In 1960, H.LA. Hart published a paper entitled “Acts of Will and Responsibility,” in which he challenges the traditional understanding of the “voluntary act requirement” in Anglo-American criminal-law theory.Footnote ¹ The traditional understanding, Hart contends, had never provided a very satisfactory account of what it took to qualify as a “voluntary act.” Hart writes that he could not “find in any legal writings any clear or credible account of what it is for conduct to be voluntary and not involuntary in the sense required.” It was dubious that the courts actually relied on any general doctrine of voluntariness and it was in any case not clear that a positive account of voluntary action was actually necessary for what the theorists wanted the so-called “voluntary act requirement” to do.Footnote ² Despite Hart's criticisms, however, the traditional understanding of voluntary action has endured in both legal theory and legal doctrine.Footnote ³

The persistence of the view that criminal responsibility requires an act, as well as its methodological corollary that legal theory requires a theory of action, is somewhat puzzling, as the view faces apparently intractable flaws. Recently skepticism has been growing, and a number of authors have pointed out some of the difficulties that plague both the doctrinal view as well as criminal theory's philosophical interpretation of action.Footnote ⁴ In this paper, I join the skeptics but seek to go further in reconstructing the intuition that I believe motivates the act requirement and may help explain the doctrine's otherwise puzzling persistence. In brief, I defend the following propositions: first, that orthodox Anglo-American criminal theory (as represented by Michael Moore's updating of Austin's volitionalism) fails to explain adequately why criminal responsibility requires an act. Second, when it comes to the just definition of crimes, the act requirement is at best a rough generalization rather than a substantive limiting principle. Third, that what is sound in the intuition underlying the so-called “act requirement” is better explained by what I call the “practical-agency condition,” according to which punishment in a specific instance is unjust unless the crime charged was caused or constituted by the agent's conduct (broadly understood) qua practically rational agent. I argue that the practical-agency condition embodies what is worth retaining in Anglo-American criminal law's notion of an “act requirement.”

I.

Anglo-American criminal theory has inherited a philosophical account of action that stems from Austin's Lectures, though it has roots going back at least to Hobbes.Footnote ⁵ According to this line of thought, a human action is a bodily movement (sometimes a “muscular contraction”) caused by a type of mental state called a “volition.”Footnote ⁶ The difference, in other words, between my arm moving and my moving of my arm is that in the latter case the movement is caused by a volition–a determination of the will–whereas in the former case the movement might be nothing more than the result of being pushed. The propositional object of a volition was taken to be merely “that my body move thus-and-so” or “that this muscle contract this far,” and specifically not “that I shoot A” or “that I open the door.” Oliver Wendell Holmes picks up this Austinian line of thought when he suggests that all it takes to satisfy the act requirement is that the agent have willed a muscular contraction and that the agent need not have conceived of the muscular contraction as an instance of a particular type of action, as taking place in a certain context, or as having any particular consequences whatsoever. All of those features are to be considered circumstances and consequences, and they figure into responsibility only insofar as the “reasonable man” should have known that the circumstances were as they were and that contracting his muscles or moving his limbs in those circumstances would have those consequences.Footnote ⁷

Postulating an act requirement so defined thus resolves two issues in legal theory in one fell swoop: first, it explains at least some of the exemptions from responsibility recognized at law (unconsciousness, seizure, physical coercion), and second, it explains why Anglo-American law has been hesitant to consider omissions, thoughts, and statuses as proper objects of criminal regulation. Defining actions in this manner thus has several advantages for a theorist interested in systematizing criminal law. Notably, it provides a unified view of criminal liability from what can be called both the legislative and the judicial points of view: a statute that purports to criminalize something other than a willed bodily movement is unjust, as is a judicial proceeding that punishes a person for something other than willed bodily movement, even if the applicable statute is facially just.Footnote ⁸

But why define actions in terms of willed bodily movement? Because, I suspect, doing so allows the theorist then to assign a clear role to mens rea doctrine: acts describe what can possibly be criminalized in the first place, and mens rea standards describe which acts are criminalized. Since the degree and kind of fault required differs in different cases–negligently dropping a brick from a rooftop and intentionally releasing a brick with intent to kill are both possible objects of criminal liability–focusing on willed bodily movement seems promising as one thing every crime would have in common.Footnote ⁹ Beliefs regarding circumstances, consequences, motives, and the like are relevant in selecting which cases from this set are actually to be punished. This suggests that determining for legal purposes whether a person has “acted” is morally neutral in the sense that it does not presuppose a moral assessment of the agent or the act. Where moral and legal assessment starts is with the standards applied to the awareness of consequences, circumstances, and motivations present in the agent's psychological set at the time of the action. In other words, to say that A acted is not itself to assign responsibility to A; the attribution of the resulting harm to the agent arises only when the act is described as accompanied by, say, negligence or criminal intent.Footnote ¹⁰ Something along these lines is, I take it, what Holmes has in mind when he describes acts as “indifferent.”

Were it not for Michael Moore's exhaustive rehabilitation of the willed-bodily-movement account of action, Austin's and Holmes's remarks might today be mainly of antiquarian interest.Footnote ¹¹ However, in Act and Crime–the most systematic and detailed account of the concept of action in the criminal law currently available–Moore undertakes an ambitious rearticulation and defense of the classical account of action, arguing both for its independent philosophical plausibility as well as its centrality to notions of criminal responsibility.Footnote ¹² There, Moore claims that his “functionalist” reinterpretation of the willed-bodily-movement view “defuses” Hart's principle objection to Austin's account, thereby vindicating Austin and orthodox criminal theory.Footnote ¹³ I now turn to consider Moore's argument.

Hart claims that, contrary to what Austin's account would lead one to expect, we are in fact not generally aware of ourselves as willing discrete muscular contractions or even movements of our limbs as such. Rather, our conscious and deliberative activity is directed to the things we do by moving our bodies; saying “hello,” for instance, rather than moving one's lips, taking the purse rather than contracting fingers. To be sure, under particular circumstances, we can and do focus our attention and reflection on our limbs and muscular contractions–for example, if I am learning to pronounce the French “r” correctly, learning how to play a difficult piano passage, or striving to correct defects in my butterfly stroke. But by and large, Hart writes:

Therefore the characterization of an agent as necessarily willing bodily movement or muscular contraction “is a fatal defect in any account of action supposed to help us to characterize the mental conditions required for the ordinary man's responsibility.”Footnote ¹⁵

In response, Moore claims that Hart's objection gets no traction against a functionalist interpretation of Austin. Moore agrees with Austin that the difference between bodily movements that are and those that are not actions lies in their causal histories: actions are those movements caused by a particular type of mental state, namely volitions. Volitions are a type of bare intention the content of which concerns nothing but bodily movements. But the fact that we do not normally deliberate about, form intentions regarding, or generally have attitudes toward muscular contractions does not show that there are no volitions; it just means that volitions should not be understood as (conscious) intentional states but as functionally defined mediators between intentional states (beliefs and desires, for instance) and bodily movements. A functionalist in this sense can be noncommittal about how those mediating volitions are realized, as conscious phenomena, brain states, or anything else. As Moore notes, the nature of functional states “is given by their functional roles and not by their accessibility to consciousness. If one has good theoretical reasons to hypothesize a state that mediates between our general preferences and our particular actions, the lack of phenomenological evidence will not be critical.”Footnote ¹⁶ Volitions thus understood bear no intrinsic connection to “phenomenology” or deliberative awareness; we need no more be aware of volitions, functionally construed, than we are of the subatomic particles our bodies are made of or the Freudian id and superego that play functional roles in our cognitive and emotional lives. Therefore, Moore concludes, Hart's objection that volitions do not enter into deliberative awareness falls wide of the mark.

Has Moore vindicated Austin? I agree with Moore that it is no objection to his account that the volitions it postulates form no part of a person's ordinary deliberative activity. What is an objection, however, is that his functionalist account has severed the connection that made the theory seem relevant to legal and moral responsibility in the first place. For it is quite obscure what the normative consequences of the presence or absence of volition, functionally understood as mediating between attitudes and bodily movements, are supposed to be for how we assess a person's responsibility. As phenomena that do not have a role in the deliberative economy, volitions are essentially subpersonal states, mental states or events that may be part of the machinery involved in executing intentions, plans, and so forth (e.g., as proximate cause of certain muscular contractions). But they do not themselves constitute ordinary intentions, plans or the like; they are not, for instance, reasons or desires, attitudes that are generally concerned with objects and goals. But since responsibility and blame are generally assessed at the personal level–such as whether the agent engaged in the questionable conduct intentionally or because an intervening external force caused him to do it–even if we assume that whenever there is action, there is volition, that alone would not show that volitions have any bearing on the theory of responsibility. A person who is exempted on grounds of not having acted is not exempted because of a lack of volition but because his conduct was not suitably attributable to him as an agent. It is more plausible that the spasms, reflexes, and the like that exempt an agent from liability do so because they show the conduct in question to be radically unhinged from the agent's deliberative attitudes and capacities, not because they did not volitionally cause the bodily movements, even if the latter is also true.

In short, to the degree that Moore's functionalism rescues him from Hart's objection, it renders the theory inert as far as normative theory–an enterprise concerned with explaining and justifying holding people responsible for various things under various circumstances–is concerned. We may still insist on the moral salience of a person acting in certain ways, but the willed-bodily-movement theory does not capture what we want when we think that acts are more deserving of punishment than nonacts, since our resentment in the case of the agent is directed toward his identification or involvement with his bodily movement and not toward the existence of subpersonal mental states functionally specified as occupying a certain position in the causal chain preceding the movement.

I should be clear that I am not claiming that criminal responsibility must always presuppose some conscious state of mind such as a plan or intention. If strict criminal liability or even criminal negligence is ever just, then that proposition must be false; for such agents need have nothing in mind when they do what they do–indeed, that is often precisely the problem with the negligent actor. Rather, my claim is that whatever plausibility there is to the view that “was X a voluntary act?” is a threshold question to any further inquiry into criminal guilt rests on the view that people are not responsible for bodily movements that are not voluntary actions; and the plausibility of this view is bolstered by the thought that non-agential bodily movements are bodily movements that are not connected in the appropriate manner to an individual's deliberative capacities.Footnote ¹⁷ But once we draw the distinction between action and nonaction in terms of functional rather than intentional volitions, that latter connection is severed.Footnote ¹⁸

The argument can be made in the form of a dilemma. If we construe volitions as necessarily characterizing an agent's conscious self-control (Austin describes volitions in terms of “my wish of the movement”), then there is some intuitive appeal to the thought that the epileptic and unconscious agents are excused because they lack volitional self-control in the relevant sense.Footnote ¹⁹ But the Austinian view faces Hart's original objection that it is not true that volitions necessarily characterize an agent's deliberative self-control. Alternately, we can sever the connection between volitions and conscious self-control, as Moore suggests, but this ends up making it quite mysterious why normative theory should care very much about volitions. Whatever interest volitions would have would be derived from their role as instrumental to A's beliefs, desires, plans, and so forth but not as objects of legal or moral concern in their own right.Footnote ²⁰ In either case, the result is that it is at best misleading to suggest that a putatively fundamental principle of criminal responsibility is reasonably founded on a volitional theory of action.

What I take to be the important positive claim to emerge from the discussion so far is this: while a conscious mental state such as a plan or an intention is not necessary for criminal responsibility, insofar as we insist on the primacy of the act/nonact distinction, this is because of an intuitive connection between acts and conduct appropriately attributable to the agent. But the nature of this connection has yet to be clarified.

II.

Suppose I am right that Moore's updated account of action as willed bodily movement is unhelpful for explaining Anglo-American criminal law's supposed act requirement. A natural reaction would be to cast about for an alternate theory of action to fit the bill. However, the conclusion I am after is more sweeping. What I am arguing is that there is no compelling reason to think our judgments of criminal responsibility are or should be held hostage to the theory of action in the way Moore suggests. (This is not to say that there might not be other ways in which the theory of mind or action might be relevant to criminal theory.) For “was X an action?” is simply not–as the act requirement suggests–a threshold question for “should we punish A for X?”

Two initial points must be emphasized. The first is that what I am challenging is the claim that the act requirement is a fundamental principle of criminal law; that is, as Antony Duff notes:

This claim implies that the act requirement cannot be a generalization from the sorts of crimes that Anglo-American law is, for various reasons, unwilling to accept. For in that case, it would be a mistake to think that X cannot be criminalized (or A punished for X) because X is not an action; action would be, at best, a rough-and-ready shorthand for representing those things that are felt to be outside the scope of criminalization, though the substantive reasons why they fall beyond the pale may not have anything to do with action per se.Footnote ²² That X is not an action must be an independent reason for not criminalizing or assigning punishment on its basis, not merely a stand-in for other (and therefore more fundamental) considerations; action should define not just the extension of the set of all crimes but their intension as well. If the act requirement is a structural feature of the notion of criminal responsibility, it should be understood as an instance of what John Rawls refers to as a “practice”- rather than a “summary”-based rule.Footnote ²³

Understood in this light, it is just as misguided to ask how many exceptions there are to the act requirement as it is to ask how many exceptions there are to the rule that solitaire is played by oneself. “Criminal responsibility requires an act” is, if valid, a constitutive feature of just punishment, not a statistical generalization. The claim that criminal responsibility “requires an act” should therefore be understood as at least partial a answer to theoretically ambitious question: What are the minimum necessary components of a morally acceptable concept of criminal responsibility?Footnote ²⁴

The second point to emphasize is that, as noted above, there is a distinction between the act requirement as a restriction on legislation–decisions about whether to criminalize certain classes of conduct–and the act requirement as a restriction on judicial decision-making–decisions about whether to impose punishment on a given individual for a given crime.Footnote ²⁵ While these aspects of the act requirement are easily conflated, it is crucial to keep them distinct. When it comes to describing patterns of legislation, both as a de facto and as a de jure matter, the act requirement is at best a wildly inaccurate generalization from the various rationales that have been given against extending punishment to omissions, thoughts, and statuses. It is only at the level of judicial decision-making that there is any indication that action is an independent explanatory variable. I agree there is a genuine and sound intuition that it is unjust to punish a person who was not in the appropriate way involved in the criminal result. But as I argue in the final section of the paper, that relation is best spelled out not in terms of action but in terms of practically rational agency.Footnote ²⁶

I start by noting that if there is a moral reason to impose criminal punishment only for a person's acts, then standard types of nonaction-based liability–notably, omissions, thoughts, and statuses–will be at most only very exceptionally permissible. The problem is that it is widely accepted that Anglo-American law does criminalize various sorts of nonactions, and properly so. For instance, one can be criminally responsible for failing to pay taxes, to stop at a red light, to care for one's minor child, or to take suitable care in one's risk-imposing activities, for conspiring to commit a crime, for overstaying a visa, and so on. Moreover, in some cases when actions are statutorily required, they are often at best only tangentially related to the object of punishment.Footnote ²⁷ And on one interpretation of the law of attempts, the object of punishment is the criminal intent, requiring the defendant to have taken steps towards its execution merely for evidentiary reasons.Footnote ²⁸ So a defender of the act requirement is now required either to explain away the apparent exceptions to the principle or to show why it is wrong to have these crimes. I cannot think of any strategy for doing this that is convincing.

First, one might try to finesse the question in this way: the act requirement is shorthand for the claim that we have reasons–moral ones, perhaps, but in any case specially weighty ones–for demanding that a person have acted before imposing punishment. But unlike requirements, reasons, while possibly weighty, are not generally peremptory and can in particular cases prove to be less weighty than countervailing reasons–of morality, social policy, political expediency, or whatever else. Thus exceptions to the act requirement are consistent with action as a rational restriction on criminal responsibility, for all the “act requirement” means is that there is a prima facie but defeasible case against criminalizing X when X is not an action (under some preferred theory of action). This move would allow a defender of the act requirement to contest the idea that the existence of exceptions shows that the act requirement is not a genuine general normative constraint.

This is perhaps a step in the right direction, but I do not think it saves the traditional view. For when we do impose punishment for omissions, say, we do not do so with a sense that we are sacrificing one value for another. But since it is not as if, in being outweighed by other considerations, a reason ceases to be a reason, we should expect that the absence of an action still provides some normative force even in cases where the act requirement is trumped by other considerations. Yet this does not appear to be the case. It is dubious that we generally feel that while it is morally objectionable to punish a tax evader (who committed no act), unfortunately that reason is outweighed by the even more objectionable alternative of allowing tax evaders to go unpunished, a result that leads us to punish them while regretting it all the while. Another case in which Anglo-American law imposes particular duties of care is that between parents and minor children such that a parent can be liable for failing to care for his child. There does not seem to be any appreciable sense in which the fact of the parent's inaction presents any exculpatory reason, even one that is ultimately outweighed. Indeed, we blame him precisely on account of his failure to act.

Contrast these sorts of cases with punishing an offender more severely than previous similarly situated offenders in order to “send a message” to potential future miscreants. Here we do sacrifice one principle–that of fairness, in the sense of treating like cases alike–to another, the promotion of socially acceptable behavior. Here, there is a legitimate sense that even if the circumstances justify us in doing so, we do not act fully justly toward the person whom we use to send our penal message. There is a sense of moral residue in this case that is not present in the case of the criminally negligent parent. The lack of moral residue in cases where we impose punishment for omissions suggests that these are not cases in which one moral principle is trumped by another.

Another line of attack is to try to recharacterize the problematic cases of what appear to be nonaction-based criminal liability as actually action-based after all. One way of doing this, at least for omissions, is to insist that some omissions are in fact disguised actions. For, it might be claimed, there is a difference between a person who “omits” to pay her taxes because of her principled political views and the person who is simply forgetful. The former deliberately omits while the latter merely fails to act. Perhaps, if we agree that bodily movement is not, as such, crucial in defining action, then willed omissions can also be classified as a kind of action, and therefore punishment on that basis is consistent with the act requirement.

The problem with this proposal is that not every criminal omission is a deliberate omission.Footnote ²⁹ Mere failure to act is sufficient, as, for instance, if a worker forgets to secure heavy machinery which then falls upon and kills passerby, or if a parent forgets his child in the backseat of his car on a hot summer day while he spends the day running errands. (“I forgot” is not generally a legally recognized excuse.) As these examples suggest, if mere failure to act were not properly subject to criminal penalty, then negligence, often premised on an agent's failure to do X as an ordinary standard of care requires, would be seriously curtailed as a basis for criminal liability insofar as it often rests on a nondeliberate failure to take certain required steps. Moreover, this proposal would effectively define an actus reus in mens rea terms because it would distinguish omissions from failures to act on the basis of criminal purposes, that is, whether the agent failed to do X purposefully or knowingly. This strategy makes a hash of what I describe above as the systematic ambitions for separating out actus reus as an independent principle of criminal law, one that can exculpate even if every mens rea requirement is satisfied (or if there is none, as with strict liability).Footnote ³⁰

A third and initially more promising way of trying to rescue the act requirement is by insisting that even in cases where it appears a person is being punished for a failure to act, his punishment is predicated on previous acts that create an obligation that he later fails to meet through his omission/failure to act.Footnote ³¹ For instance, a person who is charged with running a red light has a positive duty–given the undisputed act of getting in the car and driving–to obey traffic signals, or a person who has earned income in a jurisdiction has a duty to pay taxes on that income, and so forth. The object of responsibility is the complex of the prior act plus the later omission/failure to act.Footnote ³²

Such a construction raises difficult–though perhaps not impossible–issues as regards the limits on the stretch of conduct in which the act must take place and, crucially, how the background act must be “related” to the triggering nonact (i.e., what it means for an act to be “included” in the conduct for which the person is being punished). One problem in this vein is that even if the general strategy is sound, it is at best quite incomplete. For the strategy of relating criminally culpable omissions back to prior acts does not even work for every kind of omission. Some jurisdictions, for instance, attach criminal penalties to failure to vote in an election.Footnote ³³ In the United States, failure of male citizens between eighteen and twenty-five years of age to register with the Selective Service can (in principle) result in both serious fines and imprisonment.Footnote ³⁴ Such statutes present legal duties to act that are not conditioned on any prior act, even that of “entering” the jurisdiction. (One can be born into it and simply fail to leave). While the wisdom of such statutes is not uncontested, the statutes are not contested on grounds that the sanctions they allow for are not conditioned on prior acts. So even if we allow that some instances of punishment for nonacts presuppose prior acts, it is simply not the case that every instance does so.

The important question, however, is not whether a technical means can be found to relate every instance of culpable conduct to some prior act, but whether doing so is illuminating. It is here, I think, that the more critical problems for prior-acts analysis emerge. While it is true that one cannot be liable for failing to stop at a red light unless one has gotten into the car to drive in the first place, nor be liable for paying income taxes unless one has earned an income, it is misleading to suggest that what one is being punished for in such cases in some substantive sense “includes” these previous acts. For it is not as if driving a vehicle or earning an income itself contributes to the blame attributed to the criminal agent. It is true that the agent would not have the particular duty in question were it not for some previous voluntary act on his part, but the offense and the blame and penalty that go along with it consist in violation of the duty itself, regardless of what explains why he has the duty.Footnote ³⁵ Indeed, even by her own lights, a defender of the act requirement must insist on this point. For consider the alternative: if all the legal conditions on the acquisition of a legal duty are rolled into the offense of violating that duty, then, since presence in a jurisdiction when the crime was committed is typically a precondition for acquiring a duty to comply with the criminal law of that jurisdiction, it follows that a person who commits any crime will be punished in part for his mere presence in that jurisdiction. But mere presence in a jurisdiction is not an act at all.

Those who insist that just punishment can only be punishment for acts thus have no option but to deny that presence in a jurisdiction comprises part of the object of punishment. But if so, then what distinguishes the fact of having gotten into the driver's seat of the car or having earned an income as a prerequisite to legal duties from being within the statute's jurisdictional purview? The “agent A would not have duty D unless precondition P obtains” test does not on its face distinguish those preconditions that are acts from those that are not, and it is not clear that there are any principled and noncircular grounds for doing so. Of course, one could insist that those legal preconditions that are voluntary acts form part of the object of punishment only when accompanied by a later culpable omission, but this would be ad hoc. Therefore, since both act and nonact preconditions exist, without an explanation of why one is salient and the other not, this test is useless as a defense of the act requirement.Footnote ³⁶

Our attempts to defend at the legislative level the proposition that criminal liability requires an act have not been fruitful. Since we have been stymied in our attempts to make good on the notion that, from the legislative point of view, just punishment is always for an act, I tentatively conclude that that the principle should be abandoned. Acts are indeed the normal objects of criminalization, but the content of the assertion that criminal responsibility requires an act is exhausted by the familiar and heterogeneous reasons against criminalizing omissions, thoughts, and statuses–reasons that have everything to do with autonomy, privacy, and freedom from discriminatory abuse and nothing in particular to do with action as such.Footnote ³⁷ Thus understood, from the legislative point of view, the act requirement is wholly derivative. It is therefore not a fundamental principle of Anglo-American criminal law, for it is not a self-standing principle at all.

III.

I admit that it would be more than a little surprising if the consistent and long-standing consensus concerning the act requirement rested on a persistent illusion. In this section, I argue that there is a sound intuition underlying the act requirement, but that this intuition has traction only in its judicial rather than legislative role; and that it is better understood in terms of practically rational agency rather than in terms of acts.

The intuition is brought to mind by paradigm cases such as Martin v. State, in which the police carried the intoxicated defendant out of his home and onto the street, at which point he was arrested for public drunkenness.Footnote ³⁸ The injustice of a subsequent conviction rests, I take it, on the thought that “he didn't do anything!” It is irrelevant whether public drunkenness can be perfected with no showing of mens rea, for the issue is more fundamental than that. So it does not matter to this thought whether liability for public drunkenness is “strict.” On the other hand, it is important that Martin was physically carried onto the street rather than, say, being commanded to walk to the street. For in the latter instance, though punishment would still be unjust, it would be unjust because Martin was improperly coerced into going into a public space by the police, not because he did not “do” anything. In both cases, the conduct with which Martin is charged could be described as “involuntary,” but for quite different reasons. Anglo-American criminal doctrine respects this difference by distinguishing the “act requirement,” an element of the crime, from excuses (such as duress), which are affirmative defenses.Footnote ³⁹ My argument so far is that the development of this line of thought exclusively in terms of positive action is superficially plausible but ultimately cannot account for the great diversity in kinds and instances of justified criminal punishment. Keeping in view the diverse bases for just punishment as well as the act requirement's doctrinal role as distinct from both the mens rea and excuse doctrines, how, then, best to unpack the basis for our judgment in cases like Martin or Larsonneur?

The central concept is evidently something akin to an agent's ability to comply with the law he is charged with violating; that is, whether he was in a position to do (or refrain from doing) what the law requires. The paradigmatic cases of actus reus defenses–the person who is physically carried from one location to another, who has a seizure, sleepwalks, and so forth–are cases in which the person was in no position to comply with the law's demands. Conversely, the problematic cases for the act-based theorist–the citizen who refuses to pay his taxes, who forgets to register for the draft, who fails to take reasonable precautions in his dangerous activities, and the attempter who has decided to carry out some crime–are united in the fact that in each case the agent was in a position to comply with the law's demands but did not do so.Footnote ⁴⁰

We must, however, take some care in how we understand the notion of compliance in this context. For there are problems lurking in the wings: Control over what? What degree of control, and how far should we individualize our assessments of control? In what does control or voluntariness consist–what (if anything) unifies the various legally recognized forms of involuntariness? Does the agent who fails to carry out an affirmative duty “choose” to break the law? (When?) Why should it matter that the agent “could not have done otherwise” if she did what she wanted and intended to do all along? The remainder of this paper is devoted to sketching how to preserve what is sound in the so-called act requirement while suggesting answers to at least some of these questions.

Specifically, I think we can go a good distance in unpacking the underlying intuition if we ask: Was X (the criminal state of affairs the defendant is charged with) caused or constituted by the agent's conduct qua practically rational agent? Or, in assertoric form: Punishment is unjust unless the crime charged was caused or constituted by the agent's conduct qua practically rational agent. I believe that the practical-agency condition, as I call it, captures what cases like Martin and Larsonneur have in common with cases involving epileptic seizure, automatic reflex, and the like. The practical-agency condition isolates the fundamental sort of defect involved in such cases from those more properly treated under excuse doctrines, and manages to do so without suggesting, implausibly, that positive acts are the only proper basis of liability–all while retaining the traditional independence (such as it is) of actus reus from mens rea doctrine.

The central question for the practical-agency condition is to explain what notion of practical agency is being appealed to. As a first gloss, what I suggest is the commonsensical interpretation that a practically rational agent is an agent who is in a position to guide his conduct through the normal use of his deliberative and executive capacities–that is, those capacities for forming and acting on intentions.Footnote ⁴¹ An understanding of practical agency in these relatively vague terms is enough, I think, to show why the agent who is under extreme emotional stress (the standard example is walking in on a philandering lover), as with the agent who is threatened with harm unless he does X, cannot make use of the “act requirement” to exempt himself from liability, whereas the epileptic or unconscious agent (who may “do” exactly the same things) can. For the provoked agent is still an agent in the sense of forming and acting on intentions; his problem is rather that his judgment concerning what he ought to do in his situation is impaired. However, that defect, serious as it is, does not implicate his ability to form and act on intentions just as such.

Invocation of the practical-agency condition would require A to show not just that her grasp of which reasons to act on was impaired, but that her ability to grasp and act on reasons at all was destroyed; and that is both significantly rarer and harder to show.Footnote ⁴² Similarly, we should not deny that the individual who acts in such a way as to cause harm is nevertheless an agent, even if she acts only because she is threatened by another; and not should we deny that she is the cause of the harm in question. Given sufficient coercion, she should be exculpated under the heading of duress, but not exculpated because she is no longer a practically rational agent. (There would, after all, be little point in trying to coerce such an individual.) In contrast, if we suppose that an epileptic seizure removes any ability to control one's motions, then nothing requiring such a motion would be available to one in the grip of a seizure; one would, in that respect, no longer be a practically rational agent. The harm such an agent unequivocally causes is therefore not caused or constituted by her conduct as a practically rational agent.

As an aside, the focus on practically rational agency rather than control has an advantage here, in that the ability to control one's conduct is not evenly distributed in the population, and it is a fine question as to how much control (however we measure it) is enough in a given situation to allow a prosecution to proceed. Are extreme neurotics, the habitually impulsive, and the congenitally weak-willed–all of whom are plausibly less able than the average person to control themselves in various situations–therefore less prosecutable than the average person? (For example, the neurotic who cannot make up his mind to register for the draft, the impulsive grab for easy money, the follower who gets talked into a criminal plot, etc.). Now to be sure, the ability to form and act on intentions is also a variable trait, but the baseline below which a person has to fall to fail this condition is both lower and more discernible than what is suggested by the control criterion.

A somewhat more difficult case for the practical-agency condition is Anglo-American criminal law's exemption of crimes committed while sleepwalking, while under hypnotic control, and the like.Footnote ⁴³ The difficulty here arises when we note (as Bernard Williams and others point out) that there is an “intentional contour” to the actions of the sleepwalking or hypnotized agent, and it is accordingly difficult to assimilate such individuals to the epileptic or the comatose or the physically constrained as “nonagents.”Footnote ⁴⁴ Yet, unlike the enraged lover, we want to exempt individuals for crimes committed while sleepwalking and so on, and exempt them precisely because their will is in some sense not implicated in the crimes they commit. Like the epileptic, we imagine such agents to be in an important way dissociated from the movements of their bodies. But like the enraged lover, their conduct is clearly intentional.

Tentatively, I suggest that while such agents do act intentionally, they nonetheless are excused because they fail to be practically rational agents insofar as they do not have, one might say, executive self-control. In other words, we conceive of such agents as being in some way commandeered, as having their agendas set by something or someone else.Footnote ⁴⁵ So while the steps they take in carrying out those agendas are surely not unintentional, they are nonetheless not intended by the agents as part of some plan for which they are responsible. The radical dissociation between the agent's sense of self and the relevant conduct calls into question this agent's ability to set his own ends, as opposed to merely finding efficient means of carrying out extrinsically given ones; and this may call into question whether such individuals can be seen as practically rational agents such that prosecuting them for harms committed in that state is justified. Of course, there are likely to be very close cases and a great likelihood of self-serving dissimulation. And just as obviously, the notion of executive control on which I am relying, as well as the degree to which it can be distinguished from the means-end intentionality which characterizes the commandeered agents, requires much further elaboration.Footnote ⁴⁶ Fortunately, while the conceptual issues are murky and deep, such cases arise only infrequently in actual criminal contexts.

Before turning to address objections, I want to point out two further comparative advantages of the practical-agency condition over alternate accounts in the literature. First, the practical-agency condition has particular appeal when dealing with the “culpability-in-causing” cases, in which a person's involuntary conduct causes harm, but his involuntariness is a result of his own doing. The paradigm case is People v. Decina, in which Decina knew he was prone to epileptic seizures yet failed to take his medication before driving and, as a result, had a seizure behind the wheel and thereby killed several children.Footnote ⁴⁷ The challenge these cases pose is that while Decina was clearly culpable, not every epileptic driver is–for instance, a person who has no reason to suspect she is epileptic when she has a seizure behind the wheel. I agree with the received view that what is important is that Decina knew of his susceptibility to seizure and knew that he ought to take his antiseizure medication prior to driving. Where I disagree is that I do not view these facts as important because they establish some kind of nexus between Decina's causing of the deaths and a prior culpable act, a nexus that is necessary for punishment to be permissible, for I do not view it as important that there be a prior act at all.Footnote ⁴⁸

Nor does the practical-agency condition need to be given an ad hoc qualification to address these cases (“X is required for criminal responsibility, except when . . .”)–a result that would seemingly be at odds with the supposedly fundamental character of the condition. Rather, the practical-agency condition directly allows prosecution in cases like Decina because, given the defendant's knowledge of his condition, his deciding to drive without the antiseizure medication itself amounts to conduct as a practically rational agent that causes the harm with which he is charged. Thus, on the practical-agency condition, we can easily accommodate culpability-in-causing cases without modifying the supposedly general norms of criminal responsibility through ad hoc qualifications or tortured and underinclusive prior-acts analysis.Footnote ⁴⁹

Second, the practical-agency condition has the advantage that, compared to the control and voluntariness criteria, it is philosophically less contentious. Those criteria rest crucially on the premise that a person is not responsible if she could not have done otherwise; or, in Douglas Husak's case, if he could not have prevented the state of affairs from taking place or obtaining.Footnote ⁵⁰ But this premise (in either version) is philosophically vulnerable and, indeed, is the subject of quite a bit of ongoing controversy.Footnote ⁵¹ The practical-agency condition is not so committed and thus raises fewer philosophical difficulties. For consider: interpreted on its face, Husak's control requirement would exonerate any agent when the harm he causes is overdetermined, that is, if there exists a causal force independent of the agent sufficient to bring about the state of affairs in question. Thus, if A has decided to murder B by putting poison in his glass, then normally A would be liable given his control over whether B is poisoned. But now suppose that C, acting independently of A, has also decided to murder B by putting poison in his glass; indeed, suppose A and C put equal dosages of the same poison into B's glass, each of which is sufficient to kill B. Strictly speaking, A has no control over whether B dies or even over whether B is murdered, since no matter what A does, C will cause B to be poisoned. Moreover, since C can raise the same defense as A, the control requirement thus interpreted would entail that neither is responsible for B's murder, clearly an absurd result.

Husak can retrench and claim that the control requirement is a matter of not just what is controlled (states of affairs) but also how it is controlled (with or without the agent's participation). Retrenching along these lines brings Husak's views closer to that of P.F. Fitzgerald and A.P. Simester: that what is required for (legal) responsibility is the ability to have done otherwise. (Following standard terminology, we can call this the principle of alternate possibilities.) But the problem of overdetermination is not escaped quite so easily. For the agent's participation in the result can also be overdetermined, as Harry Frankfurt's famous examples are meant to indicate. Briefly, suppose A wants to poison B and is considering whether he shall. Now suppose that an evil demon, mad brain scientist, or rare neurological disorder lurks, ready to flip a switch and cause A to decide to poison B should A indicate (perhaps through certain neurological signs) that she is about to decide to do otherwise. But A decides to poison B and proceeds to do so of her own accord. Given the circumstances, A has no control over either the resulting state of affairs or her participation in it. But since the reason A poisons B is not because she is forced to do so by the evil demon or whatever else may lurk, but rather because of her own decision, it seems gratuitous to exculpate A on grounds of a lack of control. Thus, if the example is sound, then the claim that responsibility requires the possibility of doing otherwise is seriously jeopardized. Put in more general terms, A may be responsible for X if A's lack of control with respect to X does not contribute to explaining why A did X.Footnote ⁵²

Such a cursory sketch does not, of course, amount to an argument against the principle of alternate possibilities and so does not directly undermine Husak's control requirement. However, my claim is that if we retreat from control as ability to influence outcomes to control as ability to have done otherwise, we are inescapably premising an account of when imposing criminal punishment is just on the resolution of ongoing and potentially intractable philosophical issues. Since the practical-agency condition does not presuppose the principle of alternate possibilities, it is not troubled by Frankfurt-type examples, and thus less hangs, I think, on the outcome of that debate for the practical-agency condition than for the control requirement. The background assumption to my argument here is the belief that it overburdens criminal theory to suppose that basic legal concepts such as actus reus cannot be adequately understood until the problems of perennial philosophy are ultimately resolved. If we can avoid having to say that, we should; and I think the practical-agency condition shows us how, in this one context, we can.Footnote ⁵³

It might be objected that since the principle of alternate possibilities says that such possibilities are necessary for moral responsibility, an account of (in this case, legal) responsibility that does not require such possibilities is not agnostic on the principle; it rejects it. But if that is so, then the same should apply, and the dialectical advantage of the practical-agency condition (beyond simply providing a broadly compatibilist alternative) over the control requirement evaporates. The response to this challenge is that the practical-agency condition is agnostic. After all, it can be given an interpretation that is consistent with the principle of alternate possibilities, for instance, by arguing that if A's decision to do X was causally determined such that it is not true that A could have formed and acted on an intention to do not-X instead of X at the relevant time, then A's conduct was not in any meaningful sense guided by his deliberation as to what to do, and so A fails to be a practically rational agent in the relevant instance.Footnote ⁵⁴

Of course, one could also insist that the principle of alternate possibilities is wrongheaded, in which case it would be no argument against the practical-agency condition that it does not yield a pattern of responsibility consistent with that principle. By leaving the notion of practical agency at a higher level of abstraction, the condition is not ipso facto committed to either a compatibilist or an incompatibilist reading but remains–or so I hope–sufficiently substantive to provide principled guidance in deciding cases.

I conclude by considering three objections. First, it might be objected that the practical-agency condition requires that criminal responsibility be premised on intentional–in the sense of deliberate–conduct. But this is misleading, for failure to act, including negligent failures, is properly subject to criminal sanction. In response, while I accept the major premise of the objection, I reject the conclusion, as I think failures to act indeed do fall within the scope of the practical-agency condition. After all, the harms complained of (failing to register for the draft, imposing risk by failing to secure heavy machinery) are caused by the individual's conduct during a legally relevant stretch of time. That is, it is sufficient to bring about these harms that at no point during the relevant stretch of time did the agent undertake to do what was legally required. The practical-agency condition is satisfied so long as, during that time, the individual was not causally prevented from taking stock of what he was failing to do and taking steps to bring himself into compliance (for instance, if the agent was in a persistent vegetative state during the entirety of the relevant time frame).

Second, one might reasonably ask: If X is a requirement on every case of justly imposed punishment, how could it not be an independent basis for restricting the definition of crimes? Surely, one might insist, there is a role for the practical-agency condition in restricting the definition of crimes, since it restricts in every instance how such laws are applied. I concede that the condition could in theory be used, for instance, to prohibit crimes with which it is impossible to comply–for example, if a legislature were, for some reason, to penalize being born on an odd-numbered day of the month or require a person to do something physically impossible. Such “offenses” would not be caused or constituted by the agent's conduct as a practically rational agent and so would be unjust as such. I would not, however, expect this to be a very significant role for the practical-agency condition in a reasonably well-ordered polity, especially in comparison to its role in restricting the application of facially just criminal laws.

Finally, one might challenge not the content of the practical-agency condition but its general applicability. It is, after all, certainly not obvious that the law must in every case treat its subjects as practically rational agents. That a person was not functioning as a rational agent at the time of the offense is not always a defense, for instance, in civil tort actions.Footnote ⁵⁵ So why condition criminal responsibility in general on it? I agree that nothing said here so far provides an answer to such a fundamental challenge. It is one thing to describe a possible moral constraint on criminal responsibility, and something else altogether to provide a justification for accepting that constraint in the first place. But answering such a sweeping challenge is not my task; rather, I am proceeding by attempting to reconstruct rationally the intuition I take to underlie Anglo-American criminal law's “act requirement,” and to do so in a way that preserves some role for it as a fundamental and independent limitation on when punishment is imposed. I think such a role is sensible and makes moral sense, but I do not defend that claim here. Perhaps all that can be said here is that so long as we assume some distinctive role for the criminal law in expressing norms of moral condemnation and not simply–that is, not only–in the regulation of destructive behavior, then there is plausibility to the view that the criminal law should be concerned with individuals not just as brutely subject to the law's demands but as rationally responsive to them as well. Obviously much more could be said on this point. My aim is limited to the more modest task of rational reconstruction in criminal theory, not the laying of normative foundations.

The view I have sketched marks, in one sense, a radical departure from the orthodox view that criminal responsibility requires an act. For I argue that, contrary to the usual understanding of the act requirement as an independent and fundamental principle of Anglo-American criminal law, whether X is an act is not a threshold question for whether it is permissible to criminalize X, regardless of the content assigned to X. It may well be a bad idea to make a crime out of, say, deliberately omitting to register for the draft, forgetfully failing to feed one's child, being drunk in public, being in a country illegally, or failing to exercise due care in undertaking hazardous activities. But if such crimes are unjust or otherwise a bad idea, the explanation will not be: because these are not genuine acts. However, in another and perhaps deeper sense, the view I have sketched is not a departure at all. For if I am right, the practical-agency condition is a plausible interpretation of the intuition that an agent must be an author of that for which he is being held criminally responsible. And it is this intuition, I suggest, that has actually, if obscurely, motivated Anglo-American criminal theory's supposed “act requirement” all along.