A familiar answer goes like this. There is something that the agent wants, and there is an action that he believes conducive to its attainment. His desire for the end, and his belief in the action as a means, justify taking the action, and they jointly cause an intention to take it, which in turn causes the corresponding movements of the agent’s body. Provided that these causal processes take their normal course, the agent’s movements consummate an action, and his motivating desire and belief constitute his reasons for acting.
This story is widely accepted as a satisfactory account of human action—or at least, as an account that will be satisfactory once it is completed by a definition of what’s normal in the relevant causal processes. The story is widely credited to Donald Davidson’s Essays on Actions and Events (1980), but I do not wish to become embroiled in questions of exegesis. I shall therefore refer to it simply as the standard story of human action.
I think that the standard story is flawed in several respects. The flaw that will concern me in this chapter is that the story fails to include an agent—or, more precisely, fails to cast the agent in his proper role. In this story, reasons cause an intention, and an intention causes bodily movements, but nobody—that is, no person—does anything. Psychological and physiological events take place inside a person, but the person serves merely as the arena for these events: he takes no active part.
To be sure, a person often performs an action, in some sense, without taking an active part in it; examples of such actions will be discussed below. But these examples lack that which distinguishes human action from other animal behavior, in our conception of it if not in reality. I shall argue that the standard story describes an action from which the distinctively human feature is missing, and that it therefore tells us, not what happens when someone acts, but what happens when someone acts halfheartedly, or unwittingly, or in some equally defective way. What it describes is not a human action par excellence.
Those who believe the story will of course contend that the events recounted in it add up to the agent’s participating in his action, as components add up to a composite. The story doesn’t mention his participation, they will explain, simply because his participation isn’t a component of itself. Complaining that the agent’s participation in his action isn’t mentioned in the story is, in their view, like complaining that a cake isn’t listed in its own recipe.
But this response strikes me as inadequate, because I don’t accept the claim that the events recounted in the story add up to a person’s activity. Various roles that are actually played by the agent himself in the history of a full-blooded action are not played by anything in the story, or are played by psychological elements whose participation is not equivalent to his. In a full-blooded action, an intention is formed by the agent himself, not by his reasons for acting. Reasons affect his intention by influencing him to form it, but they thus affect his intention by affecting him first. And the agent then moves his limbs in execution of his intention; his intention doesn’t move his limbs by itself. The agent thus has at least two roles to play: he forms an intention under the influence of reasons for acting, and he produces behavior pursuant to that intention.
Of course, the agent’s performance of these roles probably consists in the occurrence of psychological states and events within him. To insist that the story mention only the agent himself as the object of rational influence, or as the author and executor of intentions, would be to assume a priori that there is no psychological reduction of what happens in rational action. One is surely entitled to hypothesize, on the contrary, that there are mental states and events within an agent whose causal interactions constitute his being influenced by a reason, or his forming and conforming to an intention.
True enough. But the states and events described in a psychological reduction of a fully human action must be such that their interactions amount to the participation of the agent. My objection to the standard story is not that it mentions mental occurrences in the agent instead of the agent himself; my objection is that the occurrences it mentions in the agent are no more than occurrences in him, because their involvement in an action does not add up to the agent’s being involved.
How can I tell that the involvement of these mental states and events is not equivalent to the agent’s? I can tell because, as I have already suggested, the agent’s involvement is defined in terms of his interactions with these very states and events, and the agent’s interactions with them are such as they couldn’t have with themselves. His role is to intervene between reasons and intention, and between intention and bodily movements, in each case guided by the one to produce the other. And intervening between these items is not something that the items themselves can do. When reasons are described as directly causing an intention, and the intention as directly causing movements, not only has the agent been cut out of the story but so has any psychological item that might play his role.
At this point, defenders of the standard story might wish to respond that it includes the agent implicitly, as the subject of the mental and physiological occurrences that it explicitly describes. The reasons, intention, and movements mentioned in the story are modifications of the agent, and so their causal relations necessarily pass through him. Complaining that the agent takes no part in causal relations posited between reasons and intention, they might claim, is like complaining that the ocean takes no part in causal relations posited between adjacent waves.
But reflection on the phenomena of action reveals that being the subject of causally related attitudes and movements does not amount to participation of the sort appropriate to an agent. As Harry Frankfurt has pointed out, an agent’s desires and beliefs can cause a corresponding intention despite him, and hence without his participation.  When an addict’s desire for a drug causes his decision to take it, Frankfurt reminds us, “he may meaningfully make the analytically puzzling [statement] that the force moving him to take the drug is a force other than his own” (p. 18), and so he may be “a helpless bystander to the forces that move him” (p. 21). Similarly, an agent can fail to participate when his intention causes bodily movements. A frequently cited example is the assassin whose decision to fire on his target so unnerves him as to make his trigger-finger twitch, causing the gun to fire. In such a case, the agent’s intention has caused corresponding movements of his body, but it has done so without the agent’s participation.
Proponents of the standard story believe that the agent’s participation is lacking from these cases only because the train of causes leading from his motives to his intention, or from his intention to his behavior, is somehow abnormal. They therefore deny that these cases demonstrate the inadequacy of the standard story. The story is committed only to the claim that the causal sequence from motives to behavior will involve the agent himself when it proceeds in the normal way.
In my view, however, the discussion of “deviant” causal chains has diverted attention from simpler counterexamples, which omit the agent without lapsing into causal deviance; and it has thereby engendered a false sense of confidence in the requirement of causal normality as sufficient to protect the standard story from counterexamples. In reality, an agent can fail to participate in his behavior even when it results from his motives in the normal way. Consequently, no definition of causal normality will fix what ails the standard story.
Suppose that I have a long-anticipated meeting with an old friend for the purpose of resolving some minor difference; but that as we talk, his offhand comments provoke me to raise my voice in progressively sharper replies, until we part in anger. Later reflection leads me to realize that accumulated grievances had crystallized in my mind, during the weeks before our meeting, into a resolution to sever our friendship over the matter at hand, and that this resolution is what gave the hurtful edge to my remarks. In short, I may conclude that desires of mine caused a decision, which in turn caused the corresponding behavior; and I may acknowledge that these mental states were thereby exerting their normal motivational force, unabetted by any strange perturbation or compulsion. But do I necessarily think that I made the decision or that I executed it? Surely, I can believe that the decision, though genuinely motivated by my desires, was thereby induced in me but not formed by me; and I can believe that it was genuinely executed in my behavior but executed, again, without my help. Indeed, viewing the decision as directly motivated by my desires, and my behavior as directly governed by the decision, is precisely what leads to the thought that as my words became more shrill, it was my resentment speaking, not I.
Of course, to say that I was not involved in the formation and execution of my intention is to concede that these processes were abnormal in some sense. My point, however, is that they were not abnormal in respect to the causal operation of the motives and intention involved. When my desires and beliefs engendered an intention to sever the friendship, and when that intention triggered my nasty tone, they were exercising the same causal powers that they exercise in ordinary cases, and yet they were doing so without any contribution from me. Hence what constitutes my contribution, in other cases, cannot be that these attitudes are manifesting their ordinary causal powers. When I participate in an action, I must be adding something to the normal motivational influence of my desires, beliefs, and intentions; and so a definition of when their influence is normal still won’t enable the standard story to account for my participation.
In omitting the agent’s participation from the history of his action, the standard story falls victim to a fundamental problem in the philosophy of action—namely, that of finding a place for agents in the explanatory order of the world. Our concept of full-blooded human action requires some event or state of affairs that owes its occurrence to an agent and hence has an explanation that traces back to him. As I have already noted, not all actions are full-blooded—witness the aforementioned raising of my voice, which owed its occurrence to my attitudes but not to me. Such an occurrence may still count as the behavioral component of an action, as something that I did; but it lacks those features which seem to set human action apart from the rest of animal behavior, and which thus provide the philosophy of action with its distinctive subject matter. What makes us agents rather than mere subjects of behavior—in our conception of ourselves, at least, if not in reality—is our perceived capacity to interpose ourselves into the course of events in such a way that the behavioral outcome is traceable directly to us.
The question whether our practical nature is as we conceive it in this respect—or in any other, for that matter—should be clearly distinguished from the question what we conceive our practical nature to be. Carl Ginet has recently argued that what happens when someone acts is that his behavior is caused by a mental event whose intrinsic qualities include feeling as if it issues directly from him; but that this feeling corresponds to no actual feature of the event’s causal history or structure.  Even if Ginet’s account correctly describes what actually happens in all or most of the episodes that we describe as actions, the question remains whether it correctly expresses what we mean to say about those episodes in so describing them.
Indeed, Ginet’s account strongly suggests that what we mean to say about an event, in calling it an action, is unlikely to be what the account itself says, since it says that an action begins with a mental event that feels as if it were something that, according to this account, it is not—namely, a direct production of the agent. If our actions always begin with mental events that feel as if they are of agential origin, then one might expect the notion of agential origin to crop up in our commonsense concept of action; whereas one wouldn’t expect a commonsense concept to include the philosophical critique of this notion, as having no realization in the history or structure of events. Ginet’s account therefore suggests that we are likely to conceive actions as traceable to the agent in a sense in which, according to Ginet, they actually are not.
Of course, if actions can fail to be as we conceive them, then the philosopher of action must specify whether his object of study is the concept or the reality. Does the philosopher seek to explain what we ordinarily mean when we call something an action, or does he seek to explain what something ordinarily is when so called? My aim is to explain the former, at least in the first instance. For I suspect that our practices of deliberation, rationalizing explanation, and moral assessment are designed for action as we conceive it to be, and that any account of a reality substantially different from this conception will not help us to understand the logic of these practices.
In saying that my aim is to explicate our concept of action, as opposed to the reality, I do not mean to imply that I have given up hope of finding that the two are in accord. All I mean is that the concept has an antecedently fixed content that doesn’t depend on what actually goes on in all or most or even a privileged few of the cases to which it’s applied, and hence that correspondence between concept and reality will count as a cognitive achievement on our part. As for this cognitive achievement, however, I do hope to show that we need not despair of having attained it. For I hope to show that our concept of full-blooded action, as involving behavior that’s ultimately traceable to an agent, can be understood in a way that may well be realized in the world, as we otherwise understand it.
The obstacle to reconciling our conception of agency with the possible realities is that our scientific view of the world regards all events and states of affairs as caused, and hence explained, by other events and states, or by nothing at all. And this view would seem to leave no room for agents in the explanatory order. As Thomas Nagel puts it, “Everything I do or that anyone else does is part of a larger course of events that no one ‘does’, but that happens, with or without explanation. Everything I do is part of something I don’t do, because I am a part of the world.”
I implicitly endorsed this naturalistic conception of explanation when I conceded, earlier, that the standard story of action cannot be faulted merely for alluding to states and events occurring in the agent’s mind. Any explanation of human action will speak in terms of some such occurrences, because occurrences are the basic elements of explanation in general.
Some philosophers have not been willing to concede this point. According to Roderick Chisholm, for example, the explanatory order must include not only occurrences but also agents, conceived as additional primitive elements.  The causation of occurrences by agents, rather than by other occurrences, is what Chisholm calls “agent-causation”.
If the phrase ‘agent-causation’ is understood in Chisholm’s sense, then the naturalistic conception of explanation implies that agent-causation doesn’t exist. Yet those who endorse the naturalistic conception of explanation, as I do, may still want to reconcile it with our commonsense conception of full-blooded action, in which behavior is traced to the agent himself rather than to occurrences within him. Such a reconciliation will have to show how the causal role assigned to the agent by common sense reduces to, or supervenes on, causal relations among events and states of affairs. And the agent’s being a supervenient cause of this sort might also be called agent-causation, in a more relaxed sense of the phrase. If ‘agent-causation’ is understood to encompass this possibility as well as the one envisioned by Chisholm, then naturalists may want a theory of agent-causation, too.
This broader understanding of the phrase ‘agent-causation’ is in fact endorsed by Chisholm himself, in a passage whose obscure provenance justifies extended quotation. Chisholm says: 
[T]he issues about “agent-causation” . . . have been misplaced. The philosophical question is not—or at least it shouldn’t be—the question whether or not there is “agent-causation”. The philosophical question should be, rather, the question whether “agent-causation” is reducible to “event causation”. Thus, for example, if we have good reason for believing that Jones . . . kill[ed] his uncle, then the philosophical question about Jones as cause would be: Can we express the statement “Jones killed his uncle” without loss of meaning into a set of statements in which only events are said to be causes and in which Jones himself is not said to be the source of any activity? And can we do this without being left with any residue of agent-causation—that is, without being left with some such statement as “Jones raised his arm” wherein Jones once again plays the role of cause or partial cause of a certain event?
As the failings of the standard story reveal, we may have difficulty in meeting this challenge even if we help ourselves to a rich inventory of mental events and states. We could of course make the problem even harder, by asking how statements about Jones’s action can be reexpressed, not just in terms of occurrences, but in terms of physical occurrences taking place among particles and fields. In that case, we would be worrying, in part, about the mind-body problem. But the problem of agent-causation lingers even if the mind-body problem can be made to disappear. For let there be mental states and events in abundance—motives, reasons, intentions, plans—and let them be connected, both to one another and to external behavior, by robust causal relations; still, the question will remain how the existence and relations of these items can amount to a person’s causing something rather than merely to something’s happening in him, albeit something mental. The problem of agency is thus independent of, though indeed parallel to, the mind-body problem. Just as the mind-body problem is that of finding a mind at work amid the workings of the body, so the problem of agency is that of finding an agent at work amid the workings of the mind.
Now, Chisholm’s non-reductionist solution to the problem of agency hasn’t been taken seriously by many philosophers, nor do I intend to accord it serious attention here. However, I do sympathize with Chisholm’s complaint that those who smirk at his solution do so unjustly, since they haven’t taken seriously the problem that it is intended to solve. Chisholm says: 
Now if you can analyze such statements as “Jones killed his uncle” into event-causation statements, then you may have earned the right to make jokes about the agent as [a primitive] cause. But if you haven’t done this, and if all the same you do believe such things as that I raised my arm and that Jones killed his uncle, and if moreover you still think it’s a joke to talk about the agent as cause, then, I’m afraid, the joke is entirely on you. You are claiming the benefits of honest philosophical toil without even having a theory of human action.
Here I think that Chisholm has come as close as anyone ever has to speaking frankly about a philosophical disagreement. And I hope that he would recognize it as a token of my respect for this accomplishment if I adopt his locution and declare that the proper goal for the philosophy of action is to earn the right to make jokes about primitive agent-causation, by explaining how an agent’s causal role supervenes on the causal network of events and states.
The best sustained attempt at such an explanation, I think, is contained in a series of articles by Harry Frankfurt. These articles begin with the question of what constitutes a person, but the focus quickly narrows to the person as an element in the causal order. What primarily interests Frankfurt, as I have mentioned, is the difference between cases in which a person “participates” in the operation of his will and cases in which he becomes “a helpless bystander to the forces that move him”. And this distinction just is that between cases in which the person does and does not contribute to the production of his behavior.
In attempting to draw this distinction, Frankfurt is working on the same problem as Chisholm, although he is seeking a reductive solution rather than a solution of the non-reductive sort that Chisholm favors. What’s odd is that Frankfurt conceives of the problem in a way that initially appears destined to frustrate any reductive solution. In the following sections, I shall first explain why Frankfurt’s project can thus appear hopeless; and I shall then suggest a conception of agency that might offer Frankfurt some hope.
Frankfurt’s strategy for identifying the elements of agent-causation is to identify what’s missing from cases in which human behavior proceeds without the agent as its cause. Frankfurt figures that if he can find what’s missing from instances of less-than-full-blooded action, then he’ll know what makes it the case, in other instances, that the agent gets into the act.
The cases of defective action that occupy Frankfurt’s attention are cases in which the agent fails to participate because he is “alienated” from the motives that actuate him and which therefore constitute his will, or (as Frankfurt calls it) his “volition”. And what’s missing when an agent is alienated from his volition, according to Frankfurt, is his “identifying” or “being identified” with it.
Although Frankfurt draws this observation from cases in which the agent consciously dissociates himself from the motives actuating him—cases involving addiction or compulsion—it can equally be drawn from cases of the more familiar sort that I illustrated above. When my latent resentments against a friend yield an intention that causes my voice to rise, for example, I am not consciously alienated from that intention, perhaps, but I do not identify with it, either, since I am simply unaware of it. Hence Frankfurt might say that I do not participate in raising my voice because, being unaware of my intention, I cannot identify with it.
From this analysis of defective actions, Frankfurt draws the conclusion that what makes the difference between defective and full-blooded actions must be that, in the case of the latter, the agent identifies with the motives that actuate him. Here Frankfurt casts the agent in a role of the general sort that I envisioned in my critique of the standard story. That is, he doesn’t think of the agent as entering the causal history of his action by displacing the motivational force of his desires or intentions; rather, he thinks of the agent as adding to the force of these attitudes, by intermediating among them. Specifically, the agent interacts with his motives, in Frankfurt’s conception, by throwing his weight behind some of them rather than others, thereby determining which ones govern his behavior.
Frankfurt thus arrives at the conclusion that if a causal account of action is to include the agent’s contribution to his behavior, it must include the agent’s identifying himself with his operative motives. He therefore looks for mental events or states that might constitute the agent’s self-identification.
Frankfurt’s first candidate for the role is a second-order motive. The agent’s identifying with the motive that actuates him, Frankfurt suggests, consists in his having a second-order desire to be actuated by that motive, whereas his being alienated from the motive consists in his having a desire not to be so actuated. These higher-order desires either reinforce or resist the influence of the agent’s operative motive, and they thereby “constitute his activity”—that is, his throwing his weight behind, or withholding his weight from, the motive that actuates him, and thereby making or withholding a contribution to the resulting behavior (p. 54).
As Gary Watson has pointed out and Frankfurt has conceded, however, the same considerations that show the standard story to be incomplete can be applied to this enhanced version of it. For just as an agent can be alienated from his first-order motives, so he can be alienated from his second-order desires about them; and if his alienation from the former entails that they operate without his participation, then his alienation from the latter must entail similar consequences. Yet if the agent doesn’t participate when a second-order desire reinforces his operative motive, then how can its doing so constitute his identifying with that motive and contributing to the resulting behavior? The occurrence that supposedly constitutes the agent’s contributing to his behavior seems itself to stand in need of some further contribution from him. Hence Frankfurt has failed to identify a mental item that necessarily implicates the agent in producing his behavior.
Watson and Frankfurt have subsequently sought alternative candidates for the role. Watson argues that Frankfurt’s references to second-order desires should be replaced with references to the agent’s values. What is distinctive about behavior in which the agent isn’t fully involved, according to Watson, “is that the desires and emotions in question are more or less radically independent of [his] evaluational systems.” Watson therefore suggests that the agent’s contribution to an action is the contribution made by his system of values.
But this suggestion solves nothing. A person can be alienated from his values, too; and he can be alienated from them even as they continue to grip him and to influence his behavior—as, for instance, when someone recoils from his own materialism or his own sense of sin. Hence the contribution of values to the production of someone’s behavior cannot by itself be sufficient to constitute his contribution, for the same reason that the contribution of his second-order desires proved insufficient.
Frankfurt has made an attempt of his own to solve the problem, in subsequent papers, but with no more success. Frankfurt now suggests that the agent’s involvement in his behavior can be provided by “decisions” or “decisive commitments” to his operative motives, since these mental items are indivisible from the agent himself. Frankfurt writes, “Decisions, unlike desires or attitudes, do not seem to be susceptible both to internality and to externality”—that is, to identification and alienation—and so “[i]invoking them . . . would appear to avoid . . . the difficulty” (p. 68, n. 3). Yet the example of my unwitting decision to break off a friendship shows that even decisions and commitments can be foreign to the person in whom they arise. How, then, can a decision’s contribution to behavior guarantee that the agent is involved?
One might wonder, of course, why Frankfurt and Watson assume that the agent’s identifying with his operative motives must consist in a mental state or event specifiable in other terms, as a particular kind of desire, value, or decision. Perhaps identifying with one’s motives is a mental state or event sui generis rather than a species of some other genus.
Tempting though this suggestion may be, it is really just an invitation to beg the question of agent-causation. The question, after all, is how an agent causally contributes to the production of his behavior; and to observe that he sometimes identifies with the motives producing that behavior is to answer this question only if identifying with motives entails somehow making a causal contribution to their operation—throwing one’s weight behind them, as I put it before. Other kinds of identification may not at all guarantee that the agent gets into the act.
Frankfurt seems to think that an agent cannot fail to get into the act when he identifies with a motive. “It makes no sense”, he says, “to ask whether someone identifies himself with his identification of himself, unless this is intended simply as asking whether his identification is wholehearted or complete” (p. 54). What this remark shows, however, is that Frankfurt is using the term ‘identification’ in a specialized sense, since ordinary talk of identifying with something often denotes a mental event or state from which the subject can indeed be alienated. For example, you may find yourself identifying with some character in a trashy novel, even as you recoil from this identification. Identifying with the character may then seem like something that happens to you, or comes over you, without your participation.
One might think that such a case is what Frankfurt has in mind when he says that an agent’s identification of himself may not be “wholehearted” or “complete”, but I think not. For if it were, then Frankfurt would in effect be conceding that self-identification can sometimes occur without the agent’s participation; and in that case, he could no longer claim that self-identification alone is what distinguishes the actions in which the agent participates from those in which he doesn’t. An agent who identifies with a motive needn’t be implicated in the behavior that it produces if he can somehow dissociate himself from the identification.
I think that what Frankfurt means, when he refuses to ask whether someone identifies with his self-identification, is that identifying oneself with a motive is unlike identifying with a character in a novel precisely in that it cannot happen at all without one’s participation. Identifying with another person is, at most, a matter of imagining oneself in his skin, whereas identifying with a motive entails taking possession of it in fact, not just in imagination. Frankfurt therefore assumes, I think, that identifying with a motive is a mental phenomenon that simply doesn’t occur unless one participates, although one may participate halfheartedly or incompletely.
Having put our finger on this assumption, however, we can see that for Frankfurt to posit self-identification as a primitive mental phenomenon would be to beg the question of agent-causation. For if self-identification is something that cannot occur without the agent’s contributing to it, then it cannot occur without agent-causation, and we cannot assume that it occurs without assuming that agent-causation occurs—which is what we set out to show, in the first place. The question is whether there is such a thing as a person’s participating in the causal order of events and states, and we can’t settle this question simply by positing a primitive state or event that requires the person’s participation.
Lest the question be begged, then, “self-identification” must not be understood as naming the primitive event or state that provides the needed reduction of agent-causation; it must be understood, instead, as redescribing agent-causation itself, the phenomenon to be reduced. When Frankfurt says that an agent participates in an action by identifying with its motives, he doesn’t mean that self-identification is, among mere states and events, the one in virtue of which the agent gets into the act; rather, he is saying that if we want to know which are the mere states and events that constitute the agent’s getting into the act, we should look for the ones that constitute his identifying with his motives. Frankfurt and Watson are therefore correct in trying to reduce self-identification to desires, values, or decisions—that is, to mental phenomena whose existence we can assume without presupposing that agent-causation occurs.
But how can such a reduction ever succeed? If we pick out mental states and events in terms that do not presuppose any causal contribution from the agent, then we shall have picked out states and events from which the agent can in principle dissociate himself. Since the occurrence of these items will be conceptually possible without any participation from the agent, we shall have no grounds for saying that their occurrence guarantees the agent’s participation in the causal order.
The only way to guarantee that a mental state or event will bring the agent into the act is to define it in terms that mandate the agent’s being in the act; but then we can’t assume the occurrence of that state or event without already assuming the occurrence of agent-causation. Hence we seem to be confronted with a choice between begging the question and not answering it at all.
We may be tempted to slip between the horns of this dilemma by characterizing some mental items in terms that are sufficiently vague to carry an assumption of agent-causation while keeping that assumption concealed. I suspect that Watson’s appeal to “the agent’s system of values” and Frankfurt’s appeal to “decisive commitments” seem to succeed only insofar as they smuggle such an assumption into the story. But a genuine resolution of the dilemma will require a more radical change of approach.
The main flaw in Frankfurt’s approach, I think, is that substituting one instance of agent-causation for another, as the target of reduction, does not advance the reductionist project. Since self-identification won’t serve our purpose unless it’s conceived as something to which the agent contributes, rather than something that happens to him, reducing self-identification to mere events and states is unlikely to be any easier than reducing action itself
The way to advance the reductionist project is, not to substitute one agent-causal phenomenon for another as the target of reduction, but to get the process of reduction going, by breaking agent-causation into its components. And surely, the principal component of agent-causation is the agent himself. Instead of looking for mental events and states to play the role of the agent’s identifying with a motive, then, we should look for events and states to play the role of the agent.
Something to play the role of agent is precisely what I earlier judged to be lacking from the standard story of human action. I pointed out that the agent intermediates in various ways between his reasons and intentions, or between his intentions and bodily movements; and I argued that the standard story omits the agent, not because it fails to mention him by name, but rather because it fails to mention anything that plays his intermediating role.
What plays the agent’s role in a reductionist account of agent-causation will of course be events or states—most likely, events or states in the agent’s mind. We must therefore look for mental events and states that are functionally identical to the agent, in the sense that they play the causal role that ordinary parlance attributes to him.
Looking for a mental event or state that’s functionally identical to the agent is not as bizarre as it sounds. Of course, the agent is a whole person, who is not strictly identical with any subset of the mental states and events that occur within him. But a complete person qualifies as an agent by virtue of performing some rather specific functions, and he can still lay claim to those functions even if they are performed, strictly speaking, by some proper part of him. When we say that a person digests his dinner or fights an infection, we don’t mean to deny that these functions actually belong to some of his parts. A person is a fighter of infections and a digester of food in the sense that his parts include infection-fighting and food-digesting systems. Similarly, a person may be an initiator of actions—and hence an agent—in the sense that there is an action-initiating system within him, a system that performs the functions in virtue of which he qualifies as an agent and which are ordinarily attributed to him in that capacity. A reductionist philosophy of action must therefore locate a system of mental events and states that perform the functional role definitive of an agent.
I sometimes suspect that Frankfurt sees the necessity of this approach and may even think that he’s taking it. My suspicion is based on the potential confusions that lurk in Frankfurt’s talk of “identifying oneself” with a motive and thereby “making it one’s own”. The reader, and perhaps the writer, of these phrases may think that when a person identifies himself with motives, they become functionally identical to him, or that when motives become his, they do so by becoming him, in the sense that they occupy his functional role. But the psychological items that are functionally identical to the agent, in the sense that they play the causal role attributed to him in his capacity as agent, cannot be items with which he identifies in Frankfurt’s sense, because identifying with something, in that sense, is a relation that one bears to something functionally distinct from oneself. The agent’s identifying with an attitude requires, not only something to play the role of the attitude identified with, but also something else to play the role of the agent identifying with it; and the latter item, rather than the former, will be what plays the functional role of the agent and is therefore functionally identical to him.
What, then, is the causal role that mental states and events must play if they are to perform the agent’s function? I have already outlined what I take to be the causal role of an agent; but for the remainder of this chapter, I want to confine my attention to that aspect of the role which interests Frankfurt, since my approach is simply a modification of his. Frankfurt doesn’t think of the agent as having a function to play in implementing his own decisions, nor does he think of the agent as interacting with reasons per se. Frankfurt focuses instead on the agent’s interactions with the motives in which his reasons for acting are ordinarily thought to consist. The agent’s role, according to Frankfurt, is to reflect on the motives competing for governance of his behavior, and to determine the outcome of the competition, by taking sides with some of his motives rather than others. For the moment, then, I shall adopt Frankfurt’s assumption that the agent’s role is to adjudicate conflicts of motives (though I shall subsequently argue that such adjudication is best understood as taking place among reasons instead).
Which mental items might play this role? Here, too, I want to begin by following Frankfurt. Frankfurt says that adjudicating the contest among one’s motives entails occupying an “identity apart” from them (p. 18); and he says this, I assume, because a contest cannot be adjudicated by the contestants themselves. When an agent reflects on the motives vying to govern his behavior, he occupies a position of critical detachment from those motives; and when he takes sides with some of those motives, he bolsters them with a force additional to, and hence other than, their own. His role must therefore be played by something other than the motives on which he reflects and with which he takes sides.
Indeed, the agent’s role is closed, not only to the actual objects of his critical reflection, but to all potential objects of it as well. Even when the agent’s reflections are confined to his first-order motives, for example, his second-order attitudes toward them cannot be what play his role; for he can sustain his role as agent while turning a critical eye on those second-order attitudes, whereas they cannot execute such a critical turn upon themselves. The functional role of agent is that of a single party prepared to reflect on, and take sides with, potential determinants of behavior at any level in the hierarchy of attitudes; and this party cannot be identical with any of the items on which it must be prepared to reflect or with which it must be prepared to take sides.
Thus, the agent’s role cannot be played by any mental states or events whose behavioral influence might come up for review in practical thought at any level. And the reason why it cannot be played by anything that might undergo the process of critical review is precisely that it must be played by whatever directs that process. The agent, in his capacity as agent, is that party who is always behind, and never solely in front of, the lens of critical reflection, no matter where in the hierarchy of motives it turns.
What mental event or state might play this role of always directing and never merely undergoing such scrutiny? It can only be a motive that drives practical thought itself. That is, there must be a motive that drives the agent’s critical reflection on, and endorsement or rejection of, the potential determinants of his behavior, always doing so from a position of independence from the objects of review. Only such a motive would occupy the agent’s functional role, and only its contribution to his behavior would constitute his own contribution.
What I’m positing here is an attitude that embodies the concerns of practical thought per se, concerns distinct from those embodied in any of the attitudes that practical thought might evaluate as possible springs of action. Frankfurt seems to assume that the concerns animating the agent’s critical reflection on his first-order motives are embodied in his second-order desires about whether to be governed by those motives—such as the desire not to act out of anger, for example, or the desire to be actuated by compassion instead. Yet these second-order desires figure in critical reflection only with respect to a particular conflict of motives, and they can themselves become the objects of critical reflection one step further up the attitudinal hierarchy. Hence the concerns that they embody cannot qualify as the concerns directing practical thought as such, concerns that must be distinct from the objects of critical reflection and that must figure in such reflection whenever it occurs. If we want to find the concerns of practical thought per se, we must find motives that are at work not only when the agent steps back and asks whether to act out of anger but also when he steps back further and asks whether to restrain himself out of shame about his anger, and so on. Only attitudes that are at work in all such instances of reflection will be eligible to play the role of agent, who himself is at work whenever critical reflection takes place.
One is likely to balk at this proposal if one isn’t accustomed to the idea that practical thought is propelled by a distinctive motive of its own. Agency is traditionally conceived as a neutral capacity for appraising and exercising motives—a capacity that’s neutral just in the sense that it is not essentially animated by any motive in particular. This traditional conception is not hospitable to the idea that the deliberative processes constitutive of agency require a distinctive motive of their own. My point, however, is that anyone who wants to save our ordinary concept of full-blooded action as involving behavior caused by the agent had better grow accustomed to this idea, because the problem of agent-causation cannot be solved without it. Some motive must be behind the processes of practical thought—from the initial reflection on motives, to the eventual taking of sides; and from second-order reflection to reflection at any higher level—since only something that was always behind such processes would play the causal role that’s ordinarily attributed to the agent.
Is there in fact such a motive? I believe so, though it is not evident in Frankfurt’s account. Frankfurt’s conception of critical reflection strikes me as omitting a concern that’s common to reflection in all instances and at all levels.
The agent’s concern in reflecting on his motives, I believe, is not just to see which ones he likes better; it’s to see which ones provide stronger reasons for acting, and then to ensure that they prevail over those whose rational force is weaker. What animates practical thought is a concern for acting in accordance with reasons. And I suggest that we think of this concern as embodied in a desire that drives practical thought.
When I speak of a desire to act in accordance with reasons, I don’t have a particular desire in mind; any one of several different desires would fill the bill. On the one hand, it could be a desire to act in accordance with reasons so described; that is, the de dicto content of the desire might include the concept of reasons. On the other hand, it could be a desire to act in accordance with considerations of some particular kind, which happened to be the kind of consideration that constituted a reason for acting. For example, I have argued elsewhere that rational agents have a desire to do what makes sense, or what’s intelligible to them, in the sense that they could explain it; and I have argued that reasons for a particular action are considerations by which the action could be explained and in light of which it would therefore make sense.  Thus, if someone wants to do what makes sense, then in my view he wants to act in accordance with reasons, though not under that description. In any of its forms, the desire to act in accordance with reasons can perform the functions that are attributed to its subject in his capacity as agent. We say that the agent turns his thoughts to the various motives that give him reason to act; but in fact, the agent’s thoughts are turned in this direction by the desire to act in accordance with reasons. We say that the agent calculates the relative strengths of the reasons before him; but in fact, these calculations are driven by his desire to act in accordance with reasons. We say that the agent throws his weight behind the motives that provide the strongest reasons; but what is thrown behind those motives, in fact, is the additional motivating force of the desire to act in accordance with reasons. For when a desire appears to provide the strongest reason for acting, then the desire to act in accordance with reasons becomes a motive to act on that desire, and the desire’s motivational influence is consequently reinforced. The agent is moved to his action, not only by his original motive for it, but also by his desire to act on that original motive, because of its superior rational force. This latter contribution to the agent’s behavior is the contribution of an attitude that performs the functions definitive of agency; it is therefore, functionally speaking, the agent’s contribution to the causal order.
What really produces the bodily movements that you are said to produce, then, is a part of you that performs the characteristic functions of agency. That part, I claim, is your desire to act in accordance with reasons, a desire that produces behavior in your name by adding its motivational force to that of whichever motives appear to provide the strongest reasons for acting, just as you are said to throw your weight behind them.
Note that the desire to act in accordance with reasons cannot be disowned by an agent, although it can be disowned by the person in whom agency is embodied. A person can perhaps suppress his desire to act in accordance with reasons; but in doing so, he will have to execute a psychic manoeuvre quite different from suppressing his anger or his addiction to drugs or his other substantive motives for acting. In suppressing his anger, the person operates in his capacity as agent, rejecting anger as a reason for acting; whereas in suppressing his desire to act in accordance with reasons, he cannot reject it as a reason for acting, or he will in fact be manifesting his concern for reasons rather than suppressing it, after all. The only way for a person truly to suppress his concern for reasons is to stop making rational assessments of his motives, including this one, thus suspending the processes of practical thought. And in suspending the processes of practical thought, he will suspend the functions in virtue of which he qualifies as an agent. Thus, the sense in which an agent cannot disown his desire to act in accordance with reasons is that he cannot disown it while remaining an agent.
Conversely, a person’s desire to act in accordance with reasons cannot operate in him without its operation’s being constitutive of his agency. What it is for this motive to operate is just this: for potential determinants of behavior to be critically reviewed, to be embraced or rejected, and to be consequently reinforced or suppressed. Whatever intervenes in these ways between motives and behavior is thereby playing the role of the agent and consequently is the agent, functionally speaking. Although the agent must possess an identity apart from the substantive motives competing for influence over his behavior, he needn’t possess an identity apart from the attitude that animates the activity of judging such competitions. If there is such an attitude, then its contribution to the competition’s outcome can qualify as his—not because he identifies with it but rather because it is functionally identical to him.
Note, finally, that this reduction of agent-causation allows us to preserve some aspects of commonsense psychology about which we may have had philosophical qualms. What we would like to think, pre-philosophically, is that a person sometimes intervenes among his motives because the best reason for acting is associated with the intrinsically weaker motive, and he must therefore intervene in order to ensure that the weaker motive prevails. What inhibits us from saying this, however, is the philosophical realization that the weaker motive can never prevail, since an incapacity to prevail over other motives is precisely what constitutes motivational weakness. Every action, we are inclined to say, is the result of the strongest motive or the strongest combination of motives, by definition.
But my reduction of agent-causation enables us to say both that the agent makes the weaker motive prevail and that the contest always goes to the strongest combination of motives. The agent can make the weaker motive prevail, according to my story, in the sense that he can throw his weight behind the weaker of those motives which are vying to animate his behavior and are therefore objects of his practical thought. But the agent’s throwing his weight behind the weaker of these motives actually consists in its being reinforced by another motive, so that the two now form the strongest combination of motives. Thus, the weaker motive can prevail with the help of the agent simply because it can prevail with the help of another motive and because the agent is another motive, functionally speaking.
Come to think of it, what else could an agent be?
1. This chapter originally appeared in Mind 101 (1992) 461–81 and is reprinted by permission of Oxford University Press. The material in this chapter was originally presented to a seminar in the philosophy of action at the University of Michigan. I am grateful to the participants in that seminar for their comments and questions. A very different paper was presented under a similar title to the philosophy departments of Yale University and the University of Dayton; this chapter shows the benefit of comments from those audiences as well. For comments on earlier drafts, I am grateful to Paul Boghossian, Sarah Buss, Daniel Cohen, John Martin Fischer, Harry Frankfurt, Carl Ginet, Brian Leiter, Connie Rosati, and several anonymous reviewers for Mind.
2. The story can be traced back at least as far as Hobbes, Leviathan, Part I, chapter vi.
3. I discuss another problem with the standard story in my “The Guise of the Good” (Chap. 4, above).
4. A critique along these lines, with special reference to Hobbes, appears in N. J. H. Dent, The Moral Psychology of the Virtues (Cambridge: Cambridge University Press, 1984), ch. 4. See, e.g., p. 99: “[A] weighty reason does not, like a weighty brick, fall upon one and impart a certain push to one’s body.”
5. Here I part company with some philosophers of action, who believe that nothing counts as an action unless the agent participates in it. See, e.g., J. Bishop, Natural Agency: An Essay on the Causal Theory of Action (Cambridge: Cambridge University Press, 1989), 41. Of course, every action must be someone’s doing and must therefore be such that an agent participates in it in the sense that he does it. But this conception of agential participation doesn’t require anything that is obviously missing from the standard story. What’s missing from that story is agential participation of a more specific kind, which may indeed be missing from doings that count as cases—albeit defective or borderline cases—of action.
6. Ibid., 72: “Intuitively, we think of agents as carrying out their intentions or acting in accordance with their practical reasons, and this seems different from (simply) being caused to behave by those intentions or reasons.”
7. See A. I. Goldman, A Theory of Human Action (Princeton: Princeton University Press, 1970), 80 ff.
8. See C. Ginet, On Action (Cambridge: Cambridge University Press, 1990), 6–7. “For a person S to cause E, it is not enough for S to be the subject of just any sort of event that causes E.”
9. The Importance of What We Care About (Cambridge: Cambridge University Press, 1988).
10. The most recent discussion of such “deviant causal chains” appears in Bishop, Natural Agency, chs. 4 and 5. See also G. Harman, “Practical Reasoning”, Review of Metaphysics 29 (1976) 445; C. Peacocke, “Deviant Causal Chains”, Midwest Studies in Philosophy 4 (1979) 124; R. Taylor, Action and Purpose (Englewood Cliffs, N.J.: Prentice-Hall, 1966), 248; Goldman, A Theory of Human Action, 54; and Donald Davidson, Essays on Actions and Events (Oxford: Clarendon Press, 1980), 79.
11. See e.g., Davidson, ibid., pp. xiii, 79, 87.
12. We can assume that this causal relation was mediated by any number of subconscious intentions—intentions to sever the friendship by alienating my friend, to alienate my friend by raising my voice, to raise my voice now . . . etc. So long as we assume that these intentions subconsciously crystallized as the conversation progressed (which is not hard to assume) we preserve the intuition that I’m currently trying to evoke—namely, that I did not participate in the resulting action. And surely, this intuition doesn’t depend on the assumption that the causal links between these intentions and my behavior weren’t “sensitive” to counterfactual differences in them (in the sense defined by Bishop, Natural Agency, ch. 5). Thus, we can conceive of cases in which reasons cause intentions, and intentions cause behavior, in all the “right ways”, and yet the agent doesn’t participate.
13. I don’t mean to suggest that these reflections absolve me of responsibility for my action. I have an obligation to be vigilant against unconsidered intentions and to keep my voice down, no matter what may be causing it to rise. The fact remains, however, that my responsibility for the action in question arises from my having failed to prevent or control it rather than from my having truly initiated it. And I am responsible for having failed to prevent or control the action because it would have yielded to various measures of self-scrutiny and self-restraint that I could have initiated. Thus, my responsibility depends on my capacity to intervene among events in a way in which I failed to intervene among my desires, intentions, and movements in this instance. If my behavior could come about only in the manner described here—that is, springing directly from intentions that have simply come over me—nothing would owe its occurrence to either my participating or failing to participate in events, and I might bear no responsibility for anything.
14. I believe that this problem is distinct from the problem of free will, although the two are often treated together. For my views on the latter problem, see “Epistemic Freedom” (Chap. 3, above).
15. On Action, 11–15. Ginet thinks that actions other than simple mental actions do issue from the agent in the sense that they involve the agent’s causing something. But he thinks that something can be caused by an agent only insofar as it is caused by one of the agent’s actions. And he thinks that the resulting regress of actions in which things are caused by other actions must terminate in a simple mental action—usually, the act of willing—which qualifies as an action only because it feels as if it was caused by the agent himself, although it hasn’t in fact been caused by him in any sense. Thus, Ginet thinks that complex actions issue from the agent only in the sense that their component behavior is ultimately caused by a mental event that misleadingly feels as if it issued from the agent. Since the agential ancestry of complex action is thus inherited from a simple mental act whose agential ancestry is itself illusory, the ancestry of all actions would seem to be tainted by illusion.
16. Here, of course, I assume that the term ‘action’ does not function like the Kripkean name of a natural kind, referring to whatever shares the essential nature of all or most or a privileged few of the episodes to which it is applied. I assume that ‘action’ has a de dicto meaning in virtue of which it may in fact fail to be a correct description of anything to which it is applied.
17. I therefore think that Ginet dismisses the causal conception of action too quickly. I do agree with Ginet that an agent, as a persisting entity, is the wrong sort of thing to cause particular events. (Ginet cites C. D. Broad, Ethics and the History of Philosophy [London: Routledge & Kegan Paul, 1952], 215, as the source of this objection.) But this objection militates only against a non-reductive theory of agent-causation. It leaves open the possibility that the causation of events by the right sort of things—that is, by other events—may in some cases amount to, or deserve to be described as, their being caused by the agent himself. It therefore leaves open the possibility of agent-causation that’s reducible to, or supervenient on, causation by events. (I discuss this possibility, and its implications, in the next section of the text.) Ginet argues against a conception that characterizes action in terms of event causation (pp. 11–13). But Ginet’s argument suffers from two flaws. Ginet’s argument is that we can conceive of a simple mental act, such as mentally saying a word, without conceiving of it as comprising a structure of distinct, causally related events. (“I mean that it is not conceptually required to have such a structure, under our concept of it as that kind of mental act” [p. 12].) Yet this point doesn’t speak to the hypothesis that we conceive of the act in question as comprising behavior caused by the agent, and that the behavior’s being caused by the agent supervenes on its causal relation to other events. Our concept of action may include agent-causation without including the supervenience base thereof. What’s more, the illustrations that Ginet provides for his argument—pairs of mental causes and effects whose structure is clearly different from that of the mental act in question—are all cases in which the imagined cause is itself a mental act. But someone who thinks that a mental act consists in mental behavior caused by the agent, in a sense that supervenes on it’s being caused by another event, is not likely to think that the causing event is yet another act.
18. T. Nagel, The View from Nowhere (Oxford: Oxford University Press, 1986), 114; cf. Bishop, Natural Agency, 39 ff.
19. R. Chisholm, Person and Object: a Metaphysical Study (London: Allen & Unwin, 1976).
20. R. Chisholm, “Comments and Replies”, Philosophia 7 (1978) 622–3.
21. Cf. Bishop, Natural Agency, 43.
22. The standard story of rational action has also illustrated that the problem is more than that of casting the agent in the role of cause. In explaining an action, we trace its history back to the agent who brought the action about; but then we trace back further, to the reasons that persuaded him to do so. And as Donald Davidson has argued (“Actions, Reason, and Causes”, in Essays on Actions and Events), the reasons cited in the explanation of an action must be, not just reasons that were available to the agent, but reasons for which he acted, the difference being precisely that the latter are the reasons that induced him to act. The reasons that explain an action are thus distinguished by their having exerted an influence upon the agent. In the explanation of an action, then, the agent must serve not only as an origin of activity, or cause, but also as an object of rational influence—and hence, in a sense, as an effect.
23. Ibid. Note the need to insert the word “primitive” in Chisholm’s phrase “the agent as cause”, which illustrates that Chisholm has reverted to understanding agent-causation in a narrower sense.
24. See Bishop, Natural Agency, 69: “Of course action differs from other behavior in that the agent brings it about, but the problem is how to accommodate such bringing about within a naturalist ontology.”
25. “Freedom of the Will and the Concept of a Person”, “Three Concepts of Free Action”, “Identification and Externality”, “The Problem of Action”, “Identification and Wholeheartedness”, all in The Importance of What We Care About (Cambridge: Cambridge University Press, 1988). Frankfurt has recently returned to the topic, in his 1991 Presidential Address to the Eastern Division of the APA, entitled “The Faintest Passion”. I shall not be discussing the new suggestions contained in this address.
26. Frankfurt says that the “essential difference between persons and other creatures” that he wishes to discuss “is to be found in the structure of a person’s will” (ibid., 12). And he later suggests that if someone becomes unable to exercise his will in the relevant way, this inability “destroys him as a person” (p. 21).
27. Ibid., 21–2. In another essay Frankfurt formulates the distinction in terms of a person’s “activity or passivity with respect to . . . states of affairs” (p. 54).
28. Ibid., pp. 18 ff., 54.
29. G. Watson, “Free Agency”, in his Free Will (Oxford: Oxford University Press, 1982), 205–20. In “The Way of the Wanton” (Chap. 11, below), I offer an interpretation of Frankfurt in which the problem pointed out by Watson doesn’t even arise. Since Frankfurt acknowledges the problem and attempts to solve it—unsuccessfully, in my view—that interpretation is probably not what he has in mind, though I think it gives the most plausible version of a view like his.
30. The Importance of What We Care About, 65–6.
31. Watson, “Free Agency”, 110.. For a recent discussion of Watson’s view, see S. Wolf, Freedom Within Reason (Oxford: Oxford University Press, 1990), ch. 2.
32. I owe the latter example to Elizabeth Anderson, “The Source of Norms” (MS).
33. Of course, Watson refers not just to values lodged in the agent but to the agent’s evaluational system; and he might argue that values are no longer integrated into that system once the agent becomes alienated from them. But in that case, Watson would simply be smuggling the concept of identification or association into his distinction between the agent’s evaluational system and his other, unsystematized values. And just as Frankfurt faced the question how a volition becomes truly the agent’s, Watson would face the question how a value becomes integrated into the agent’s evaluational system.
34. Again, the discussion that follows does not deal with Frankfurt’s 1991 Presidential Address to the Eastern Division of the American Philosophical Association, in which he outlines a somewhat different solution.
35. I can of course imagine defining a phrase ‘decisive commitments’ denoting only those commitments which an agent actively makes. In that case, decisive commitments will indeed be such as cannot fail to have the agent’s participation; but in what that participation consists will remain a mystery, and the claim that the agent participates in his actions by way of decisive commitments will be uninformative. A related criticism of Frankfurt’s solution appears in J. Christman, “Autonomy and Personal History”, Canadian Journal of Philosophy 21 (1991) 8–9.
36. See nn. 26 and 28, above.
37. Frankfurt, The Importance of What We Care About, 18.
38. This possibility may be ruled out by an argument in Bernard Williams’ paper “Internal and External Reasons”, in his Moral Luck (Cambridge: Cambridge University Press, 1981), 101–13. In any case, Williams’ argument does not rule out the alternative possibility, which is the one that I favor. I discuss Williams’ argument in “The Possibility of Practical Reason” (Chap. 7, below).
39. Practical Reflection (Princeton: Princeton University Press, 1989).
Originally published at http://dx.doi.org/10.3998/maize.13240734.0001.001
This article and its reviews are distributed under the terms of the Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and redistribution in any medium, provided that the original author and source are credited.