for the sake of the joys of San Marco.


All the same, the determinist's complaint is founded upon a legitimate concern, which begins to emerge more clearly if we realize that the two types of explanation we are considering function somewhat at cross purposes.  As indicated, the primary focus of the libertarian's teleological explanations is not on the ontological foundation of decisions.  Practically speaking, there is no need for that, because in practical affairs we concern ourselves with the details of how events come to pass only when we are trying to bring them under control.  With decisions this is neither necessary for agents nor possible for observers, if libertarianism is right.  By their nature, libertarian decisions are events only the agent controls, and he controls them "at will."  Rather, reason explanations address the dimension of action appropriate for dealing with purposive beings who initiate new sequences of change in the world:  they focus on the way decisions are inspired.  This has reference to beginnings, of course, but not in a way calculated to providing a key to control.  The function of reasons is not to control but to persuade.

The concerns of the determinist, by contrast, are much more like those we have when we are not able to control events directly, but must instead will a sequence of change that will eventuate in a desired result.  Such indirect control requires understanding how events come to pass, and that is the sort of understanding the determinist would like to have of the operations of the will itself.  Here the focus is fully on ontological origins, and in terms of that focus the objection to libertarian decisions is easily renewed.  Whatever the shortcomings of nomic causation for explaining why we have this world rather than some other, determinism at least avoids a corresponding failure in dealing with events within the world we have.  A deterministic world is a seamless fabric:  every transformation of things, all that occurs, may be seen to emerge in a law-governed way from what already exists.  We can tell where present events come from, it is said, because they are brought about by past events in accordance with scientific law.  Within a deterministic world there are no explanatory gaps, because each event is made inevitable by those that went before.  The libertarian's world, by contrast, is filled with such gaps.  At innumerable junctures, decisions and other acts of will go one way rather than another, and we have no sufficient explanation why.  To be sure, the agent's reasons may make his decision fitting.  But they do not make it inevitable, and in that sense we do not and cannot know where the decision comes from.  It is a discontinuity in the world, an event that simply crops up, and whose ontological origin, if any, is utterly lost to us.  In short, however congenial they may be to our theories of moral responsibility, uncaused decisions are theoretically anomalous and rationally unacceptable;  they violate our expectation that the world be an intellectually comprehensible place.

One response to this version of the determinist's objection is that the issue here is in the end an empirical one:  if acts of will turn out to be nomically undetermined, we will simply have to accept that fact, as we have with other natural phenomena, and try to understand them as best we can.  Often, however, the libertarian response goes further.  That exercises of agency are not subject to nomic causation does not, it may be argued, mean they have no causes at all.  Rather, when I decide to vacation in Italy, it is I, the agent, who am the cause of my decision.  Instead of being brought about by other events, it is brought about by me, through an exercise of voluntary power.  It is therefore wrong to assert that we do not know where undetermined acts of will come from.  True, they are nomically discontinuous with the rest of the world;  but they come from their agents, who knowingly and intentionally produce them.  This may not solve every problem about understanding the will, but it does provide an alternative to the idea that nomically uncaused decisions present a final and impenetrable mystery of ontogenesis.  That, the argument runs, is simply an error, which arises from taking nomic causation to be the only acceptable way in which an event may be brought to pass.[21]

There are, however, some problems with this answer.  It has reference, of course, to the other aspect of exercises of agency mentioned earlier:  the fact that decisions and other operations of the will seem to us not just to happen, but to be manifestations of voluntary control on our part.  It is, of course, this dimension of agency that constitutes its core.  What makes my decisions exercises of agency is not just that they are not determined by other events and states, nor even that they are intrinsically intentional.  There is also the fact that when I decide, I am actively doing something about the intention intrinsic to that act.  Like intentionality, this feature of exercises of the will is intrinsic and essential to them, and it has a certain sui generis character that renders it incapable of being reduced to anything else.  When we decide, we take ourselves to be actively carrying out a commitment to deciding which is itself brought to pass solely through that very act, and extends to all dimensions of it.  My decision, its timing, and its content are all "up to me," in that I am actively and intentionally given over to them through the act of deciding itself,  and there is no condition independent of that act that makes this so.[22]

It is not obvious, however, that we should take this aspect of deciding as providing any useful sense in which I may be said to be the "cause" of my deciding.[23]  It is, perhaps, a natural temptation to say that, as a means of trying to capture the unique aspect of voluntariness I have just tried to describe.[24]  It may even be fair to think of me as the "source" of my decisions, inasmuch as their occurrence is a manifestation of a power or capacity of mine—as opposed, say, to a power of something acting upon me.  But a source in this sense is simply a point of origin.  A cause is supposed to be more than that:  it should have some sort of explanatory priority with respect to the phenomenon to be explained.  But that notion has no purchase with substances as such.  Qua acting subject, I don't explain anything;  I only act.  Explanations have to invoke the descriptive and/or valuational aspects of things.  And in the case of deciding, as we have seen, explanation is a matter not of nomic determination but of teleology:  of the goals an agent is advancing in the act of deciding.  This sort of explanation does not require our speaking of the agent as causing his decision, nor is it strengthened by it.  Teleological explanations are about goals, not causes.

Even more problematic is the idea that voluntary agency is a matter of my somehow bringing about my decisions, in the sense of conferring existence upon them.  Again, one can understand the temptation:  the active spontaneity of deciding, together with its intrinsic intentionality, may suggest that when I decide, I do something that amounts to producing that very act—ex nihilo, as it were.  But it is very hard to find coherent content in this idea, taken literally.  For in what would the supposed productive relation consist?  If it is centered in some operation on my part independent of the act of deciding, then that operation, whatever it is, will become the true locus of agency—posing all of our problems anew, and threatening an infinite regress.  If, on the other hand, the productive relation is thought to reside in some aspect of deciding itself, then the occurrence of my act of deciding must have ontological priority.  Only  with its appearance could the causative aspect find reality.  But once the act does appear the causative aspect has no work to do, and so is redundant.  The idea of agency as a causal or productive relation seems, then, to be of no value.  It may help to convey the intrinsically actional quality of operations of the will, and to emphasize the point that we should not look for a "source" for them in the other events of the world.  But it creates the impression that there is some higher-order doing by which agents somehow bring their acts of will into existence, and that is not a workable view.


When the determinist's objection to libertarian freedom is framed as one about how the operations of will originate, then, the libertarian has no truly satisfying response.  The essence of his position is to postulate ontological discontinuity between acts of will and other events and states.  Once that is done, it is no longer possible to explain a decision as we would an event like a solar eclipse or the acceleration of a billiard ball—that is, as a natural outcome of the continuing dynamic processes that constitute our world.  And the void cannot be filled by other explanatory devices.  Teleological explanations may be suited to free decisions, but they are chiefly concerned with matters of content.  They address the rational grounds for decisions, not their ontological underpinnings.  And agent causation is simply a misfire—metaphysically misconceived, and lacking any real explanatory value.  Decisions are, of course, comprehensible in part via their antecedents.  They are influenced by character and circumstance, and they are made by rational agents, who in order to achieve anything in life must display at least some stability of purpose.  One can expect, therefore, that statistical generalization about decisions will often be possible, and that knowing an agent's character and projects will often enable us to predict how he will decide.  But this kind of order is more an upshot of free decision than a limitation on it.  If the operations of the will are undetermined, they really do represent discontinuities in the world, and their provenance is as hidden from us as that of the world itself.
This concession is not, however, as damaging as it first appears, for it turns out that provenance is a problem on the other side as well.  We have already seen this to be so with respect to the question why we have the world we do, rather that some other, or none at all.  Determinism is not equipped to deal with that kind of issue, because there are no nomic processes for getting us to this world from some other, or from none at all.  It might be thought, however, that once it is in place nomic principles can, if the world is deterministic, account for its continued existence, as well as the coming to be of new events and states within it.  Indeed, it might be thought, this is precisely what the metaphysical seamlessness of a deterministic world consists in:  that its history is a matter of law, in which present events produce or bring about future ones, in accordance with principles which, though not logically true, nevertheless articulate a kind of necessity which orders the progress of all things.  How else could it be that in a deterministic world we know where events come from, whereas in an indeterministic one we do not?

The matter is not, however, so simple—and not only because present scientific knowledge makes it all but ridiculous to suppose we live in a deterministic world.  The idea that present events are somehow able to produce or generate future ones is no more successful than that of agent causation.  For, again, in what is the alleged generative relation supposed to consist?  It cannot consist in some further event, something the causing event does, or an operation in which it engages.  There are not that many things events may be said to do in any case, and it is impossible to imagine any operation that would count as one event bringing another into existence.  We could, of course, make one up:  we could postulate some kind of intervening event–producing, say, or bringing about, or necessitation—and claim that it binds event-causes to their effects.  But we could have done the same with agent causation, and it would not have availed very much.  Nor does it avail anything here.  As Hume pointed out, all such claims are empirically vacuous:  we observe no causal nexus in the world, only the orderly succession of events.[25]  Moreover, as with agent causation, the kind of generational event imagined here would simply raise anew the problems it is supposed to solve.  Only now the problem is twofold.  First, if causal production does count as an additional event in the world, we will need to explain how events of this sort arise, again threatening an infinite regress.  Second, to account for the appearance of the effect the supposed "causing" would have to be independent of it, in which case some new generational relation would have to be postulated to bind these two as well, so that a second regress looms.  We cannot, then, postulate occurrences in the world that consist in one event causing another.  But if there are no such occurrences, then the claim that events can produce other events has no more substance than the idea that agents can produce the operations of their wills.[26]

It might be thought that recourse to scientific laws can save the day at this point—that what is crucial to the relation of causal productivity is that when it obtains, there will be a law which states that events of the kinds in question are bound by a relation of necessitation, by which the cause event must give rise to the effect.  I think, however, that this move only compounds the problem.  Scientific laws are, after all, propositions—that is, they are the means by which we report the way things are in the world.  But laws do not operate in the world:  they are not pieces of legislation, and they can neither put in place nor strengthen any real relation between events.  Rather, if there is some viable notion of necessity attaching to scientific laws—which is itself a contested question—it has to be because there is already an answering relation in the world, among the events laws purport to describe.  But then we are back where we began:  the necessity of laws has to be grounded in an appropriate relation in the world, which we are unable to find.  Indeed, the invocation of laws as a foundation for causal efficacy actually leads to a new embarrassment.  Whereas the causal relation is usually taken to be diachronic, scientific laws, classically at least, are not.  Newton's first law does not tell us that a body not acted upon by a net force will be at rest or in uniform motion a moment from now;  nor does the second law say a force applied to a rigid body at t will yield an acceleration an instant later.  Rather, the world of classical physics is one of simultaneous action and reaction, in which changes of state occur as a body is acted upon, not afterward.  I am not fit to judge whether all of physics is this way, but to the extent that it is so, diachronic productive relations among events are not called for by scientific laws.  They are, as Hume said, a product our imagination, something we have added on.

The idea that there are generative relations by which earlier events give rise to later ones is, then, no more successful than the notion that there is a generative relation by which agents produce their actions.  But then how is event causation to be understood?  And how do we account for our sense, which is surely legitimate, that we understand the origin of nomically determined events far better than that of a free decision?  The answer, I think, lies in treating the changes we observe in the world not as a matter of things coming to be and passing away, but as variations in the way what is preserved in the dynamic flux of things is manifested.[27]  The explanation of a solar eclipse is not that the event of the moon moving between the sun and the earth generates an eclipse.  It is that when the moon moves to this position the flow of energy from the sun to the earth is interrupted, which we observe as a darkening of the sun.  The explanation for the acceleration of a billiard ball is not that when the cue-ball strikes it, an event of acceleration is produced.  It is that when the cue-ball contacts the object ball a transfer of momentum occurs, which is manifested in the acceleration of the latter.  These are not stories about a mysterious nexus by which the universe somehow bootstraps itself into the future.  On the contrary, these explanations, and the laws that underlie them, deal entirely with the interplay of masses and of energy, whose existence throughout the processes in question is presupposed.  True, that interplay involves the entities of the world assuming new properties with the passage of time.  But nomic explanation, properly understood, treats this not as a generational process but as a transformational one.  The primary concern of science is not the existence of things, but rather their nature;  and it explains the present constellation of the world as emerging from what went before, not as created by it.

Taken in this way, nomic explanation is freed of the intractable task of describing processes of ontogenesis.  True, we have to give up the pretense that the present was brought into being by the past, but that was never anything but a false pretense anyway.  Rather, the way in which nomic explanations tell us where events "come from" is by making them continuous with what went before.  The nature of the entities composing a deterministic world is essentially passive:  their changes of state are entirely interdependent, so that an entity acts only insofar as it is acted upon.  The effect of nomic explanations is in part to trace those relations of interdependence, so that events need never surprise us.  But that is not all.  A good explanation will also describe the phenomena in terms that display the common nature of the entities involved, so that their interactions can be seen as transformations of a shared underlying reality.  And of course that is what is lost when phenomena are not determined.  Entities capable of spontaneous behavior act without being acted upon, so that events cease to be completely interdependent, and outcomes are no longer inevitable.  In effect, to the extent an event is undetermined, it has no natural ancestry.  That is the sense in which we do not know where a free decision comes from.  Notice, however, that this is simply a fuller description of the problem we saw in the last section, the problem of explanatory discontinuity.  There is not a further problem of a free decision failing to participate in a nexus that would otherwise explain its existence.  There is no such thing, even in a determined world.


Thomas Reid held that our idea of active power was derived not from observation of the world, but from the inner experience of voluntary exertion.[28]  If that is right, it is only to be expected that the illusion that we generate the operations of our own will should be translated into the world, and reappear as the illusion that events generate other events.  A more plausible view results if we give up the idea that causation is a matter of bringing events into existence.  Rather, event causation is a matter of present states of the world giving way to later ones by way of ongoing processes, in which the crucial participants neither come to be nor pass away, but simply display new arrangements and manifestations.  Agent causation, if we wish for it to be a viable concept, consists in the fact that through the intrinsically actional operations of the will, we are able to enter that causal stream, and thereby influence events at a distance from the will.  But it does not consist in our somehow "bringing about" the operations of the will itself.

The world that results is not, of course, the orderly world of the determinist, but it is not one for which the libertarian need apologize.  It is, rather, the familiar world of experience, in which agents act freely, and are responsible for what they do.  There is, of course, a large remaining problem about that world:  namely, where it, along with the free decisions and actions it includes, comes from.  And here comes from has the stronger sense that is concerned primarily with the sheer existence of things, and worries over why we have this world rather than some other, or none at all.  But that was going to be a problem in any case.  Relations of event generation, whether by agents or by other events, were not going to solve it either globally or in detail, because they do not exist;  to believe in them is to adopt a superstition as the solution to a mystery.  As to why we do have a world, that is not a problem that can be addressed here, but there are two things to be said.  First, it is only one problem, not two:  there is not one problem as to why this world exists, and then another as to why the events that make it up do.  And finally, it is only incidentally a problem for the philosophy of action.


[1]  For a classic example of this sort of view see Moritz Schlick, "When Is A Man Responsible?" in B. Berofsky, ed., Free Will and Determinism (New York:  Harper & Row, 1966), pp. 54-63.

[2]  It is sometimes claimed that compulsive behavior like kleptomania is owing to a deterministic "irresistible impulse," given which the agent had to behave as he did.  My inclination is to think that if libertarian free will is a reality this is wrong—that a faculty of free choice cannot be sometimes determined and sometimes not.  Rather, I think, situations in which freedom appears to be lost are actually just cases where, in the agent's deliberation, one option overshadows all others, thereby driving them away from our attention, so that it becomes the only valid choice.  We seldom call this compulsion when the values involved are reasonable, as for example in most cases of self-preservation;  but we do tend to do so when they are irrational.

[3]  See my The Works of Agency (Ithaca, New York: Cornell University press, 1998), chapter 8.

[4]  Harry G. Frankfurt, "Alternate Possibilities and Moral Responsibility," The Journal of Philosophy 66 (1969), 829-39, p. 835;  a more recent defense of such examples can be found in John Martin Fischer, The Metaphysics of Freedom (Cambridge, Massachusetts:  Blackwell Publishers, 1994), chapter 7.

[5]  Fischer, Ibid.  The assumption that Jones's reaction time is less than what it would take for me to complete my decision is, of course, utterly wrong;  but it is not logically impossible.

[6]  Immanuel Kant, Groundwork of the Metaphysics of Morals, trans. H.J. Paton, (New York:  Harper & Row, 1964), p. 62.

[7]  This argument is presented by David Widerker, in "Libertarian Freedom and the Avoidability of Decisions," Faith and Philosophy 12 (1995), 113-18.  Fischer responds in "Libertarian Freedom and Avoidability:  A Reply to Widerkur," Ibid., 119-25.  See also Robert Kane, The Significance of Free Will (New York: Oxford University Press, 1966), pp. 142-43.

[8]  It is worth mentioning that even if Jones were miraculously informed of my decision in advance, it is hard to see how anything he could replace it with would count as a decision in the libertarian sense.  The problem is the same as that with viewing kleptomania as an irresistible compulsion.  To treat decisions as intrinsically autonomous while at the same time subject to determination from without is rather like granting that the phenomenon of beta-decay is intrinsically undetermined, yet still claiming to be able to control it.

[9]  David Hume, An Inquiry Concerning Human Understanding, ed. Charles W. Hendel (New York:  Liberal Arts Press, 1955), p. 104.

[10]  "Alternate Possibilities and Moral Responsibility," pp. 836-37.

[11]  John Locke, An Essay Concerning Human Understanding, ed. J.W. Yolton (New York:  E. P. Dutton, 1961). Vol. 1, pp. 197-98.

[12]  Even the science fiction examples cut both ways.  Imagine an alien intervener who, rather than preventing my deciding in accordance with my strongest motive, actually speeds up the process by electronic means.  Here too, I am unfree, even if we make it out that had my motives been different, the friendly intervener would have made me decide accordingly.  Cf. Robert Audi, "Acting for Reasons," The Philosophical Review 95 (1986), 511-46, p. 531.

[13]  For more on reasons responsiveness see Fischer, The Metaphysics of Freedom, chapter 8, especially pp. 164-168, on which the discussion given here is based.  Fischer suggests that what he calls weak reason responsiveness is sufficient for responsibility, but I think that is mistaken.

[14]  For an example of the former see R. E. Hobart, "Free Will as Involving Determination and Inconceivable Without It," in Berofsky, Free Will and Determinism, pp. 63-95, esp. pp. 72-77;  for the latter, see Kant, Critique of Practical Reason, trans. L.W. Beck (Indianapolis:  Bobbs-Merrill, 1956), p. 99.

[15]  An Inquiry Concerning Human Understanding, p. 107.

[16]  A. J. Ayer, "Freedom and Necessity," in Ayer, Philosophical Essays (London, Macmillan, 1954) pp. 271-84, p. 275.

[17]  "Free Will as Involving Determinism and Inconceivable Without It," p. 70.

[18]  For fuller development of the points in this and the next two paragraphs, see The Works of Agency, chapter 8.

[19]  For a recent example of this kind of argument see Thomas Nagel, The View From Nowhere (New York:  Oxford University Press, 1986), pp. 116-17.

[20]  This kind of charge is associated, of course, with cosmological arguments for the existence of God.  But this consideration is at most secondary to the present discussion. The fact is that the charge is available to be made, and it makes exercises of agency that are determined as much a matter of “luck” as any that are not.

[21]  Defenders of agent causation include Roderick M. Chisholm, "Freedom and Action," in Keith Lehrer, ed., Freedom and Determinism (New York:  Random House, 1966), pp. 11-44; and "Human Freedom and the Self," in Gary Watson, ed., Free Will (New York:  Oxford University Press, 1982). pp. 24-35;  Timothy O'Connor, "Agent Causation," in O'Connor, ed., Agents, Causes, and Events (New York:  Oxford University Press, 1995), pp. 173-200;  and William L. Rowe, "Two Concepts of Freedom," Proceedings and Addresses of the American Philosophical Association 61 (1987), pp. 43-64.

[22]  Compare Kane's condition of Ultimate Responsibility, The Significance of Free Will, p. 35.

[23]  Kane would agree, Ibid., pp. 188-90.

[24]  Compare the suggestion of Carl Ginet, On Action (New York:  Cambridge University Press, 1990), p. 13.  Ginet too rejects the concept of agent causation;  but he also finds that it offers the best available simile for describing what he calls the "actish phenomenal quality."

[25]  An Inquiry Concerning Human Understanding, pp. 74-75.

[26]  The case against productive causal relations is presented at greater length by Jonathan L. Kvanvig and I in "The Occasionalist Proselytizer:  A Modified Catechism," in J.E. Tomberlin, ed., Philosophical Perspectives 5 (Atascadero, California:  Ridgeview Publishing Company, 1991), pp. 587-615.

[27]  For this kind of approach to understanding causality see Phil Dowe, "Wesley Salmon's Process Theory of Causality and the Conserved Quantity Theory," Philosophy of Science 59 (1992), 195-216;  and Wesley C. Salmon, "Causality Without Counterfactuals," Philosophy of Science 61 (1994), 297-312.

[28]  Thomas Reid, Essays on the Active Powers of Man (Cambridge, Massachusetts;  MIT Press, 1969), p. 36.