Printer Friendly

Causing, aiding, and the superfluity of accomplice liability.

 I.  THE PUZZLE OF ACCOMPLICE LIABILITY
II.  SOME BAD REASONS FOR THINKING ACCOMPLICE LIABILITY TO
     BE NONCAUSAL OR OTHERWISE DISTINCT IN ITS STRUCTURE
       A. Liability as an Accomplice Does Not Depend on One Being a
          Necessary Condition of the Harm, but Causation Does
       B. Accomplices Cannot Cause the Results Brought About by
          the Actions of Their Principals Because Those Latter
          Actions Are Intervening Causes
       C. Accomplices Cause the Harms They Aid but Only in a
          Distinct, Secondary, and Anemic Sense of "Cause".
       D. The Supposed Inapplicability of the Causative Verbs of
          English to the Causings of Accomplices
       E. Nonproxyable Crimes and the Need for Accomplice Liability
III. BEGINNING AGAIN: THE BASES OF ACCOMPLICE LIABILITY
       A. Truly Causal Accomplices
       B. Necessary Accomplices
       C. Chance-Raising Accomplices
       D. Necessary to Chance Accomplices?
       E. Subjectively Culpable Accomplices
       F. Vicarious Accomplices
IV.  SOME CONCLUSIONS ABOUT THE FOUR
     KINDS OF ACCOMPLICES


I. THE PUZZLE OF ACCOMPLICE LIABILITY

There are two doctrinal puzzles about accomplice liability in Anglo-American criminal law. One is the puzzle about the mental states required for conviction as an accomplice. Hornbook law has it that accomplice liability is a "specific intent" offense, (1) a requirement that the accomplice have "purpose" and not merely "knowledge," as those terms are used in the Model Penal Code. (2) Yet, are there two mens rea requirements here, a "primary" mens rea having as its object the aiding of the conduct of another person, and a second requirement having as its object the elements of the underlying crime aided? If so, does the secondary requirement expand or limit the liability otherwise permitted by the primary requirement? What is the relationship between the mens rea required for conviction of guilty principals and the secondary mens rea required for conviction as an accomplice? Does this vary depending on the kind of element (circumstance or result) of the underlying offense involved?

Interesting and important as these mens rea questions are, my focus here (3) is on a second puzzle about accomplice liability, having to do with the actus reus of being an accomplice, not its mens rea. Put generally, what does one have to do in order to be guilty as an accomplice to someone else's crime? Again, the hornbook law answer (in the ancient language of the common law) is that one must "aid and abet" another's commission of a crime in order to be guilty as an accomplice to that crime. (4) So one can phrase my inquiry as a question about what it is to aid or abet another to do something.

What is puzzling here can be got at more precisely by stepping away from accomplice liability temporarily to reflect on three other bases of criminal liability. One is the liability as a principal for some completed crime. If I shoot and kill you, I have committed the actus reus of homicide; if I drive away in your car without your permission, I have committed the actus reus of theft. In such cases, it is my agency that is involved, not someone else's; in addition, my agency has "completed" the crime in the sense that it has produced each of the elements required for conviction of the offense.

The structure of nonomissive, principal liability for completed offenses is essentially causal. (5) Sometimes this is obvious, as when a statute prohibits one from "causing the death of a person" or "causing the disfigurement of a person." Yet this causal structure is only slightly less transparent when criminal statutes use causally complex verbs of action such as "killing" or "hitting" a person or "abusing" a child. Such verbs transparently require that one cause death, contact, or abuse, respectively. This causal structure only becomes less obvious for what many criminal law theorists call "conduct crimes." It is commonly said that there are no "result elements" (i.e., no causal requirements) for crimes such as burglary, rape, theft, kidnapping, defacing public property, or drunk driving. (6) Yet there plainly are causal requirements for such offenses. Rape is done only when the perpetrator, by his bodily movements, causes penetration; burglary, when she causes a breaking and entering; theft, when he causes movement--"asportation"--of the thing stolen; kidnapping, when she causes confinement and movement of the person kidnapped; defacing, when he causes marks constituting defacement to appear on public property; drunk driving, when she causes a car to be in motion while she is drunk; etc.

There are two reasons why theorists have been misled about the universal causal structure of nonomissive, principal liability for completed offenses. One is the immediacy of causal chains connecting acts of bodily movement and prohibited states of affairs in the kinds of crimes considered here--so-called "conduct crimes." Because of the structure of such chains--between movements of the relevant sort and penetration in rape, for example--there is little room for complex causal questions to arise. By contrast, causing death can be done in quite complicated ways. The upshot is that it is easy to miss the causal requirements in the former type of cases; being unproblematically simple, these are too obvious to notice.

The second reason for not seeing the obvious here lies in the seeming triviality of saying things like, "to move some object is to cause the movement of that object." That smells suspiciously like a famous definition of a rose. Yet the language misleads here. To use "move" as a transitive verb, as in "I moved the table," is different than using "move" as an intransitive verb, as in "the table moved." (7) To move a table does require that it move, but that does not prevent the bodily movements (that constitute my act of moving it) from causing the separate event that is the table's movement.

Obvious or not, nonomissive, principal liability for completed offenses is causal in structure. To be guilty as such, a principal is to act in such a way as to cause some legally prohibited state of affairs. Contrast this form of liability with a second form, that of inchoate liability. For inchoate liability, while the principal's act may have to cause something, it need not cause the state of affairs the law ultimately cares to prevent. Attempt crimes, crimes of reckless endangerment (when the danger is not known to the victim), solicitation, and conspiracy are the usual examples of inchoate liability. So are many specific intent crimes, such as traveling across state lines with the intent to bribe a state official.

Inchoate crimes lack the causal structure of completed crimes in that the perpetrator need not cause the state of affairs the law seeks to prevent. Indeed, typically, that undesired state of affairs has not occurred so one couldn't have caused it. But one does need to have that state of affairs in mind as one acts in order to be guilty of an inchoate crime. For example, one needn't cause death to be guilty of attempted murder, but one needs to have such a death as the object of one's purpose (or in some cases, belief) for such liability.

The third major form of criminal liability is vicarious liability for the acts of others. In vicarious liability, one needn't cause, nor even try to cause or risk causing, a legally prohibited state of affairs. If someone else has caused one of a number of legal wrongs, and if one stands in a certain relation to that perpetrator, then one is liable oneself for the offense committed. Vicarious liability is a form of agency well known to the civil law, in doctrines such as that of respondent superior in torts. Criminal law is much more circumspect in imposing liability vicariously. Still, it is not unknown, the most notable example being liability of conspirators for the crimes of their co-conspirators even though there is no more than a general agreement between them. (8)

Accomplice liability is usually presented as a fourth and distinct basis of criminal liability, in addition to principal liability for completed offenses, inchoate liability for incomplete offenses, and vicarious liability for members of criminal combinations or groups. The puzzle is this: what relationship must exist between the act of the accomplice and the state of affairs the law seeks to prevent? We know, for example, that to be guilty of the completed offense of murder as a principal the accused must cause death. We also know that to be guilty of the inchoate crime of attempted murder the act of the accused must be in execution of an intention that has as its object such a death, and the act must go some distance toward causing such death even though no death is caused by such an act. And we know that to be guilty for the crimes of a co-conspirator, a conspirator must have established a relationship amounting to a conspiracy with that coconspirator, who herself causes the legally prohibited state of affairs. What are we to say is the analogous relationship between the act of an accomplice and a result such as death that someone else has caused, if the accomplice is to be guilty of aiding and abetting murder?

A natural temptation is to assimilate accomplice liability, either to the vicarious liability of co-conspirators, (9) to inchoate liability, (10) or to principal liability for completed crimes. (11) Consider first the assimilation to vicarious liability. The often-expressed idea is that complicity, like conspiracy, is an agency form of vicarious liability. For this form of liability, one needn't have contributed in any way toward the occurrence of some legally prohibited result; one only need be a member of some group or combination, the other member(s) of which does cause some legally prohibited result to occur. On this view, an accomplice stands as the principal liable for the action of his "agents" (i.e., the perpetrator).

There is some truth to this basis for complicity, at least as complicity is reflected in certain doctrines. Specifically, in some states merely agreeing with another that she will do some crime makes one liable as an accomplice to the crime that the other commits. (12) Secondly, a number of states make one liable as an accomplice not only for crimes that one agrees should be committed but for all crimes committed by any member of the group, so long as the commission of such crimes was foreseeable to the accomplice. (13) If one puts these doctrines together, a form of vicarious liability results: merely by joining a group, but doing nothing else oneself, one is liable as an accomplice for all the crimes of that group's members that are foreseeably incident to those crimes that may have motivated the group's formation to begin with.

Yet this vicarious (or agency) interpretation of complicity does not begin to cover cases where accomplice liability has been imposed. There is no requirement that one be a conspirator to be an accomplice. If I aid you by finding a ladder, placing a gun where you can find it, and making sure the victim is where you can find him, I am liable as an accomplice for whatever crimes I am trying to promote with such aid, even if there is no prior agreement between us. Moreover, in most states, complicity requires more than mere agreement or group membership; one has to aid the commission of a crime to be an accomplice, and in such states the aiding required cannot be reduced to mere group membership or general agreement.

Now consider the second assimilation, that to inchoate liability. The reasoning here is this: to be guilty as an accomplice requires the same mens rea toward the evil the law seeks to prevent (e.g., death in homicide) as does inchoate liability--at least if one resolves in certain ways and not others the doctrinal ambiguities about the mens rea requirements for accomplice liability mentioned above. So all that is different in the case of accomplice liability is that the evil the law seeks to prevent has occurred, whereas for inchoate liability it typically has not. To be guilty as an accomplice to murder requires that a murder has taken place--i.e., that there be a death. By contrast, for liability for attempted murder, no death need have happened. So (on this view) accomplice liability is just inchoate liability in the special cases when the evil sought to be prevented by the law has occurred (even though the accomplice did not cause it to occur).

Such assimilation to inchoate liability plainly will not work. Suppose I shoot at someone, intending to kill him. I miss, but at just the moment I would have killed him if I had hit him, another bullet fired by someone else arrives and kills him in just the way my bullet would have killed him. I have tried to kill the victim, and the result (broadly speaking) that I was trying to bring about has occurred (viz., he was killed); yet it is plain that these two facts do not make me an accomplice to some form of homicide. I have only inchoate liability, not accomplice liability, showing that accomplice liability requires something more than attempt plus occurrence of the result attempted. To be an accomplice, my act must have something to do with why, how, or with what ease the legally prohibited result was brought about by someone else.

Seeing this difference between accomplice liability and inchoate liability (even where the result attempted or risked does occur) tempts a third assimilation: my act of shooting above must in some sense cause the victim to be killed by the second bullet for me to be an accomplice in his murder. (14) Accomplice liability is thus likened to principal liability in that both require that an accused's acts cause some legally prohibited state of affairs. The caveat for accomplice liability would be that the accomplice's way of causing death operates through the action of a third person, whereas principals may directly cause prohibited states of affairs like deaths.

By and large, this assimilation also will not work. Yet why this is so is a complicated business. In the first two sections of Part II, I will explore two commonly voiced reasons for thinking that accomplice liability must be noncausal in its structure, concluding that they are unsound. I will then explore reasons for thinking that causing by aiding another involves a different kind of causation than that needed for principal liability, concluding again that there is no reason to think this. I will close the Part by considering two commonly voiced reasons for thinking that even though accomplices cause death by aiding another to kill, they do not themselves kill, which is the prerequisite for liability for homicide as a principal. This line of thought too I by and large reject. In Part III I will begin again in the search for the bases for accomplice liability. The five bases I will explore involve the sometimes causal, sometimes noncausal relationship(s) that must exist between an accomplice's act and a legally prohibited state of affairs, for accomplice liability to attach. My conclusion is that aiding another to cause a harm is not a distinct basis for blame and punishment.

II. SOME BAD REASONS FOR THINKING ACCOMPLICE LIABILITY To BE NONCAUSAL OR OTHERWISE DISTINCT IN ITS STRUCTURE

Even paranoids have real enemies--or so the saying goes. True or partially true conclusions are indeed often supported by bad reasons. Such is the case here.

A. Liability as an Accomplice Does Not Depend on One Being a Necessary Condition of the Harm, but Causation Does

A very standard view of why accomplice liability is noncausal contains two premises. (15) It first asserts that the criminal law's causal requirements include a "necessary condition" (a "sine qua non," or counterfactual element): for an act A to cause a harm B, A must be necessary for B (i.e., if A had not happened, then B would not have occurred either). So if accomplice liability had a causal structure, then the act of an accomplice would have to be a necessary condition for the prohibited harm to occur. To be an accomplice to murder, for example, one would have to have done something necessary to the death occurring. The second premise is that there are many cases of accomplice liability in which the act of the accomplice is not necessary to the occurrence of the harm. Ergo, the conclusion: accomplice liability is noncausal in its structure.

On the second premise, consider the well known case of State ex rel. Martin v. Tally. (16) In Tally, the alleged accomplice, Judge Tally, learned of the plans of the four Skelton brothers to ride to the next town and shoot and kill one Ross (who had seduced the Skeltons' sister). Judge Tally discovered that a warning telegram was to be sent to Ross and ordered the telegraph operator not to deliver the warning telegram. Ross was not warned, and the Skeltons found and killed him.

On these facts, it is possible that Judge Tally's action was necessary for Ross's death. Perhaps with a warning Ross could have fled or defended himself successfully. But it is also possible that Tally's act was not necessary; whether warned or not, the Skeltons may well have caught up with Ross and killed him. (This may even be true if we focus in on the exact time, place, and manner of Ross's killing; even if the warning had been received, Ross might well have been killed exactly when, where, and how he was actually killed.)

The important point is that under the law of accomplice liability it doesn't matter which of these scenarios is true. As the Tally court accurately stated the law on accomplice liability, "The assistance given ... need not contribute to the criminal result in the sense that but for it the result would not have ensued." (17) If one assumes that the counterfactual ("but for") test must be satisfied for there to be causation in the criminal law, then one has the desired conclusion: "The upshot of these cases is that causal responsibility is not necessary to complicituous criminal liability." (18)

The problem with this argument lies in the first of its two premises. While theorists such as Kutz are not alone in thinking that the "general causal requirement[s]" of the criminal law are "expressed in counterfactuals," (19) this in fact is not so. Saying why this is not so is a complicated business, but fortunately I have gone through the complications elsewhere, (20) so only a summary need be given.

To begin with, it is very implausible to identify the causal relation as counterfactual dependence. When two names or descriptions that putatively are different nonetheless refer to one and the same thing, then what is true of that thing under one name or description must be true of it under any other name or description. Leibniz taught us as much about identity. (21) To use a time-honored example: if "the Evening Star" and "the Morning Star" both refer to one and the same thing, viz., the planet Venus, then if it is true that the Morning Star rises in the morning, so it must be true that the Evening Star does too, and so does Venus.

If causation is counterfactual dependence, then it can have no properties not possessed by counterfactual dependence, and vice-versa. Yet causation is a transitive relation. With qualifications not here important, (22) if C causes E, and E causes F, then C causes F. It is well known that counterfactual dependence is not transitive. (23) F may counterfactually depend on E, and E may counterfactually depend on C, and yet F may not counterfactually depend on C. Moreover, causation (again with qualifications not here relevant) is an asymmetric relation. If C causes E, E does not cause C. Counterfactual dependence is not necessarily asymmetric. E can counterfactually depend on C, even while C counterfactually depends on E. Finally, causation relates temporally ordered states of affairs. If C causes E, then E must not precede C--causation does not work backwards through time. Counterfactual dependence, however, knows no such limitation. E can counterfactually depend on C without regard to whether C precedes or succeeds (or is simultaneous with) E in time. (24)

So there is little plausibility in identifying causation as counterfactual dependence. The more plausible claim is that counterfactual dependence is a good test of causation--which it certainly would be if "C causes E" is true when but only when "E counterfactually depends on C" is true. The weaker, more plausible, claim here is of extensional equivalence between statements, not of identity between the things such statements are about.

Biconditionals such as "C causes E if and only if E counterfactually depends on C" are the conjunction of two conditionals: (1) that the counterfactual dependence of E on C is sufficient for C to cause E and (2) that the counterfactual dependence of E on C is necessary for C to cause E. Although the matter is hotly disputed in the philosophy of science, it is difficult to sustain either of these conditionals.

Take the sufficiency claim first. If there is causation whenever there is counterfactual dependence, then the following four statements would have to be true: (a) Absent events must be causes, for an effect E can counterfactually depend equally on the absence of a type of event occurring as it does on the presence of certain other events. This means my failure (while in Oslo, where I am writing this) to prevent George from shooting you (in New York) is as much the cause of your death as was the event of George shooting you. I don't know about you, but I take my cues on causation here from Julie Andrews in The Sound of Music: "Nothing comes from nothing; nothing ever could." (25) (b) Logical and mereological relations are causal relations, because the existence of bachelors counterfactually depends on the existence of unmarried male persons, and because the existence of a whole word like "Larry" counterfactually depends on the existence of part of the word such as "rr." (c) Epiphenomenal events (i.e., co-effects of a common cause) must be causes, because there are counterfactual dependencies between such pairs of events. So when I run with my dog this causes my feet to get tired; it also causes, somewhat later, my dog to get tired. It is plausible to think that if my feet were not tired, then my dog wouldn't be tired either because I wouldn't have gone running. So my feet getting tired caused my dog to get tired. (d) Later events must cause earlier events whenever those earlier events counterfactually depend on those late events. For example, if my dog weren't tired, then I wouldn't have run as far as I did (the only way he gets that tired is by my running a certain distance with him). Thus, his getting tired caused me to run as far as I did.

"Counterfactualists" about causation vigorously dispute each of these last contentions. (26) While to my mind their responses do not meet the challenges I raise, in this context this is a bit by-the-by, for it is the second conditional that is crucial to the argument about accomplice liability being noncausal. This is the claim that counterfactual dependence is necessary for causation. And even those who defend the sufficiency conditional above much more rarely defend the necessity conditional needed here.

The main stumbling block to the view that counterfactual dependence is necessary to causation is to be found in those cases where there is more than one set of conditions sufficient for the occurrence of some one state of affairs. I refer to cases of concurrent overdetermination, preemptive overdetermination, and asymmetrical-concurrent overdetermination. Here are three examples (respectively): two bullets simultaneously strike the victim, each sufficient to kill her instantly, and she dies; my bullet kills the victim but another just behind it was sufficient to kill the victim if she weren't already dead; my lesser wound joins a larger wound, sufficient by itself to kill the victim, who bleeds to death from both wounds. In each of these cases some factor other than the accused's act is sufficient for some harm; that means that the accused's act is not necessary.

Despite the enormous attention given to these cases in the philosophy of science these past three decades, (27) the straightforward conclusion much of that literature attempts to avoid is at the end of the day the correct one: a cause of a certain harm need not be a necessary condition of that harm's existence. Particularly instructive are the three generations of attempts to resist that conclusion by David Lewis, (28) who in the end does no better than the drafters of the Model Penal Code fifty years earlier. (29) Both ended up with the fine-grained individuation of effects solution to this problem which is no solution at all. This is the view that if you individuate some harm finely enough, by the exact time, place, and manner of its occurrence, then whatever caused that harm was necessary to it being just like that. This doesn't work at all for the concurrent overdetermination cases-a one-fire destruction of a building can be exactly like a two-fire destruction (when the fires have joined) save for the fact that the destruction was caused by one fire. And the fine-grained solution is unprincipled and ad hoc in the preemptive cause cases because one individuates to the point that one gets the right causal answer, arrived at by some noncounterfactual means.

The upshot is that some act may be causative of some harm even though it is not necessary to the occurrence of that harm. Legal theorists have simply been misled in their long-held view that the "scientific" or "factual" part of legal causation is counterfactual in nature. Now, of course, some legal theorists would claim a freedom for the law to define its notion of cause-in-fact as it pleases, to serve its own purposes. On this view, "cause-in-fact" in the criminal law need not correspond to the actual nature of causation, but need only serve the artificial purposes of the law. Yet that old view, bequeathed to us by the Legal Realists, (30) has no place here. Causation is one of those doctrines adopted by the criminal law because of its role in determining moral responsibility. If criminal liability is to track moral responsibility, the law cannot simply define items like causation as it pleases. It must discover, not stipulate, what causation is, so that the law can parcel out legal liability on the same grounds as it measures moral responsibility. The law thus has no Humpty Dumpty-like freedom to mean what it pleases by "cause-in-fact." If counterfactual dependence does not constitute or even measure causation in reality, then it does not do so in the criminal law either.

The case law sustains this rejection of stipulative definitions of causation, whether by the highly influential American Law Institute, or by anyone else. Courts do what the more perspicuous First and Second Restatement of Torts told them to do: when the but-for test leads to counterintuitive results in the concurrent, asymmetrical, and preemptive overdetermination cases, then they ignore the definition and go with some other, more intuitive notion of causation. (31)

B. Accomplices Cannot Cause the Results Brought About by the Actions of Their Principals Because Those Latter Actions Are Intervening Causes

The second major explanation for why accomplice liability is noncausal stems from certain views about proximate causation (as opposed to the views about cause-in-fact that motivated the explanation just examined). The pertinent views on proximate causation are those that go under the label "direct causation." The direct cause version of proximate causation goes like this. Most versions of proximate causation--such as those based on foreseeability, harm within the risk, space/time proximity versions--are scalar in the sense that they conceive of causation as gradually petering out. The direct cause version at least purports to be nonscalar; causal chains are dramatically cut off by "intervening" (also known as "superseding" or "extrinsic") causes. On the direct cause view, causal chains continue across considerable links but suddenly end whenever one of those special events constituting an intervening cause intervenes between the defendant's act and the harm.

The direct cause notion is thus given whatever content it has by the notion of an intervening cause. Generically, an event is an intervening cause if and only if (1) it is an event, not a state or an omission; (2) it occurs after the defendant's act but before the harm in question (i.e., it "intervenes"); (3) it is itself a cause of the harm; (4) it is causally independent of the defendant's act (i.e., it is not an effect of that act); and (5) it is either such an extraordinary natural event as to amount to a "coincidence," or it is the "free, informed, voluntary act" of some third-party human agent. (32)

It is this last bit that is of greatest relevance here. A human act is "free, informed, and voluntary" only when (a) it is the act of a responsible agent (meaning an adult, sane, rational, and autonomous person); (b) it is a chosen act in the sense that the bodily movements that constitute the act are willed by the agent (sleepwalking, reflex reactions, hypnotic behavior, etc., do not qualify); (c) it is not done in ignorance or under mistaken belief about the harm-producing aspects of the action; and (d) it is free in the sense that it is not coerced or otherwise compelled by external human threats or natural necessity, or by internal, ego-alien emotion. (33)

Examples sometimes convey information better than generalities, so suppose I know that you are an extremely jealous, gullible, and volatile person. I get you to assault your wife by (a) getting you so drunk that you are barely rational; (b) hiding your insulin while you are locked in a room with her so that you strike her when you go into a hypoglycemic episode; (c) deceiving you about her presence behind a punching bag you are about to use; or (d) threatening you with the death of your children unless you strike her. In none of these cases is your hitting her an intervening cause. In all of them, therefore, I am "guilty as a principal" because I have proximately caused her to be struck.

Contrast these with cases where I (a) tell you where she is, knowing that you will hit her because you have just learned of her affair with another; (b) tell you about the affair, with the same expectation; (c) suggest to you that no red-blooded man should put up with such cheating behavior without some violence on his part; or (d) offer you money to go beat her up. In all of these latter cases, your hitting of your wife will constitute an intervening cause, breaking any causal chain that might otherwise exist between my acts and her injury, and thus precluding any criminal liability on my part as a principal.

Enter accomplice liability. As Sanford Kadish has shown in great and persuasive detail, the doctrines of accomplice liability fit like soft clay around the shape of intervening causation. (34) Wherever an intervening agent is himself guilty of a crime and is an intervening cause vis-a-vis some earlier act by another, that other is eligible for accomplice liability. However, if the intervening agent is not an intervening cause, then the earlier actor is eligible for liability as a principal. The picture is reminiscent of the hit tune by the former punk rock band, Blondie: we're going to get you, "one way or another." (35) Intervening cause doctrines will determine which way it will be.

Notice that on the Kadish view accomplice liability is always and necessarily noncausal. Acts of accomplices cannot be causal because of the intervening causation of their principals. And if the acts of their principals are not intervening causes, then the would-be "accomplices" cannot be accomplices--because they will be principals (i.e., causers) themselves.

This is a quite accurate description of Anglo-American legal doctrine on accomplice liability. Yet making normative sense of this doctrine is much more challenging. Just because it is elegantly systematic does not mean it makes normative sense. It might mean instead the whole edifice falls to the ground once any of it is removed.

There are two routes one might take in seeking to make normative sense of the intervening cause/accomplice liability doctrines. One route is to seek to ground the legal doctrines in the facts about causation. Because the dominant policy of the criminal law is to punish only those who are morally responsible and only in proportion to the degree of their responsibility, and because the degree of moral responsibility is measured in part by the degree of causal responsibility, (36) this first route can succeed in its normative task if it succeeds in grounding the legal doctrine in causal fact. That requires that the distinctions drawn by the legal doctrines of intervening causation turn out to be true, causal distinctions.

The second route to making normative sense of these legal doctrines is quite different. This route admits that intervening actors do not break causal chains in fact. In fact, accomplices cause harms through the acts of their principals. But there are nonetheless policy reasons justifying the law in treating intervening actors as breaking causal chains even when they do not do so in fact. This is the avowedly fictional approach that I eschewed in the context of but-for causation above, but that has greater prima facie plausibility for intervening causation. (This is because all proximate cause doctrines have something of this fictional, constructed cast to them, in contrast to the prima facie literalist pretensions of cause-in-fact doctrines such as the "but-for" test.)

Turning to the first route, literal chain-breakers are hard to find in science. "Shit happens," as a popular 1990s bumper sticker had it. But surely that does not mean that such bad things are inexplicable because uncaused. The notion of prima causa in theology gives us a picture of what such chain-breaking, uncaused causes would have to look like. Theologians give that power to god; libertarian metaphysics gives it to each free human choice. We are all "little gods," causing things while free of being caused to do so by anything else. Such libertarian metaphysics shows us what we need here if we are to make sense of the legal doctrines of intervening causation in causal terms, (37) as Kadish recognizes. (38)

Yet to my mind this is like saying we need theism to be true for intervening causes to exist in fact. Indeed, is not such libertarianism a kind of theism, a pantheism? The libertarian is committed to there being lots of gods, namely, each person. These gods are like the ghost Casper in that they can cause things to move whilst themselves being immune to causal influences. This strikingly unscientific posture is a desperate and panicky metaphysics that should be put aside as not serious. Kadish himself does not subscribe to this metaphysics. (He doesn't unsubscribe either--he simply doesn't tell us what he thinks to be true.) Rather, he explains why the law of intervening causation/accomplice liability has the shape that it does in terms of common belief: those who produced this legal doctrine believed in libertarian metaphysics. (39)

This bit of sociological speculation by Kadish may well be true. Yet can we turn this plausible explanation for why we have such doctrines into a plausible justification of why we should keep them? Suppose it is true, as H.L.A. Hart and Tony Honor6 also claimed, that "the common sense view" of causation incorporates the libertarian metaphysics that alone could make sense of the legal doctrines of intervening causation. (40) Would those doctrines be justified because they are supported by popular belief, no matter how false such belief might be? Moral responsibility does not depend on what most people believe about it. Nor are the natural properties on which such moral responsibility supervenes--such as intention, causation, rationality, voluntariness, etc.--a function of popular belief. False beliefs, if popular, may win out in the democratic process that makes law. Such beliefs may thus explain why we have the legal doctrines we do. However, such doctrines are still unjustified.

The second route to justifying judges saying that the intervening acts of principals break causal chains would admit that in fact this is not true. (41) But some reasons of legal policy justify legal doctrines that bid judges to treat it as if it were true. Some rivers are legally navigable (for various legal purposes such as federal admiralty jurisdiction) even though, in fact, one cannot float a canoe on them.

The problem with this fiction-based justification of treating accomplices as noncausal lies in finding the appropriate legal policies. Prima facie, proximate cause doctrines (including those of intervening causation) serve a grading policy in the criminal law. We punish more severely those who (fully and unproblematically) cause the harm they intended to cause than we do those who try but fail to cause that same harm. (42) Our proximate cause doctrines exist to resolve the tough cases, where the harm intended or risked has come about but the causal route from defendant's act to the harm is freakish, attenuated, or in some other way abnormal. The doctrines serve to sort such cases, allocating some to be punished severely for causing harm, others less severely for only trying to cause (or risking the causation of) that harm.

This policy is one that seeks to correctly ascribe the degree of moral blame attaching to the actions of defendants. Yet, since moral blame is here supervenient upon causation, that requires that we get it fight about causation. There is no room for a fictional causation here. In fact, the one operative policy demands that we not fictionalize causation so that, if the acts of principals do not really break the chain of causation between accomplices and some harm in fact, neither should they be said to break such chains in law.

C. Accomplices Cause the Harms They Aid but Only in a Distinct, Secondary, and Anemic Sense of "Cause"

A view that turns out to be related to the libertarian view just considered is what I shall call the causal dualist view. On this view, the relation between an accomplice's act and some harm for which she is responsible is a causal one, but the kind of causation is said to be different when a second human agent is involved. Recall the earlier examples where I got you to assault your wife by providing opportunity, information, encouragement, or financial reward. In such cases, Hart and Honore tell us, you do not cause the apprehension of contact sufficient for an assault in the primary, strong, simple, or strict sense of cause. But there is a secondary, weak, complex, broad sense of cause according to which you can be said to cause this apprehension. Such broad sense of cause is said to be captured by the expression, "he occasioned the harm." (43)

It is important to see that this is not a merely quantitative distinction. The claim is not that there is but one type of causal relation and accomplices simply make a smaller causal contribution to some harm than do principals. Rather, the claimed difference is qualitative; there are "two types of 'causal connections,'" (44) or two "varieties": (45) a central type (or variety) that is broken by the intervening act of a guilty principal, and a secondary type (variety) that is not. The first is the causation between physical events, the second, causation between human agents. This then gives us an answer to our question about the relationship of accomplices to the harms they aid: "The 'causal connection' between a defendant's act and the harm may be succinctly described by saying that he has 'occasioned' it." (46)

While admirably succinct, this answer is, less admirably, false. Indeed, this answer is false for exactly the same reasons as is the answer just examined--that the intervening acts of principals break off the causal connection otherwise existing between accomplices and the harms they aid. For the only reasons given to support the existence of this weaker, lesser kind of causal connection ("occasioning") are the very same reasons as were given to support the idea of intervening causation: human beings are not billiard balls, but are free and not subject to the necessitation of causal laws. True enough, though here the claim is not that the intervening human choices of a principal are fully free of the causal influence of the acts of an accomplice; rather the claim is only that the accomplice weakly causes ("occasions") the principal's acts. In Anthony Flew's vocabulary, the accomplice inclines but does not necessitate that the principal so act. (47) Yet doesn't this make even less sense than a robust libertarianism? We might think libertarians can at least conceptualize an uncaused event, but what would it mean to speak of a partly caused event? It was sixty percent caused, forty percent uncaused? (48) People have said such things, but that does not mean they make any sense. (49)

D. The Supposed Inapplicability of the Causative Verbs of English to the Causings of Accomplices

The next argument I shall consider leaves the verbs referring to causation as such to what linguists call "causative" verbs, those verbs of action such as "kill" that seemingly require the actor to have caused something (such as death). The idea is that although accomplices may cause both their principal's actions and the harms those actions cause, nonetheless the accomplice does not herself do the action prohibited. On this view, an accomplice who gives a principal a gun (which the latter uses to shoot some victim dead) may be said to be a cause of the victim's death, yet that accomplice did not kill the victim. Only the principal did that. And since traditional criminal and tort law doctrine prohibits killings, hittings, maimings, etc.--not causings of death, causings of contact, causings of disfigurement, etc.--the idea concludes that accomplices cannot be guilty as principals even if they are causers of legally prohibited states of affairs.

The policies supporting this argument are two. First, there is the policy of legality. This policy regards the law we have as being a function of how ordinary people would understand it if they read it. If ordinary people distinguish killings from causing deaths, and if the law of homicide prohibits only killings, then accomplices cannot fairly be liable for the homicides they aid even if such aiding amounts to a causing of death.

The second policy is that of legislative supremacy. This policy regards the law we have as being a function of how legislators meant their words to be understood. Again, the claim is that legislators, like most native speakers of English, distinguished killings from causings of death; yet they explicitly only prohibited killings, so that accomplice-like causers of death were not meant to be included among those forbidden to kill.

Both policy arguments are premised on the truth of a linguistic claim, viz., the claim that accomplices who cause death by aiding another to kill are not themselves killers. The claim is that killing requires an act directly causative of death, a directness lacking when the principal's act intervenes.

This supposed differential closeness required for killing (as opposed to causing death) is not unique to cases where someone else more directly does the act in question. Consider this old example of Donald Davidson's: a doctor runs down a pedestrian in just such a way that the pedestrian's appendix is cut out by the doctor's sharp edged Lincoln Continental. (50) Has the doctor removed the pedestrian's appendix? Davidson's intuitions were to say that although the doctor caused the removal of the appendix, he did not perform the action of removing the appendix. (51)

My own sense of these kinds of examples, argued for elsewhere, (52) is that the oddness of saying that the doctor removed the appendix is only pragmatic, not semantic. When we say, "the doctor removed the appendix," we have a picture, a stereotype of how it transpired: white gowns, surgical instruments, in a hospital, etc. Yet these are only context-driven features of appropriate utterance (i.e., pragmatics). Literally, the doctor did remove the pedestrian's appendix with his Lincoln. So if removing an appendix is a crime--a kind of mayhem, let us suppose--the doctor has done the actus reus of it.

This expands the reach of causative verbs like "remove," "disfigure," "kill," etc., toward the reach of "causing removal," "causing disfigurement," "causing death," etc. Yet it does not reach the kind of case relevant here. Suppose A tells P where V can be found, P finds V, and hits him. Even if we say that A's act was a cause of the unlawful contact on V's body, we should be reluctant to say that A hit V. P did the hitting of V, no matter how much A helped P to get into a position to be able to do it. So if battery is defined as hitting another (rather than "causing contact"), only P but not A can be guilty of battery.

The problem with this entire line of justification (for hiving off accomplices from principals in the context of criminal liability) is the arbitrary contingency on which this argument turns. If accomplices cause legally prohibited harms just as do principals, and if the causative verbs of English require more than mere causation, then it is a poorly drafted code that defines battery as hitting (rather than causing contact), or that defines homicide as killing (rather than causing death), or that defines mayhem as maiming (rather than causing disfigurement). The obvious remedy is not to use causatives in the drafting of penal codes, given their alleged divergence from what is important--causation. (53) Such more careful drafting, of course, alleviates any problem with legality or legislative supremacy that may exist under codes as currently drafted.

More careful drafting would of course be desirable only if it is indeed causing that is normatively important, not killing (hitting, maiming, raping, etc.). One might think, for example, that the agent-relative obligations of morality (on which the malum in se crimes are based) are directed at killings, not at causing deaths; at hittings, not at causing contact; etc. If this were so, then the restriction of principal liability to killers, hitters, etc. (leaving accomplice liability for mere causers of death who are not killers), might well be justified. If, in other words, the linguistic difference between killers and causers of death is also a moral difference, then there may be good reason for there also to be a legal difference here.

Some of the British homicide cases are seemingly illustrative of this line of thought. Unlike the Model Penal Code, British homicide law purports to mark the categorical nature of the obligation not to kill by refusing duress as a defense to murder, no matter how serious the harm threatened to the defendant. Yet at one point in time the House of Lords allowed an accomplice to murder to avail himself of the duress defense; the driver who aided the death of an innocent was not categorically obligated not to so cause death in the way the actual killers were so obligated. (54)

Yet it is not the killing/aiding-another-to-kill distinction that is doing the moral work here. Rather, it is the degree of causal contribution. One who merely drives the IRA gunmen to where they can kill the British policeman is much less of a cause of the policeman's death than are the acts of shooting done by the gunmen themselves. One can see this by increasing the causal contribution of the aider--suppose he holds the victim in a position where the gunman can then shoot and kill. Now the categorical obligation seems to be as much in force as it is against the literal killers. (55)

What makes this point harder to see is that killers (hitters, rapists, etc.) are, as a class, more significant causal contributors than are those who enable others to do such acts. So killing, hitting, etc., are proxies for the kind of significant causal contributions required for there to be a breach of our categorical obligations. But we must not confuse the proxy for the underlying desert determiners; what matters morally is significant causal contribution, not the kinds of limitations marked by the causative verbs of English. (56)

To see the same point from another vantage point, repair to another linguistic feature of causative verbs like "kill." Linguistically, we distinguish people who kill from those who do acts enabling large forces of nature to kill. Suppose a freakish wind blows off the roof of a factory, blowing the roof weighing many tons a considerable distance; such freakish forces of nature are as regularly regarded as intervening causes as are the free, informed, voluntary acts of human agents. (57) Now suppose that the defendant foresees the wind and accurately predicts its effect on the roof; wanting some victim V dead, she drives V to just where the roof falls on him, crushing V to death. We might well be tempted to say that the defendant's driving V to a certain location causally contributed to V's death, but the defendant didn't kill V--the wind (or roof) did that.

Suppose one thought that such "aiding of nature" in this case was outside of morality's categorical obligation not to kill (i.e., that such acts of driving a person to where he would be killed were eligible to be justified by good consequences). But wouldn't one nonetheless distinguish the case of a more significant causal contributor? Suppose the defendant holds V in just the right position for the wind-driven roof to crush him. (58) Linguistically, it may still be the wind/roof that may properly be said to have killed V. But morality is not so formalistic and obtuse for us to think that that matters. Those who manipulate nature in this way are as surely obligated categorically not to so cause death as are literal killers. Again, it is the degree of causal contribution doing the heavy lifting here, not the niceties in the semantics of the causative verbs of English.

E. Nonproxyable Crimes and the Need for Accomplice Liability

There is a related but distinct problem left unaddressed by this redrafting proposal. This is the problem Sanford Kadish refers to as "nonproxyable" crimes. (59) Suppose (as many first year torts students tend to think) battery were defined not just as causing contact with the body of another, but as requiring that such contact be with the actor's own body. Then the crime would be nonproxyable, to use Kadish's term, because even if A caused P to hit V, he does not cause P to hit V with A's own body. Nonproxyable crimes, as Kadish defines them, are crimes that "can be done by the actor only with his own body and never through the action of another." (60)

Battery of course is not defined this way--causing contact of the victim's body with the floor, a bat, someone else's body, or anything else is enough. So consider some of Kadish's examples of nonproxyable crimes:
   A sober defendant may cause an insensate and disorderly drunk to
   appear in a public place by physically depositing him there. But
   we could hardly say that the sober person has, through the
   instrumentality of the drunk, himself committed the criminal
   action of being drunk and disorderly in a public place. A
   defendant may cause a married person to marry another by falsely
   leading the married person to believe his prior marriage was
   legally terminated. But the defendant could hardly be held liable
   for the crime of bigamy, since one does not marry simply by
   causing another person to marry.... [A] visitor to a prison who
   abducts a prisoner could not be said to have committed the crime
   of escaping from prison by forcing a prisoner to do so. (61)


Redrafting such statutes using "cause" language (rather than causatives) will not make these crimes "proxyable"--i.e., doable--through the actions of another. Suppose we change the actus reus of bigamy from "marry while still married to another" to "cause oneself to be married while oneself is still married to another." This will not make Kadish's supposed deceiver guilty of bigamy, because she isn't causing herself to get married (any more than she is doing any marrying). This is because the event such nonproxyable crimes seek to prohibit essentially involves reference to the actor being charged: it isn't criminal to cause another to get married, only to cause oneself to get married.

This linguistic feature of pronominal cross-reference may also be arbitrary, depending on what one thinks is wrong about bigamy, public drunkenness, prison escapes, and the like. Is the evil sought to be prohibited by the escape statute, for example, the evil of prisoners being outside the prison walls, no matter how they got there? Or is it the evil of prisoners getting themselves out of prison? If the latter, then we wouldn't rush to redraft because the class of persons we are aiming at--prisoners--aren't involved in these visitor-caused "escapes," except as victims. In such cases, we neither want to expand principal liability nor to use accomplice liability, because the evil we wished to prevent never happened.

As another example, consider rape. Is rape a nonproxyable crime? Should rape be defined as causing penetration of another or as causing penetration of another by one's own body? (62) That depends on what one takes the evil of rape to be. Suppose (as in one well-known case) the defendant inserts the penis of another into the female victim. (63) That action should constitute rape if the evil is the violation of the female's body, but not if the evil is violation of the female's body by the perpetrator's body. In making the crime nonproxyable, the Virginia Supreme Court implicitly took the latter view. (64) On that view, there is no rape (and no aiding of rape) because the evil of rape did not occur.

Nonproxyable crimes thus present no reason to require a separate form of liability called complicity. Doubtless most crimes should not be defined in such a way as to be nonproxyable--battery, homicide, and rape coming to mind. But for those crimes that are rightly so-defined, such as (perhaps) bigamy and escape, no form of liability should attach, accomplice or otherwise, when the evil prohibited hasn't occurred--because the perpetrator hasn't caused himself to be in a certain state. For proxyable crimes, there is no need for accomplice liability, for principal liability will attach to accomplices who cause the prohibited harm by their aid; and for nonproxyable crimes, there is no need for liability of any kind for "accomplices" who do not cause the prohibited harm (because that harm has not occurred at all). In short, the existence of nonproxyable crimes goes no distance in showing a need for accomplice liability.

III. BEGINNING AGAIN: THE BASES OF ACCOMPLICE LIABILITY

If we put aside these bad reasons for thinking accomplice liability to be noncausal (or not fully causal, or not describable with causative verbs), our puzzle recurs: is the relationship between the acts of accomplices and the legally prohibited state of affairs they aid causal or not, and if not, what is the relationship?

We now need to employ a battery of distinctions argued for here and elsewhere. We need to distinguish among (1) true causation, (2) counterfactual dependence (as we did earlier), and (3) probabilistic dependence. We further need to distinguish (4) what I shall dub "intentional dependence," the kind of relation that exists between a harm and an act, when that act executes an intention whose object is a representation of that harm. The actus reus of accomplice liability is constructed out of one or more of these four relationships (between the accused's act and the prohibited result) being present in some combination--which is why that actus reus has so resisted easy characterization. Sometimes it is one thing, sometimes another, and sometimes it is a combination of two or more things. Let me begin with what I shall call the "truly causal" accomplice.

A. Truly Causal Accomplices

As we have seen, it is not essential that an accused's act be a necessary condition of some harm for that act (both in law and in actual fact) to be a cause of that harm. Nor is counterfactual dependence of the harm on that act a sufficient condition of causality. This means we need something other than the presence or absence of a "but-for" relationship to separate causal from noncausal accomplices. What might this be?

We fortunately do not need a fully worked-out theory of causation here. Negatively, we do need causation not to be counterfactual dependence, probabilistic dependence, or intentional dependence. We need this in order to retain the distinctions organizing this discussion. But positively, we need only what the first and second iterations of the Restatement of Torts needed: to treat causation as a primitive--a "factor"--save for one characteristic, namely that there can be more or less of it. (65) Causation needs to be a scalar quality, a matter of continuous variation, as opposed to a binary, black and white sort of relationship. (66)

Then (again with the First and Second Restatements of Torts) we mark causal responsibility by using some quantitative measure: X causes Y if and only if X is (say) a "substantial" cause of Y. (67) De minimis amounts of causal contribution do not make for cause-based responsibility. This line (between substantial and de minimis amounts of causal contribution) is not governed by counterfactual dependence. X can be a substantial cause of Y without X being necessary for Y--witness the fire that joins another fire of equal size, and the two together jointly burn down the victim's house. Moreover, X can be a de minimis causal contributor even though X was necessary for Y--as where I remind you of how much you hate V, and (with that reminder) you go off and kill V, when otherwise you wouldn't have.

This spare notion of causation gives us all we need to assess the causal contributions of accomplices. And, not surprisingly, if we apply this substantiality criterion, we find that some accomplices are causes of the harms they aid another to cause, whereas other accomplices either make no causal contribution to a harm by their aid, or they make quite small causal contributions to that harm.

Consider first the substantially causally contributing accomplices. It is perhaps helpful to follow the law here, dividing these cases along a temporal dimension--between those whose acts of aid are simultaneous with the acts of the principal causing the harm and those whose acts of aid precede the relevant act(s) of the principal. The law makes this division because it treats simultaneous, substantial causal contributors as coprincipals, not as aiders. (68) Thus, one who holds up the barrel of the gun so that a (weak) shooter can make the shot that kills is treated as a concurrent causer of the death, not as an aider and abettor of another's causing of death. Likewise, the defendant who knifes the victim while another repeatedly strikes her with a saber is held to concurrently cause her death, not to be an accomplice.

Some of the sequentially contributing cause cases also treat those who aid as co-causers and thus as principals, not accomplices. I refer to the innocent (or partly innocent) agent cases. (69) These are the cases where the putative accomplice (a) gets the perpetrator so drunk that he barely knows what he is doing; (b) induces a hypoglycemic episode in another, who involuntarily does what is needed; (c) deceives the actor about the kind of act he's doing; or (d) coerces the actor with serious threats* In such cases there is no guilty principal, so there can be no accomplice liability. Yet the law here has no problem in seeing the first actor as a co-causer along with the second, thus treating that actor as a principal in his own right. (70)

Whenever the doctrine of intervening causation treats the second actor's choice as an intervening cause, the sequential cases come out quite differently. In such cases, the law classifies earlier (but substantial) causal contributions as not amounting to causal liability; rather these actors are treated as accomplices. (71) Yet once one sees that the doctrine of intervening causation does not mark a true causal distinction, there is no justification for treating the two classes of sequential causal contribution cases differently. Indeed, both sorts of sequentially co-causing cases are in reality no different than the simultaneous co-causer cases above. All of these three classes of cases represent the true causal liability of co-causers, despite the fictionally "broken" causal chains supposedly existing in the last of these three sorts of cases.

Thus, one who picks the victim of the murder, orders a hit man to do it, pays him well for it, locates the victim for the hit man, brings the gun and ammunition, and drives the hit man to the location of the killing, substantially causes the death of the victim. We should thus say plainly that one way to be an accomplice is by causing the harm through the actions of another. Substantially aiding another to cause some harm is to substantially cause that harm oneself, whatever the pretensions of the intervening causation fiction.

Seeing the parallel between the various kinds of substantial causers--whether simultaneous or sequential--raises the question of why the criminal law would distinguish between them. All substantially cause the harm, so why is one treated as an accomplice and the others treated as principals? One could say that, on average, accomplices are less-substantial causers than are the principals they aid, and this is true enough. Yet this is only a rule of thumb, something that is true in the general run of cases. It does not justify the bright-line rule marked by the distinction between principals and accomplices. Some very substantial causers will fall on the accomplice side of the line, as in the last example; conversely, some not-so-substantial causers (such as one of thirteen concurrent stabbers of a single victim) (72) will fall on the principal side of the line. Indeed, under current doctrines, the very same defendant will be treated as an accomplice or as a principal depending not on the degree of his causal contribution but only on the relative innocence of his co-causer.

Let us next examine the smaller causers of the harm aided. Take everyone's favorite here--the Wilcox case in England. (73) Wilcox was charged with aiding an alien saxophonist in performing for money whilst in England, in violation of the immigration laws then in effect. The only thing Wilcox did was buy a ticket and attend the concert where the saxophonist was playing, for the purposes of writing it up for his journal, Jazz Illustrated. The court held that Wilcox's "presence and his payment to go there was an encouragement" sufficient for accomplice liability. (74)

It is important first to see that Wilcox did causally contribute to the legally prohibited result--here, the performance for pay of an alien musician. That is because Wilcox was a member of the paid audience, and presumably the presence of a paid audience entered into the motivations of the saxophonist who played. Wilcox is in this respect like other of what I have called mixed concurrent cause cases, such as pollution and elections. (75) Suppose it requires five units of pollution to cause some harm, but each of several hundred polluters contributes one unit; or suppose it requires fifty-one votes to win an election, but ninety people vote for the winner. In such mixed concurrent cause cases no individual polluter, voter, or audience member is necessary to the harm occurring, nor are they sufficient; still, they are causes of that harm.

Yet they are, considered individually, very small causal contributors to the result. At some point, surely, they become de minimis causes, to be treated for moral and legal purposes not as causes at all. Wherever that point is, it should be the same for cases where the result is the action of someone else (as in Wilcox) as it is where the result is an event that is not a human action (as in the pollution and election examples). Whatever is a sufficient causal contribution for liability as a principal, in other words, should be sufficient for liability as an accomplice, and vice-versa. Causal accomplices should again be treated like principals, because both groups equally cause the harm.

B. Necessary Accomplices

Once one distinguishes causation from counterfactual dependence--as we did earlier--then it becomes possible to see one noncausal basis for being an accomplice, namely, counterfactual dependence. Suppose, in the Tally case earlier discussed, (76) the warning telegram would have saved the victim, Ross; this means that Tally's sending of the countermanding telegram was necessary to Ross's death--that is, but for Tally's act, Ross would not have died. In such a case, Ross's death counterfactually depended on Tally's action, even if that action in no sense caused that death. Let us call such noncausal, counterfactually related accomplices "necessary accomplices."

What we ultimately wish to examine is how the presence of counterfactual dependence can be a sufficient desert basis for necessary accomplices, even in the absence of causation. Preliminarily, however, consider a related point: can the absence of counterfactual dependence limit or eliminate the responsibility of admittedly causal accomplices? In other words, is counterfactual dependence necessary to blameworthiness, so that its absence limits the cause-based liability of causal accomplices?

The general answer, as we have seen, is in the negative here: accomplices are liable for aiding their principals even if their aid was not necessary to the harm caused by the principals. Yet consider again the triad of (a) concurrent overdetermination cases, (b) asymmetrical overdetermination cases, and (c) preemptive overdetermination cases. (77) Suppose that Tally was a causal accomplice in the Skeltons' killing of Ross--he supplied the guns and ammunition, helped position Ross so he could be shot, etc. But also suppose (a) a large boulder happened to fall on Ross just when the Skelton/Tally group shot him, and immediately mortal wounds were inflicted by both the boulder and the shots; (b) the Skelton/Tally band inflicted non-mortal wounds, whereas the boulder inflicted wounds that by themselves would eventually prove mortal, and Ross died from loss of blood from both sets of wounds; or (c) the Skelton/Tally band shot and killed Ross, but seconds later the boulder fell on Ross's body in such a way that it would have killed Ross if he were not already dead from the shots of the Skeltons and Tally.

Notice that Tally's acts, like those of the Skeltons themselves, were not necessary for Ross's death; in each case, the boulder was sufficient to cause the death. So long as the alternative factor is a natural condition (such as a boulder falling of its own accord), not wrongdoing by some other parties, there is some authority that Tally and the Skeltons are not liable (at least in tort) for Ross's death. (78) This is because Ross's death does not counterfactually depend on what Tally and the Skeltons did, even though they caused (or causally contributed to) the death in each case.

I have argued elsewhere that each of these three doctrines is by and large a mistake and that causation of death by Tally should be sufficient to ground his liability. (79) There are only two qualifications to this conclusion. First, in the asymmetrical case, if Tally's causal contribution drops below the de minimis line, he is not blamable for Ross's death. But even here, it is not the absence of counterfactual dependence that is doing the work, but rather the absence of "substantial" causation. Second, in the preemptive case, the absence of counterfactual dependence does make a striking difference to a consequentialist justification for killing Ross. I refer to the "almost dead" cases, where, because one is merely accelerating a death about to happen anyway, one may kill to achieve a net savings of lives. It is a puzzle why absence of counterfactual dependence makes such a striking difference to such permissibility when (on my view) it makes no difference to culpability. (80)

With these qualifications, my own conclusion is that counterfactual dependence is not necessary to accomplice liability in any cases for which cause-based liability can be made out. Causation, in other words, is an independent desert basis. So, I shall now argue, is counterfactual dependence. Return to my hypothetical Tally, who merely sent a telegram, countermanding the telegram that would have saved Ross, and notice two facts about him. First, Ross's death in this case does depend on Tally's act, and second, that act did not cause that death.

The first point is obviously true, if for no other reason than that I stipulated it to be so--and it's my hypothetical! The second point is less obvious. The Tally we are now discussing is a double-preventionist, not a causer, of Ross's death. My hypothetical Tally prevented something (the warning telegram) from doing what it would have done (prevented Ross's death), and by preventing this prevention Tally allowed nature (here, the Skeltons) to take its course.

To see why such double preventions are noncausal requires that we return to the basic notion here, that of a simple prevention. (81) When I prevent someone's death by throwing her a rope, it is idiomatic to describe what I did as causing something--namely, the would-be drowning victim's survival. Yet the victim's survival is not a positive event or state of affairs; surviving something is just the absence of dying because of that thing. (82) So one might say we caused a negative event or state of affairs to exist, viz., a not-dying of the victim.

Negative events, negative states of affairs, and negative properties are difficult items to admit into one's ontology. There are of course negative propositions (sometimes called "facts"), such as the fact that V did not drown today. Yet such negative propositions do not demand negative entities or properties. Rather, what we rather transparently mean by "V did not drown today" is that "it is not the case that there was some instance of the type of event, drowning of V, that occurred today." We do not here refer to a negative event as a particular--a not-drowning event--any more than we refer to a particular non-elephant in the sentence, "there were no elephants here today." Rather, such statements are negative existentially quantified statements requiring reference only to types, either of events (like drownings) or objects (like elephants).

If there are literally no such things as negative events, negative states of affairs, and negative properties, how can such "items" enter into singular causal relations? More specifically, how could one cause something to exist (a not-dying) that we admit does not and never did exist? We cannot, so preventions of death cannot be causings of not-dyings. What preventions are is a somewhat tricky business to spell out. But part of their nature we already have before us: when I prevent the death of V, the fact that V did not die does counterfactually depend on the fact that I did what I did (such as throw a rope to V). Such counterfactual dependence does supervene (in part) on certain causal facts--such as that my throwing the rope caused it to be where V could reach it, and V's reaching of the rope and hanging on to it while it was raised out of the water is inconsistent with V's drowning. But the details of this need not here detain us. (83)

Preventions themselves are not all that important to the issues that here concern us. But the (noncausal counterfactual) idea of a prevention is the basic building block for two items that do concern us. First, consider the notion of an omission. Although one can omit to do any type of action (such as killing, maiming, etc.), the omissions of interest to us are failures to prevent something. Thus, in the much-publicized New Bedford rape case of some decades past, several patrons of the bar where the rape took place stood by and did nothing. (84) They didn't cheer or in any other way positively aid the rapists; but they also did nothing, such as intervening or calling the police, to prevent the rape. They simply watched and, in so doing, omitted to prevent the rape. A much-discussed issue was whether such bar patrons' failure to prevent the rape should be sufficient for accomplice liability for the rape.

Suppose the chief stumbling block to accomplice liability in the actual case was here removed. Suppose, that is, that the bar patrons had a duty not to omit, as they would, for example, if the rape victim were their child. How should we analyze their liability/ blameworthiness for the rape? Not causally, from what we concluded about preventions. If preventing a rape is not to cause a non-rape, then not doing anything to prevent a rape is surely not to cause a rape. If there are no negative events that can be effects of some cause, surely there can be no negative events that can be causes of certain effects. "Nothings" cannot cause "somethings" any more than "somethings" can cause "nothings." Absent events cannot serve as relata (of either kind) of the singular causal relation.

Plainly, what is sufficient for omission liability in such cases is counterfactual dependence. The fact that the victim was raped counterfactually depended on the fact that the bar patrons omitted to prevent it. Put more simply, they had the ability to prevent the rape and didn't, and that is sufficient to ground their responsibility. Causation has no role to play in this kind of responsibility, with ability (i.e., counterfactual dependence) doing all of the work that needs doing.

Secondly, return to the notion of a double prevention. If preventions and omissions are noncausal, then so too are double preventions. Consider an old example of mine. (85) D sees his old enemy, V,, drowning in the ocean of natural causes. D rejoices. Then D sees L, the Mark Spitz of lifeguards, preparing to save V; to prevent this, D ties up L so he cannot swim out to save V,, and V drowns. D's act of tying up L prevented L from preventing D's death. D is plainly responsible for V's death, but not because D caused that death. Rather, V's death counterfactually depended on D's act, and this is enough to establish responsibility.

The conclusion that there is no causation here follows from what we said about prevention and omissions. D did not cause an absence of saving by L, and L's failure to save did not cause V's death. Double preventions cannot be causal if preventions and failures to prevent are not causal.

D in my hypothetical is what I have elsewhere called an "aider of nature," (86) not an accomplice to some human killer. Because of this, the law would thus classify D's liability as principal liability and not accomplice liability. Yet the difference is artificial and incidental. Tally's double prevention would ground accomplice liability because he aided the Skeltons rather than a boulder or an ocean current; still, the basis of Tally's accomplice liability is no different than D's principal liability. Both are noncausal, double-preventionists whose liability is based on the fact that the deaths in question counterfactually depended on their actions.

An interesting question is whether such necessary but noncausal accomplices are less blameworthy as a class when compared to causal accomplices. Does mere counterfactual dependence, unaccompanied by causation (because of the double-preventionist nature of such examples), make for a lesser blameworthiness? Perhaps it helps to creep up on this issue by considering a subclass of such cases where there is plainly a large moral difference. I refer to what moralists for centuries have been calling cases of allowings. (87)

Consider the paradigm cases of nonomissive allowings, the passive euthanasia cases. When the doctor disconnects the respirator from a patient, he is said to let the patient die, not to kill the patient. More generally put, he allows the patient to die, he does not do the action of killing, nor does he cause the patient's death. These are double-preventionist cases with counterfactual dependence. But such cases have a third property, one that makes a large difference morally. I call this third element the return to a morally appropriate baseline. (88) When the doctor unhooks the machine, she returns the patient to the state he would have been in had the machine never been hooked up to start with--that is, a dying state. What typically makes such state a morally appropriate baseline to which to return is that (with the advantage of the hindsight available at the time of disconnection) we can now see that the machine is incapable of doing any real good for this patient; so if we knew then what we know now, one would have been justified in not hooking the patient up to start with.

While the use of such baselines is close to the surface in many other standard instances of allowing, consider how it could be applied to variations in the facts of my hypothetical D and of Tally. Suppose it was D himself who was about to swim out to save V, or suppose it was Tally himself who had sent the warning telegram. Suppose further there was no duty making it incumbent on either of them to do these things. Then when D ties himself up (so as not to yield to temptation to save V), and when Tally sends the countermanding telegram, I take it that they only allow the death of V and of Ross. Because these acts do no more than return them to a morally appropriate baseline (one where they had no duty to intervene to start with) they have no liability--this, despite being double-preventionists whose later acts are necessary to the deaths of V and of Ross.

Indeed, in these revised scenarios their blameworthiness/liability seems no greater and not much different from simple omission cases, where they do not even begin to save their victims. To be sure, such allowers are not literally omitters; they do act. Yet their two actions together are almost self-cancelling. Hooking up a respirator that one then later unhooks allows the patient to die pretty much the death he was headed toward before there was any intervention. I say "pretty much" because the patient's death is delayed somewhat by the combined actions of hooking and unhooking. Yet like the acceleration cases, (89) the allower leaves the world pretty much where it was headed without him. The accelerator makes such death slightly earlier; the allower makes it slightly later. Both are close to the pure omitter, who leaves things completely alone.

Now subtract the moral baseline element of allowings. On the actual facts of the Tally case, Tally was not the one who sent the warning telegram. So such a telegram unsent was not a morally appropriate baseline to which Tally could return. This crucial element of an allowing is missing. Tally is only what might be called a "partial allower"--a double-preventionist whose acts are necessary to some harm but whose acts of prevention do not return the victim to some morally appropriate baseline (from which nature can take its course without blame). Still, aren't partial allowers somewhat less blameworthy than causers?

Consider in this regard the much-discussed interrogation techniques used by American intelligence officers in the contemporary "war on terror." (90) The core of these techniques was derived from those developed by the British in their own war on terror in Northern Ireland, labeled "stress and duress" techniques. (91) Some of these are cases of potential allowings, such as sleep deprivation and sensory deprivation. (Others are pure omission techniques, such as failing to feed, hydrate, or treat medically.) Sleep deprivation techniques (such as bright lights or loud and irritating music) prevent sleep and, as the Bard tells us, sleep "knits up the raveled sleave of care" (92) (i.e., it prevents a disoriented state). Sensory deprivation techniques prevent the perception of light, sound, touch, taste, and other stimuli; such perceptions again prevent disorientation of various kinds. So such techniques are double preventions on which the disorientation, anxiety, etc., of the victim counterfactually depend. And there is no pretense that such states (of over- or under-stimulation, respectively) are morally appropriate baselines to which return may be made with impunity.

Even so, such techniques are distinguished from the kinds of causation of pain and distress involved in true torture (electric shocks, cuttings, beatings, and the like). The European Court of Human Rights termed these "inhumane and degrading treatment," but not torture, in its review of British policies. (93) Even torture victims acknowledge that these so-called "torture-lite" techniques are regarded differently than are the cause-based, more traditional modes of torture. (94) Both, of course, merit heavy blame (in the absence of compelling justifications). Yet isn't there a difference in the degree of blame, just the same? Acts of "torture lite," like other partial allowings, involve less of the actor's agency because they are noncausal. Nature does the dirty work, so to speak, and the actor's agency merely removes the impediment that was holding nature back.

Notice that to so conclude is not to justify a differential punishment for accomplices vis-a-vis principals. That distinction I will get to at the end of this Article. The pie-slicing here is different. The idea is that both necessary, noncausal accomplices and necessary, noncausal principals are somewhat less blameworthy than are both causal accomplices and causal principals. The difference here cuts across the accomplice/principal line, and cannot justify whatever moral difference that line is supposed to mark.

C. Chance-Raising Accomplices

We saw earlier that a common shibboleth in law about causation was that causation of E by C is identical with counterfactual dependence of E on C. An almost as common saying about causation within legal circles is this one: a cause is the raising of the probability of its effect. More exactly: "C causes E" means P(E/C) > P(E). (95) The idea is that what causation amounts to is the chance-raising of an effect by its cause. We first need to show why this too is a myth in order to see the possibility of a third desert basis for accomplices--that of a noncausal, chance-raising kind.

In considering the identity or extensional equivalence of causation with chance-raisings we need not take a position on the semantics of probability statements. By "semantics" I refer to the various interpretations of what probability itself is--whether it is subjective or objective, Bayesian or non-Bayesian, relative frequency or propensity, etc. (96) For now, we can do with a syntactic notion of probability: the probability function (P in the earlier statement) is one that obeys the axioms of the probability calculus. This puts aside the semantic question of what corresponds with that function in the world.

As with counterfactuals, it is implausible to identify causation with the raising of conditional probabilities. Again, if two descriptions refer to the same thing, then that thing will possess all the same properties no matter which description is used. And again, causation is a transitive, asymmetrical, and temporally ordered relation, whereas chance-raising (like counterfactual dependence) shares none of these properties. Suppose C causes E, and E causes F; then (with qualifications not relevant here) C causes F. Whereas if the probability of F given E is fifty percent and the probability of E given C is fifty percent, then (by the probability calculus) the probability of F given C is only twenty-five percent, and transitivity is not fully preserved. Likewise, if C causes E, E does not cause C. By contrast, if the probability of E given C is fifty percent, then the probability of C given E can also be fifty percent. Symmetrical probabilistic relations are possible. Finally, if C causes E, E does not precede C in time. Whereas when C increases the conditional probability of E, C can succeed as well as precede E in time. Epiphenomenal forks provide frequent examples. (97)

So identification of causation with an increase in conditional probability is implausible. But (again parallel to our discussion of counterfactuals), perhaps chance-raising is a good test of causation; perhaps there is an extensional equivalence of chance-raising with causation even if there is no identity. Again, we should examine the alleged equivalence in two steps. First, is the chance-raising of E by C sufficient for the truth of "C causes E"? And, second, is the chance-raising of E by C necessary for the truth of "C causes E"?

The claim that chance-raising is sufficient for causation must cross most of the same hurdles as must the like claim on behalf of counterfactuals. Absent events must be causes, because the fact that something did not happen can raise probabilities as well as the fact that something did happen. Also, epiphenomenal events are indistinguishable from causally related events. Recall that when I run in the morning, this causes my feet to get tired, and, a little later, it causes my accompanying dog to get tired. Since my tired feet raise the conditional probability of my tired dog--the probability of my dog getting tired is greater if my feet are tired than if they are not--the one must cause the other. Worse, given the possibility of symmetrical probabilistic dependence, my dog getting tired can raise the probability of my feet getting tired. So causation works backwards through time, too.

In addition, consider a final hurdle, a case having no parallel to the counterexamples to the counterfactualists' sufficiency claim. There are two would-be assassins, A and B, each of whom fire at a single victim at the same time. A's bullet would have hit and killed V, except for the totally unforeseeable, coincidental flight of a large bird that took the hit instead; B's bullet hit V just where and when A's bullet would have hit, and B's bullet kills V. Presumably, A's shot raised the probability of V's death--if, for example, each shot had a fifty percent likelihood of killing V, two such shots (when the success of each is independent of the other) has a seventy-five percent likelihood of killing V. So A caused the death of V too? (98)

Now consider the necessity of chance-raising to causation. Take one of the kinds of cases we used against the counterfactualists' sufficiency claim, that of preemptive causation. Suppose in the two-shooters case just imagined there is no bird. Instead, A fires first, and this deters B from shooting at all. Suppose further that B was a better shot than A; B had a seventy-percent chance of success, whereas A had only a thirty-percent chance. Despite this, A's shot killed V. It looks like A's shot caused V's death even though A's shot lowered the probability of V's death vis-a-vis what it would have been had A taken no shot (because B would then have shot). (99)

Another variety of counterexamples where admitted causes lower (rather than raise) the probabilities of their effects is the "hard way" cases. (100) Consider a classic: at one time in the past taking birth control pills was known to cause thrombosis (through known causal processes). (101) Yet pregnancy is also a cause of thrombosis. Indeed, getting pregnant raised the probability of thrombosis more than did the taking of those kinds of birth control pills. Because taking birth control pills prevents pregnancy (to a very high probability), the net effect of taking such pills was to lower the probability of thrombosis--this despite the admitted fact that taking the pills causes thrombosis.

As with counterfactuals, there is a vast literature on all of this. (102) Having surveyed much of that literature elsewhere, my conclusions on counterfactuals have been uniformly negative: counterfactual dependence is not to be identified with causation, nor is such dependence either necessary or sufficient for causation. My conclusions on probabilistic dependence are less uniformly negative: singular causation is not to be identified with such dependence, nor is such dependence sufficient for causation. By contrast, probabilistic dependence is a necessary concomitant to causation (with some stipulations to accommodate the "hard way" and preemptive overdetermination counterexamples). (103)

These metaphysical conclusions are enough to make possible the moral thesis I now want to advance: that chance-raising is an independent desert basis, along with causation and counterfactual dependence and, further, that some accomplices are only blameworthy on this basis, since they are neither causal nor necessary accomplices.

Liability for chance-raising is well known in both civil and criminal liability doctrines. Examples of the latter are the attempt and reckless endangerment liabilities in criminal law. An example of the former is the tort of "lost chance," where a negligent doctor deprives terminal patients of what little chance of life they may have had. (104) Accomplice liability doctrines are also often stated in probabilistic terms; many cases conceptualize aid as acts that raise the chance of success of the principal's acts. (105)

Underlying all of these doctrines is the moral fact that risk imposition is a desert basis independent of causation and of counterfactual dependence. When I refer to risk impositions, I refer to some objective notion of risk. (I consider a purely subjective version of inchoate liability in a subsequent section.) Some contend that there is no objective notion of probability that can make sense of this objective idea of risk imposition. Such theorists urge that only subjective interpretations of probability make any sense. I shall, however, continue to suspend the debate about the correct interpretation(s) of probability. Even working with subjective probabilities alone, we can distinguish the wrong of risking from the culpability of the actor who subjectively thinks he is imposing risk.

To maintain neutrality about subjective versus objective interpretations of probability only requires that we idealize the subjective assessment of probabilities. Probability can then remain a purely epistemic notion, yet we can distinguish unreasonable risk impositions as judged from some idealized epistemic position from unreasonable risks as judged from some particular actor's actual epistemic position. The former gives us all the objectivity to risk we need to make out the desert basis with respect to risk imposition.

To assess how blameworthy one is, we need to add a culpability judgment to this judgment of objective risk. In cases of inadvertent risk creation (such as in the tort of negligent risk creation for terminal patients), we would ask, first, whether some act imposed an unreasonable risk from some idealized epistemic vantage point (which may well be the factfinder's own, with the advantage of hindsight) and, second, whether from the information base reasonably accessible to the actor, the act created an unreasonable risk. In cases of knowing risk imposition (such as in the crime of reckless endangerment), we should ask the same "objective" (i.e., idealized) question about risk creation, and then ask the culpability question of whether the actor subjectively perceived such a risk. In cases of purpose (or "specific intent"), such as in attempt liability, we again ask an "objective" risk question: was the risk substantial enough to warrant liability? And separately, we ask the culpability question of whether the actor either intended the harm risked or (in rare cases) intended the risk itself. (106)

To say that riskings are wrongs in the sense just described is not to say that "being hit with a risk" is a harm to the person risked. Nor is it to say that the victims of unrealized risk impositions have a right not to be risked. If the probabilities involved in riskings are subjective only, then one's interests are neither harmed nor violated by risks that have not materialized. Risking a harm is not like causing a harm in these ways. (107)

Still, to unreasonably risk is to be blameworthy, the degree of blame (as for causing) depending on the culpability with which the risking is done. On retributive principles, the unreasonable risker deserves punishment as does the unreasonable causer. It may well be, as I have argued elsewhere, (108) that such unreasonable riskers deserve substantially less punishment than causers of equal culpability. Even so, the risking is its own desert basis.

Criminal law presupposes this as much in its accomplice liability doctrines as in its attempt/endangerment doctrines. In the actual Tally case, for example, the Alabama Supreme Court first noted that Tally need not have been a causal or a necessary accomplice:
   The assistance given ... need not contribute to the criminal result
   in the sense that but for it the result would not have ensued. It
   is quite sufficient if it facilitated a result that would have
   transpired without it. It is quite enough if the aid merely
   rendered it easier for the principal actor to accomplish the end
   intended by him and the aider and abettor, though in all human
   probability the end would have been attained without it. (109)


The court then went on to define "facilitating," "aiding," and "rendering easier" in probabilistic terms:
   If the aid in homicide can be shown to have put the deceased at a
   disadvantage, to have deprived him of a single chance of life which
   but for it he would have had, he who furnishes such aid is guilty,
   though it cannot be known or shown that the dead man, in the
   absence thereof, would have availed himself of that chance; ...
   [as] where he who facilitates murder even by so much as destroying
   a single chance of life the assailed might otherwise have had, he
   thereby supplements the efforts of the perpetrator, and he is
   guilty ... notwithstanding it may he found that in all human
   probability the chance would not have been availed of, and death
   would have resulted anyway.


Applying these concepts to the facts of Tally, Judge Tally is guilty of aiding murder--even though the warning telegram would not have saved Ross even if that telegram had been delivered to Ross--so long as Tally's prevention of the delivery of the warning telegram raised the Skeltons' chances of success in their killing of Ross.

This is chance-raising (or risk-based) liability, pure and simple. See this clearly, and two fundamental features of the actus reus of accomplice liability at common law become very anomalous. The first of these is the firm common law requirement that there be a guilty principal before there can be accomplice liability. (111) At common law, if the Skeltons had not killed Ross, then Tally could not be an accomplice to murder, no matter how much he raised the chances of Ross being killed. Nor, at common law, could Tally be guilty as an accomplice to attempted murder by such chance-raising if the Skeltons themselves were not guilty of such a crime. The common law mantra is that accomplice liability is essentially derivative, that is, dependant on there being someone else who is liable as a principal.

This is a very odd doctrine. To see the oddity, transpose such a requirement onto the liability of another class of culpable riskers, namely, attempters. We would then have the doctrine that, to be guilty of attempted murder, the defendant must not only do an act--substantially risking death while intending to cause that death by that act--but, in addition, the victim risked must die. The victim must not die by the act of the accused, for that would be cause-based murder liability. Rather, the victim must die gratuitously--that is, by means unconnected to the accused's attempt. This would be a ridiculous requirement for attempts, would it not? Attempters are blameworthy because they culpably risk. Whether the victim dies--by lightning, another assassin's bullet, old age, etc.--is completely irrelevant to such risk-based blameworthiness.

The Model Penal Code sees this point with admirable clarity. Under section 5.01(3), that Code would make Judge Tally liable for his purposeful, chance-raising aid even if the Skeltons fail to kill Ross, fail to attempt to kill Ross, or fail to be guilty of any crime whatsoever. (112) To put some words in the mouths of the drafters of that Code: culpable risking is blameworthy in its own right; such blameworthiness does not depend on the fortuitous fact that the harm risked happens to occur or not. The liability is imposed for raising the chances of the harm, not for causing it, failing to prevent it, or allowing it to occur, and especially not for the occurrence of such harm in a manner unconnected in any of these ways with the defendant's act.

The second common law requirement pertinent here has to do with the details of how the accomplice's acts must do their chance-raising if they are to amount to aiding and abetting. Consider, by way of example, the common law rule about what used to be called "principals in the second degree," namely, those present at the crime with the intent to assist the perpetrator if she needed it. The rule is that the would-be accomplice's intent to aid must be communicated to the perpetrator before one's presence without more can amount to aiding. (113)

The problem is that many cases of presence with uncommunicated intent to aid are cases of culpable chance-raisings. Suppose the would-be accomplice is present with the intent to use his rifle to kill the victim if the principal perpetrator misses with her shot. Suppose the probability of each shooter hitting and killing the victim at her respective distances is fifty percent. The existence of a back-up mechanism--the would-be accomplice---increases the likelihood of the victim's death to seventy-five percent, irrespective of any knowledge by the principal that there is such a back-up mechanism.

It is easy to see what tempts courts to adopt the common law rule here. If the accomplice doesn't communicate his intent to aid to the principal, then the principal's decision to shoot cannot have been influenced by the presence of the accomplice. Yet this is to require causal contribution of the accomplice to the result, even if indirectly (viz., the presence of the accomplice causally contributed to the principal's decision to shoot, which itself causally contributed to the death of the victim). Yet courts cannot consistently say both that causal contribution is necessary to be an accomplice and that something else (such as culpable chance-raising) is sufficient by itself. Moreover, and passing by this inconsistency, culpable chance-raising/s an independent desert basis, and criminal liability (either as an accomplice or as a principal) should reflect this moral fact.

The Model Penal Code again sees all of this with admirable clarity. That Code's section 2.06(3) (a) (ii) makes attempting to aid a form of accomplice liability. As the official commentary notes, "Where complicity is based upon agreement or solicitation, one does not ask for evidence that they were actually operative psychologically on the person who committed the offense; there ought to be no difference in the case of aid." (114) Thus, culpable chance-raising is sufficient aid without also (and inconsistently) requiring causal contribution.

This second common law requirement for the actus reus of complicity is thus as anomalous as the first. The first requires the harm assisted to have actually occurred; the second requires that the assistance causally contribute to that harm. Neither requirement is consistent with the common law's own insight that culpable chance-raising deserves punishment on its own.

D. Necessary to Chance Accomplices?

If we attend carefully to the language earlier quoted from the Alabama Supreme Court's Tally opinion, we may think we see a somewhat distinct desert basis for accomplice liability. Note that the court does not talk in terms of Tally's act raising the chances of the Skeltons killing Ross. Rather, the court goes counterfactual on us: if but for Tally's act Ross would have had a "single chance of life," (115) then Tally has done the actus reus of aiding. Equivalently, if but for Tally's act Ross would have had a single chance less of death, then Tally has done the actus reus of aiding. We seem to be looking, not for chance-raisings in the actual world done by the presence of Tally's act, but for chance-lowerings in a possible world where Tally's act is absent.

This might seem to be a fourth sort of desert basis, one that in essence combines counterfactual with probabilistic notions. Thus, one might think some action C is necessary, not to the existence of some state of affairs E, but rather, to E having the probability that it possessed just after C occurred. But for C, in other words, there may well have been an E, but the chances of E with no C would have been lower than in fact they were with C occurring. The absence of C is thus chance-lowering for E in some close possible world where C does not occur.

This could be an independent desert basis only if such counterfactual probabilities differ from causation, counterfactual dependence, and the kind of raising of the conditional probabilities I called chance-raising. Consider each of these in turn, beginning with causation. There is a well-known theory of causation set forth by David Lewis that identifies causation with counterfactual probabilities. (116) According to the probabilistic counterfactual theory of causation, "C causes E" means C is necessary to E having as high a chance of occurring as it does. (117) And, as before, a weaker form of the theory would make causation extensionally equivalent to probabilistic counterfactuals, even if the two relations are not identical: C causes E if and only if but for C, the P(E) would be lower than it is.

It hopefully is obvious that the probabilistic counterfactual theory of causation will inherit the problems of both the counterfactual theory of causation and the probabilistic theory of causation. I will not in any case rehearse the arguments, but assume causation is neither identical to, nor extensionally equivalent with, probabilistic counterfactuals.

The difference between probabilistic counterfactuals and the kind of counterfactual dependence earlier discussed should be apparent. Counterfactual dependence made C necessary for E; probabilistic counterfactuals only make C necessary for E's likelihood of occurring, not E's existence.

The difference with conditional probabilities is much more subtle. David Lewis put the difference this way: "My analysis is in terms of counterfactual conditionals about probability; not in terms of conditional probabilities." (118) What I gather Lewis means is that the conditional probability theorist looks for a conditional relation in this world, namely, the probability of E given C. The probabilistic counterfactualist looks for an unconditional probability of E in an only possible world, one exactly like this world save the absence of C.

This distinction does indeed make for some subtle differences between the two approaches to probability. One difference, for example, is the differential ability of the two theories to handle deterministic cases, where the probability of C occurring is one and so the probability of its not occurring is zero. Conditional probability theorists cannot define a raising of probability in deterministic cases (because the probability of E given no C is undefined and so cannot be less than the probability of E given C); but the probabilistic counterfactualist can define the chance of E occurring in a possible world where C is absent. (119)

Yet this is an in-house dispute about how best to conceptualize probability and risk. What is needed for there to be a distinct desert basis here is something distinct from risk, not different analyses of what risk is. The only difference is that for the counterfactualist probability theorist, the lowering of probability by the absence of C is tested in a possible world where all features that are the same as the actual world are held fixed, while for the conditional probability theorist only causally relevant factors are being held fixed. (120) That makes a difference in how we should conceptualize probability and risk, but it does not make for a new kind of desert basis.

It might seem that there is more of a difference here. It might look like there should be as much difference here as there is between sufficiency and necessity for a determinist. In the deterministic view of the world, if C is sufficient for E, that does not mean that C is also necessary for E. Analogously, it may seem, if C is sufficient to raise the chances of E, that does not mean that C is also necessary for E to have the chances of occurring that it has. Yet this analogy is an illusion. According to the conditional probability theorist, when C raises the chances of E, P(E/C) > P(E/~C). If we were to translate the probabilistic counterfactual theory into the language of conditional probabilities, the necessity of C to E's chances would be given as P(E/~C) < P(E/C). Such chance-lowerings (by C's absence) is not different than what the conditional probability theorists asserts about chance-raisings. Chance-lowerings of E in possible worlds where C is absent are not different from chance-raisings of E in this world (where E is present).

E. Subjectively Culpable Accomplices

There is a class of persons traditionally held liable as accomplices that we have not yet accounted for. These are the individuals who neither causally contribute to some legally prohibited result, nor commit their acts or omissions necessary to that result occurring. Further, their acts do not elevate the likelihood of the harm occurring. Still, they are held as accomplices, and rightly so.

I refer to individuals who, for example, seek to shout encouragement to some principal but the latter either doesn't hear them, mishears them so that she thinks she is being discouraged from her crime, or discounts those who encourage her as untrustworthy fools to whom she pays absolutely no attention; who replace a perfectly reliable getaway car with one that is, unbeknownst to them, about to break down; who stand lookout for a bank robbery but get the time of the planned robbery wrong, so that they do their "looking out" after the robbery has already occurred; (121) who solicit a strangely silent hit man to do a murder for them by various inducements offered at a bar, when the hit man, unbeknownst to them, has already died of an overdose of heroin.

For some of these the common law would find there to be no aiding and, therefore, no accomplice liability, whereas for others, rather inconsistently, there would be liability. The Model Penal Code would hold all of them for attempting to aid the commission of a crime. (122) When that crime was committed, the liability of the secondary party would be for complicity; where there is no crime done by a principal, the liability would be for an attempt. (123) Since the Code generally punishes attempts the same as completed crimes, the differing basis of liability would make no difference to punishment.

The Code here articulates a clear, consistent vision of what is truly a fourth ground of responsibility. Since on this desert basis there is no wrongdoing by the defendant, only culpability, I shall call this the subjectively culpable basis for holding accomplices and others blameworthy and liable.

One can see this desert basis most easily by leaving accomplice liability temporarily, and focusing on attempts. The common law of attempts does not belong here. Its vision is of a different sort, the kind of chance-raising risk imposition we saw before. This comes out clearly in the common law's "dangerous proximity" requirement for the actus reus of attempts. (124) It also is evidenced by the common law version of the defense of "legal impossibility," according to which objectively nonrisking behaviors are not regarded as attempts despite the actors' perceptions that they are likely to bring about some legally prohibited state of affairs. (125)

The Model Penal Code's version of attempt liability is strikingly different. It makes no bones about jettisoning any objective risk component; instead, the subjective perceptions of the actor are alone relevant. The desert determiner for a Model Penal Code attempt is an executed intention. Attempted murder, for example, is simply the doing of some act motivated by an intent to kill. Thus, the Code's actus reus requirements are only that some step be taken strongly corroborative of the actor's intention; the step must be "substantial," but this only means, "strongly corroborative." (126) Such substantiality thus serves a purely evidentiary function, not a function of measuring degree of objective risk. Likewise, the Code explicitly jettisons the common law's version of legal impossibility. Sticking pins in a voodoo doll is attempted murder if the actor believes such action will kill the person the doll represents. Such extreme lack of any objective risk is to be taken into account, if at all, only as a matter of a court's sentencing discretion. (127)

The moral desert basis underlying this version of attempt law is even more severely subjective than the Code recognizes. What is deserving of punishment here is a trying. Notice there is no particular significance for this vision (as opposed to the objective risk vision) on how far the actor proceeds in her trying. That she tries--that she not only has an intention, but executes it--is all that matters. Such trying is a purely mental action. (128) When a defendant intends to kill V and decides here and now to do it by pulling the trigger of the gun in her hand, which results in a willing of the finger movement needed, she has tried to kill V. This is true even if V receives immediate and exceptional medical treatment so that the normally mortal wound does not kill him, V receives a nonmortal wound, V is missed by the bullet, the gun misfires because the bullet is a dud, the trigger is stuck and cannot be pulled, someone holds defendant's finger so she cannot move it, a sudden paralyses besets the defendant so that no nerve signals go to her finger from her brain, etc. So long as the defendant willed an act that she thought would cause death in the service of an intention to cause that death, she tried to kill.

This purely subjective desert basis is also broader than even the Model Penal Code recognizes in a second dimension, namely, in that it extends to the culpability of belief and negligence as well as to the culpability of intention. Take belief first. Suppose the would-be shooter hypothesized above does not intend to kill V. Rather, she intends to shoot out a light, but she believes (a) that she almost certainly will hit V (who is standing behind the light) if she hits the light; or (b) that there is a substantial risk that she will hit V if she shoots. (129) If the defendant then wills the movement of her finger on the trigger, she has again tried to do an act she believes would cause (or seriously risk) the death of V. Such mental acts of trying, accompanied by one or the other of these predictive beliefs, also make the actor culpable and deserving of blame on that basis alone.

Now take negligence. Suppose the would-be shooter in the above hypothetical believes there is a person in close proximity to her target, but she lacks any predictive belief about that person getting hit if she shoots at the target. She again wills the movement of her finger on the trigger, but now the willing is unaccompanied by either an intention to kill or any predictive belief that she will or might kill. Suppose further that, contrary to the shooter's belief, there is no one anywhere in the vicinity of his target. From the epistemic vantage point of the defendant (which includes her information base), what she tried to do was unreasonable--that is, negligent. She should have inferred there was a risk even though she in fact did not. This remains true, even though from our epistemic vantage point--with our knowledge that no one could be hit--her trying to shoot at the target is perfectly reasonable. I take the would-be shooter even here to be culpable, and therefore blameworthy, on this exclusive culpability basis for blame.

It is interesting how this ground for blame presses our distinction between what Michael Zimmerman calls "hypological" judgments (judgments of responsibility for harms real or imagined) and "aretaic" judgments (judgments of responsibility for character). (130) For those of us who distinguish these two kinds of judgments and refuse to relegate one to being a mere proxy for the other, the puzzle arises because a usual ingredient of hypological judgments is here missing. Usually when we judge someone to be responsible for some harm, real or imagined, that hypological judgment is based in part on a deontic judgment. Deontic judgments are those judging actions as right or wrong. With this purely subjective desert basis, there is no wrongdoing. The actor neither causes the harm, nor does she allow it or fail to prevent it when she had the ability to do so; nor is there even wrongdoing in the anemic sense of objective risk imposition. The only wrongdoing is in the actor's mind--as she sees the world, she is doing something that morality regards as wrong and that the law forbids. Even so, this remains a hypological judgment of responsibility and therefore blameworthiness. Such judgments are not general as are aretaic judgments of character. Rather, such judgments are about what an actor did at one razor's edge of time: she tried to do something with the intent or beliefs that make her culpable.

This is enough for some lesser form of blameworthiness. Even for those of generally excellent character, such culpable tryings are an appropriate basis for blame. Our beliefs create a possible world, one that differs from the actual world by the degree to which our beliefs are false. In that possible world created by her own beliefs, the would-be shooter is doing an act that she intends to cause (or believes will or might cause, or should believe will or might cause, in light of her other beliefs) a bad result. She is just as culpable (131) as one whose identical acts take place in a possible world much closer to the actual world and merits blame on that basis.

We can now return to the accomplice liability cases with which we began this section. On the common law view of inchoate liability, limited as it is to objective risk imposition, there should be no accomplice liability in these cases. Yet on the subjective view of inchoate liability championed by the Model Penal Code, there would be liability in all such cases. On such a view, it is arbitrary whether one calls it accomplice liability or liability as a principal attempter, for this turns on the irrelevant issue of whether there is some second party who is a guilty principal.

F. Vicarious Accomplices

Perhaps surprisingly, even with four distinct desert bases in play, we have not accounted for all cases in which courts have found accomplice liability. For there are those who have been held to be accomplices who have not causally contributed to some result, not allowed or failed to prevent a result that they had the ability to prevent, not risked the result, and not tried to do some action that in their own mind would have helped to cause some result. Yet in each case they have been blamed and held liable because someone else did one of these four things.

This is vicarious liability. Liability is vicarious for some defendant D when (1) someone else has caused, risked, tried, allowed, or failed to prevent some harm; and (2) D stands in some relation R to that person, so long as R does not make D himself stand in one of the four relations to that harm that would make him blameworthy. A threatening relation between D and the principal, for example, will not do, because then D causally contributed to the harm. But if R is mere group membership, then D's liability for other group members' actions is vicarious.

The well-known Pinkerton case illustrates the extension of accomplice liability on a purely vicarious basis. (132) Walter and Daniel Pinkerton were brothers who lived on the same farm. They were whiskey runners, transporting and selling bootleg liquor. Each was convicted of his own separate acts of possession, transportation, and sales of the bootleg whiskey. In addition, they were both convicted of conspiracy to do these things--that is, an agreement between them to cooperate in these ventures. Finally, Daniel was convicted of Waiter's transportation, possession, and sales of bootleg whiskey committed while Daniel was in prison on unrelated charges.

The Pinkerton majority held that the agreement to cooperate created an agency relationship between the brothers, much like a legal partnership. And thus, "so long as the partnership in crime continues, the partners act for each other in carrying it forward." (133) As the dissent recognized, this basis for liability "is a vicarious criminal responsibility as broad as, or broader than, the vicarious civil liability of a partner for acts done by a copartner in the course of the firm's business." (134)

Vicarious liability is a derivative liability doctrine. It piggybacks on some true basis for responsibility, such as the four we have examined. It is not, nor does it pretend to be, some distinct, fifth basis for blame. As such, it has no place in any punishment scheme linking legal liability to moral blameworthiness. This is as true of accomplice liability as of any other form of criminal liability.

Vicarious "moral" blame is not unknown. It is said that Ghengis Khan issued standing orders to execute all ten members of any ten-membered cohort in his army should any one member desert in time of battle. Some would hold present-day Germans responsible for the evils of Hitler and other Nazis simply on the basis of membership in the class "Germans." Such vicarious blamings may make for efficient armies or compensation schemes, but surely, on reflection, there can be no doubt that blame is undeserved in such cases.

In a rightly conceived penal code there should be no vicarious accomplices. The Model Penal Code's rejection of the Pinkerton doctrine, now joined by a majority of American jurisdictions, is a step in that direction. (135)

IV. SOME CONCLUSIONS ABOUT THE FOUR KINDS OF ACCOMPLICES

There are four kinds of accomplices in a rightly conceived penal code, corresponding to four kinds of general desert bases: causal contributors, omitters and allowers (whether partial or full), riskers, and culpable tryers. There are very large differences in degrees of blameworthiness within each of these classes, due in part to the degree of causal contribution (of some bad result) the accomplice makes, the degree of necessity (to that result) possessed by an allower's act or an omitter's omission, the magnitude of the risk imposed by the accomplice, or the degree of culpability of the tryer. Even so, as a general matter, the degree of blameworthiness descends as one goes down this scale. That is, on average and holding the mental states of culpability constant, causing is worse than allowing, which is worse than omitting; causing is also worse than risking and worse than trying. It thus matters what kind of an accomplice one is--matters in the sense of its being relevant to the amount of punishment one deserves.

Given the four different desert bases grounding accomplice liability, it should be unsurprising that the actus reus of such liability-aiding--has proved so elusive. The accomplice liability cases and doctrines are not like the trunk, tusks, tail, feet, etc., of an unseen elephant. They are rather like the trunk, hooves, wings, stripes, spots, claws, and talons of a rather difficult-to-picture animal. To search for some one thing that is "aiding" is to search for a will-o'-the-wisp, because there are four distinct items sufficient for aiding under rightly conceived accomplice liability doctrines.

This might suggest a disjunctive definition of the actus reus of accomplice liability, a disjunction with four distinct desert bases for this liability. Remembering the different degrees of blameworthiness attaching to these different desert bases, one would not want to define aiding as any one of the four and leave it at that. Rather, one needs to separate degrees of complicity, reserving the highest degree of punishment for causal accomplices, less for double-preventionists, still less for omitters, etc. (136) The general statutory equation of accomplices with principals (in terms of deserved punishment) would also have to go, save perhaps for causal accomplices.

Yet any such reforms would be a mere tidying up of complicity, when what we have seen shows it to be in much too serious trouble for mere rearrangement of deck chairs. What we have seen should be sufficient to sink the good ship Complicity, not rearrange its furnishings. Notice that the four bases for accomplice liability are not unique to that form of liability. These four are the desert bases for moral blameworthiness generally. As the disjunctive proposal reveals, there is no unique desert basis for accomplice liability. Aiding another to cause some bad result is not an independent desert basis. It is a mere stand-in for one of the four general bases on which we are rightly blamed.

Another way of saying the same thing is to repeat what was said earlier: the existence of some second party who more directly causes some bad result is irrelevant when judging the blameworthiness of accomplices. Their blameworthiness is established by their own causal contribution to the result, their own allowing or failing to prevent that result, their own risking of that result, or their own culpable tryings. Their blameworthiness is just like that of any principal in this regard. They are in no relevant respect different, and thus there is no need for a separate form or theory of liability for accomplices.

The most that might be said in favor of retaining an accomplice form of liability is this: those we now call accomplices, in general and on average, present lesser degrees of blameworthiness on each of the four bases for blame we have examined. More precisely, (1) causal accomplices, on average, are lesser causal contributors to some harm than those we now distinguish as principals; (2) necessary accomplices, on average, are less necessary to the harms they allow or fail to prevent than are those allowers or omitters we now call principals (where "less necessary" means the relevant counterfactual holds true only in possible worlds closer to the actual world); (137) (3) chance-raising accomplices, on average, raise the probability of some harm less than those riskers we now call principals; and (4) subjectively culpable accomplices, on average, are less culpable than those attempters we now call principals (because the former typically try to make lesser causal contributions than do the latter). But these are matters of degree and rules of thumb. One would need some good reason to magnify and rigidify these differences into two distinct theories of liability--that of a principal versus that of an accomplice.

And if one believes that there are such good reasons and thus goes this route, then one still has more tidying up to do than I recommended a moment ago. Once one sees that the existence of a "guilty principal" is merely a proxy for diminished causal contribution by the accomplice, diminished risk, etc., then another equally good proxy presents itself in the form of "guilty nature." Recall that abnormal natural events metaphorically "break" causal chains under standard intervening cause doctrines just as do guilty human actors. This is because such large storms, strange coincidences, etc., are as good at showing some defendant's causal contribution (risk, etc.) to be less than normal as is the existence of some guilty principal. So if accomplice liability is to be retained for cases involving the latter, such liability in all consistency should be created for the former. That is, aiding nature should be a basis for complicity too. (138) If it were up to me, however, I would dump them both in favor of having no form of accomplice liability.

It might seem as if my skepticism about the principal/accomplice distinction should infect the completed/inchoate crime distinction as well. After all, inchoate liability is ultimately predicated on either objective chance-raisings or culpable tryings, and liability as a principal lumps harm-causing with harm-allowing and both together with failure to prevent harm. Yet notice that this is not the "Chinese-cut chicken" that is accomplice liability. Accomplice liability as it is presently constituted not only encompasses four (or possibly five) distinct desert bases, it also shares these four distinct desert bases with principal liability. Like the Chinese technique of segmenting a chicken, the law's blade here hits not a single joint in nature.

Principal liability for completed crimes--once it includes accomplices-does encompass all three desert bases of causing, allowing, and failing to prevent. It may well be a good idea to tidy this up as well--separating causing from failing to prevent, and both from full and partial allowings. Yet, whether we tidy up or not within principal liability for completed crimes, cases where an actor either causes a harm, or on whose acts the harm counterfactually depends, are neatly separated from cases where no such harm occurs (or, if it does occur, it is unconnected to the actor in either a causal or counterfactual way)--i.e., cases of inchoate liability. Inchoate liability is thus worth distinguishing as a separate form of liability from that of a principal for a completed crime. The relationship(s) between an actor's act and the type of harm the law cares about is different for inchoate liability. It is noncausal, noncounterfactual; rather, it is either that the type of harm was objectively risked or subjectively intended.

We might also tidy up, within the category of inchoate liability, just as we could within the category of completed crimes. The complexity within the current law on inchoate liability stems from our uncertainty as to which of two desert bases should govern. On the "nearmiss" view of it, inchoate liability only attaches when there is a substantial, objective risk of harm. On the culpability view of it, such liability attaches when there is a trying by the actor (modified only by the evidentiary demand of corroboration of that trying by real world events). This is a conflict in doctrine due to indecision about the most appropriate desert basis. A simple tidying up within inchoate liability would be either to resolve the indecision in favor of one desert basis or the other, or to embrace both disjunctively as sufficient for such liability.

In any case, tidied up or not, we have good reason to retain the completed crime/inchoate crime distinction. The bottom-line conclusion of this Article is that we have no equally good reason to retain the accomplice/principal distinction. It only remains to say how eliminating the latter distinction would affect the shape of criminal liability.

Eliminating the distinction between principals and accomplices would not be academic or theoretical in the pejorative sense of making no doctrinal difference. Such elimination would not leave liability where it presently rests, only placing it under one label rather than another. Rather, such elimination would both decrease and increase liability from where it presently rests. The decrease in liability would come in cases like Wilcox, where the alleged accomplice is held for the crime of another despite de minimis aid being given. Defendants like Wilcox should be held to the completed crime only if the standard bases for such liability are present, viz., they either substantially caused the criminal result, or that result counterfactually depended upon their action. If defendants like Wilcox would not pass these tests for principal liability, they should not pass them as supposed "accomplices." That someone else completed the crime is irrelevant to the desert of such defendants who should be judged by what they and they alone contributed to the result. If their contribution is insufficient for liability for the completed crime, then they should be liable only for the lesser punishments of inchoate crimes, just like those now called "principals."

An increase in liability would occur (once the principal/accomplice distinction is eliminated) in cases where a more culpable accomplice aids a less culpable principal. Under present law, accomplice liability is derivative of that of the principal, so that the accomplice can only be convicted of the crime of which the principal can be convicted. A much-discussed example is that of Isabel Richards, who hired two men to beat up her husband, with the intent they beat him severely; they in fact only hit him without causing any grievous bodily harm. (139) Since they only intended to beat him in this moderate way, they were convicted only of misdemeanor assault. Isabel was convicted of assault with intent to cause grievous bodily harm, a crime for which a moderate beating was sufficient actus reus but which is a much more serious crime because of the more culpable mens rea. The conviction was reversed because of the derivative nature of Richards' accomplice liability. (140) Eliminating the accomplice/principal distinction would eliminate this limitation on Richards' liability; she would be judged by her own mens rea, which is as it should be.

Putting these last two points together allows us to see that eliminating the principal/accomplice distinction has very practical payoffs, not just theoretical ones. Eliminating the distinction would allow each of us to be judged as we should be judged: by what we individually contributed to some legally prohibited result, and by the state of mind we had when we did what we did. Standing in the shoes of someone else and being blamed for what she did, and for the culpability with which she did it, violates proportionality between desert and punishment.

(1) See, e.g., WAYNE R. LAFAVE & AUSTIN W. SCOTT, JR., CRIMINAL LAW 579-80 (2d ed. 1986).

(2) MODEL PENAL CODE [section] 2.02 (1985).

(3) This paper is the tenth in a series of papers on causation, both its metaphysics and its relevance to responsibility. The predecessor papers are: Heidi M. Hurd & Michael S. Moore, Negligence in the Air, 3 THEORETICAL INQUIRIES L. 333 (2002); Michael S. Moore, Causal Relata, in 13 ANNUAL REVIEW OF LAW AND ETHICS: PHILOSOPHIA PRACTICA UNIVERSALIS 589 (B. Sharon Byrd & Jan c. Joerden eds., 2005); Michael S. Moore, Causation, in 1 ENCYCLOPEDIA OF CRIME & JUSTICE 150 (Joshua Dressler ed., 2d ed. 2002); Michael S. Moore, Causation and Responsibility, SOC. PHIL. & POL'Y, Summer 1999, at 1; Michael S. Moore, The Destruction of the World Trade Center and the Law on Event-Identity, in AGENCY AND ACTION 259 (John Hyman & Helen Steward eds., 2004); Michael Moore, For What Must We Pay ? Causation and Counterfactual Baselines, 40 SAN DIEGO L. REV. 1181 (2003) [hereinafter Moore, Causation and Counterfactual Baselines]; Michael Moore, The Independent Moral Significance of Wrongdoing, 5 J. CONTEMP. LEGAL ISSUES 237 (1994); Michael S. Moore, The Metaphysics of Causal Intervention, 88 CAL. L. REV. 827 (2000); and Michael S. Moore, Patrolling the Borders of Consequentialist Justifications: The Scope of Agent-Relative Restrictions, 27 LAW & PHIL. (forthcoming 2008) [hereinafter Moore, Patrolling the Borders]. These papers together form the bulk of my forthcoming book, Causation and Responsibility, to be published by Oxford University Press in 2008.

(4) E.g., Rollin M. Perkins, Parties to Crime, 89 U. PA. L. REV. 581, 584 (1941).

(5) I argue for this at some length in MICHAEL S. MOORE, ACT AND CRIME: THE PHILOSOPHY OF ACTION AND ITS IMPLICATIONS FOR CRIMINAL LAW 189-238 (1993).

(6) See, e.g., Glanville Williams, The Problem of Reckless Attempts, 1983 CRIM. L. REV. 365, 366 (discussing the absence of causal requirements in "conduct crimes" and providing the example of rape). For other examples, see the discussion and citations in MOORE, supra note 5, at 209-10, 214-25.

(7) See generally MOORE, supra note 5, at 60-77.

(8) See, e.g., Pinkerton v. United States, 328 U.S. 640 (1946).

(9) Joshua Dressier considers but rejects the vicarious basis for accomplice liability. Joshua Dressier, Reassessing the Theoretical Underpinnings of Accomplice Liability: New Solutions to an Old Problem, 37 HASTINGS L.J. 91, 109-11 (1985).

(10) See, e.g., Richard Buxton, Complicity and the Law Commission, 1973 CRIM. L. REV. 223, 223-30; Richard Buxton, Complicity in the Criminal Code, 85 L.Q. REV. 252, 265-73 (1969); J.R. Spencer, Trying To Help Another Person Commit a Crime, in CRIMINAL LAW: ESSAYS IN HONOUR OF J.C. SMITH 148, 148 (Peter Smith ed., 1987); Daniel Yeager, Helping, Doing, and the Grammar of Complicity, 15 CRIM. JUST. ETHICS 25, 30-33 (1996). Recently, Christopher Kutz seems to have joined the view that accomplice liability is a form of inchoate liability. See Christopher Kutz, Causeless Complicity, 1 CRIM. L. & PHIL. 289, 301-03 (2007).

(11) See, e.g., K.J.M. SMITH, A MODERN TREATISE ON THE LAW OF CRIMINAL COMPLICITY 124-27 (1991); Paul H. Robinson, Imputed Criminal Liability, 93 YALE L.J. 609, 613-14 (1984); Francis Bowes Sayre, Criminal Responsibility for the Acts of Another, 43 HARV. L. REV. 689, 702 (1930).

(12) The Pinkerton doctrine, 328 U.S. at 646-47, is a federal doctrine, but it is followed in a minority of American states.

(13) See, e.g., People v. Luparello, 231 Cal. Rptr. 832, 846-51 (Cal. Ct. App. 1986).

(14) This is the suggestion of Smith, Robinson, and Sayre, supra note 11.

(15) This view is held by Dressier, supra note 9, at 102-03, and CHRISTOPHER KUTZ, COMPLICITY: ETHICS AND LAW FOR A COLLECTIVE AGE 220-27 (2000).

(16) 15 So. 722 (Ala. 1894).

(17) Id. at 738.

(18) KUTZ, supra note 15, at 217. Kutz has recently come to a somewhat more nuanced view of causation. See Kutz, supra note 10, at 290 ("Causation may be present, I argue, and causal relations feature in the justification of accomplice liability overall, but causation is not necessary to complicity.").

(19) KUTZ, supra note 15, at 215; see also Dressier, supra note 9, at 99; Yeager, supra note 10, at 29.

(20) Moore, Causation and Counterfactual Baselines, supra note 3.

(21) The substitivity of identicals is often called "Leibniz's Law." For an introduction, see MICHAEL MOORE, PLACING BLAME: A GENERAL THEORY OF THE CRIMINAL LAW 372 (1997).

(22) For the qualifications, see Moore, Causal Relata, supra note 3, at 629-34.

(23) I have made this argument in Moore, Causation and Counterfactual Baselines, supra note 3, at 1214-15.

(24) At least this is true so long as one does not stipulate (as did David Lewis and his followers) that counterfactuals that "backtrack" through time are "deviant." For a hypothetical demonstrating this, see id. at 1227-37.

(25) THE SOUND OF MUSIC (20th Century Fox 1965).

(26) Counterfactualists argue this point mostly by a stipulated narrowing of the class of counterfactuals eligible to serve as sufficient conditions for causation. Thus, one response to (a) is the denial by counterfactualists that counterfactuals relating omissions are like ordinary counterfactuals relating events because the former but not the latter are about types, not particulars; to (b), the response is to stipulate that counterfactuals must be between "distinct events"; and to (c) and (d), the main response is to treat "backtracking" counterfactuals as deviant and not included in the class of counterfactuals relevant to causation. For a discussion of these responses, see Moore, Causation and Counterfactual Baselines, supra note 3, at 1238-57.

(27) See id. (discussing this attention in detail).

(28) See id. at 1213-37 (tracing these attempts).

(29) The Model Penal Code's failure to cover cases of concurrent causation is ac knowledged at MODEL PENAL CODE [section] 2.03 cmt. 2 (1985).

(30) See generally Wex S. Malone, Ruminations on Cause-in-Fact, 9 STAN. L. REV. 60 (1956). For the critical legal studies version of this view, see Mark Kelman, The Necessary Myth of Objective Causation Judgments in Liberal Political Theory, 63 CHI.-KENT L. REV. 579 (1987). My early commentary on Kelman appeared in Michael S. Moore, Thomson's Preliminaries About Causation and Rights, 63 CHI.-KENT L. REV. 497, 501-06 (1987).

(31) Both Restatements eschew but-for causation for the "substantial factor" test. RE STATEMENT (SECOND) OF TORTS [section] 431 (1965); RESTATEMENT OF TORTS [section] 431 (1934).

(32) I explore the case law establishing these conditions in Moore, The Metaphysics of Causal Intervention, supra note 3, at 832-52.

(33) See id. at 839-46.

(34) Sanford H. Kadish, Complicity, Cause and Blame: A Study in the Interpretation of Doctrine, 73 CAL. L. REV. 323 (1985).

(35) BLONDIE, One Way or Another, on PARALLEL LINES (Chrysalis Records 1978).

(36) I argue for these propositions in Moore, The Independent Moral Significance of Wrongdoing, supra note 3, at 258-80.

(37) The fit of libertarian metaphysics to the legal doctrines of intervening causation is in fact a rough one, as I examine in Moore, The Metaphysics of Causal Intervention, supra note 3, at 869-71. This is one argument more than I need here, so I shall ignore it.

(38) Kadish, supra note 34, at 360.

(39) Id. at 334-36.

(40) H.L.A. HART & TONY HONORE, CAUSATION IN THE LAW 28-32 (2d ed. 1985).

(41) The nineteenth century's "last wrongdoer" rule is an example of such an avowedly fictionalist approach to intervening causation. See Lawrence S. Eldredge, Culpable Intervention as Superseding Cause, 86 U. PA. L. REV. 121, 124 (1937). On this rule, the law only need punish the (temporally) last wrongdoer in some complicated chain of causation leading to some legally prohibited result. Such "last wrongdoer" thus cuts off all earlier wrongdoers from liability for that harm. Id. The problems with the rule have been well articulated for the last one hundred years, starting with Holmes: (1) the cases don't actually fit the doctrine, especially in cases of merely negligent or reckless wrongdoers where the last wrongdoer often does not insulate others up the chain of causation from liability; (2) there is no policy that can justify the significance given such temporal ordering of wrongdoing. Id. at 124-25; see also Clifford v. Atl. Cotton Mills, 15 N.E. 84, 87 (Mass. 1888) (Holmes, J.) (describing the "last wrongdoer" rule as no more than a "general tendency" of courts).

(42) See, e.g., CAL. PENAL CODE [section] 664 (West 1999).

(43) HART & HONORE, supra note 40, at 6. Sanford Kadish and John Gardner have also explicitly endorsed this kind of causal dualism. Kadish, supra note 34, at 334; John Gardner, Complicity and Causality, 1 CRIM. L. & PHIL. 127, 134 (2007).

(44) HART & HONORE, supra note 40, at 186.

(45) Id. at 388.

(46) Id. at 195.

(47) Antony Flew, Psychiatry, Law and Responsibility, 35 PHIL. Q. 425, 430 (1985) (reviewing MICHAEL S. MOORE, LAW AND PSYCHIATRY: RETHINKING THE RELATIONSHIP (1984)).

(48) This is not to be confused with statements of comparative causation, as in, "Factor X was sixty percent the cause, factor Y forty percent the cause of harm Z." Some of us think such scalar statements of degrees of causal contribution make perfectly good sense. See Moore, The Metaphysics of Causal Intervention, supra note 3, at 874 & n.229.

(49) For real world examples (Sheldon Glueck and Norval Morris) and an analysis of why this is truly metaphysical gibberish, see Michael S. Moore, Causation and the Excuses, 73 CAL. L. REV. 1091, 1114-19 (1985).

(50) DONALD DAVIDSON, ESSAYS ON ACTIONS AND EVENTS 110-11 (1980).

(51) Id. at 111.

(52) See MOORE, supra note 5, at 230-31.

(53) Thus, for example, the Model Penal Code eschews the biblical "thou shalt not kill" for one shall not "cause[] the death of another human being." MODEL PENAL CODE [section] 210.1 (1) (1985).

(54) Dir. of Pub. Prosecutions for N. Ir. v. Lynch, [1975] A.C. 653, 659 (H.L.) (appeal taken from NICA (Crim)).

(55) The House of Lords recognized as much by refusing the duress defense on roughly these facts (though involving stabbing instead of shooting) in Abbott v. The Queen, (1976) 3 All. E.R. 140, 142, 148 (P.C.) (appeal taken from Trin. & Tobago Ct. App.).

(56) This is argued for more extensively in Moore, Patrolling the Borders, supra note 3 (manuscript at 36-37).

(57) See, for example, Kimble v. Mackintosh Hemphill Co., 59 A.2d 68, 71 (Pa. 1948), where the defendant successfully argued that extreme winds were an intervening cause.

(58) The hypothetical slightly alters the facts of Abbott, [1976] 3 All. E.R. at 141.

(59) Kadish, supra note 34, at 372-85. In a recent paper, John Gardner does not see this problem as distinct from the general semantics of causatives just discussed. See Gardner, supra note 43, at 135-36. Gardner's mistake here is due to his bloated notion of "proxyability." Gardner mistakenly assumes all (or most) criminal prohibitions utilizing the causative verbs of English create "nonproxyabile crimes." As we will see shortly, this is not Kadish's notion, which is much narrower and thus much more interesting.

(60) Kadish, supra note 34, at 373.

(61) Id. (footnotes omitted).

(62) The explicit definition in some states, such as North Carolina, provides that only penetration by the defendant's penis constitutes rape. N.C. GEN. STAT. [section][section] 14-27.1-14-27.10 (2005).

(63) Dusenbery v. Commonwealth, 263 S.E.2d 392 (Va. 1980).

(64) Id. at 393.

(65) RESTATEMENT (SECOND) OF TORTS [section] 431 (1965); RESTATEMENT OF TORTS [section] 431 (1934).

(66) This "causation as primitive except for scalarity" approach was adopted by the American Law Institute from Jeremiah Smith, Legal Cause in Actions of Tort (pts. 1-3), 25 HARV. L. REV. 102, 223, 303 (1911-1912). Cause-as-primitive is a respectable view within the contemporary philosophy of science. See generally D.M. ARMSTRONG, WHAT Is A LAW OF NATURE? (1983); MICHAEL TOOLEY, CAUSATION: A REALIST APPROACH (1987).

(67) RESTATEMENT (SECOND) OF TORTS [section] 431; RESTATEMENT OF TORTS [section] 431.

(68) See, e.g., Perkins, supra note 4, at 594.

(69) For a summary, see Kadish, supra note 34, at 369-72.

(70) Id.

(71) Id.

(72) E.g., AGATHA CHRISTIE, MURDER ON THE ORIENT EXPRESS (1960). For a real-life, if scaled-down, version with only two wounders, see People v. Lewis, 57 P. 470 (Cal. 1899).

(73) Wilcox v. Jeffery, (1951) 1 All E.R. 464 (K.B.).

(74) Id. at 466.

(75) Moore, Causation and Counterfactual Baselines, supra note 3, at 1248-49.

(76) State ex. rel. Martin v. Tally, 15 So. 722 (Ala. 1894); see supra notes 16-18 and accompanying text.

(77) See supra text accompanying notes 26-27.

(78) See Cook v. Minneapolis, St. Paul & Sault Ste. Marie Ry. Co., 74 N.W. 561 (Wis. 1898) (concurrent overdetermination); City of Piqua v. Morris, 120 N.E. 300 (Ohio 1918) (asymmetrical overdetermination); Dillon v. Twin State Gas & Elec. Co., 163 A. 111 (N.H. 1932) (preemptive overdetermination).

(79) Moore, Causation and Counterfactual Baselines, supra note 3, at 1264-66.

(80) Compare the differing views discussed in Causation and Counterfactual Baselines, id. at 1266-67, with Moore, Patrolling the Borders, supra note 3.

(81) The discussion that follows summarizes Moore, Causal Relata, supra note 3, at 614-25.

(82) One needn't agree with the example to agree with the point being illustrated. If to your sense of how things are, dyings are just not-survivings, then transpose the example. And if your view is that surviving and dying are both positive events or states of affairs, then (given the contradictory or at least contrary relation between them) your "positive" events are at the same time "negative" events--that is, you seem committed to there being negative events, negative states of affairs, and negative properties.

(83) For more details, see Moore, Causal Relata, supranote 3, at 617-18.

(84) For a description of the rape that occurred in New Bedford, see Lynn S. Chancer, New Bedford, Massachusetts, March 6, 1983-March 22, 1984: The "Before and After" of a Group Rape, 1 GENDER & SOC'Y 239, 244-46 (1987). For a fictionalized account, see THE ACCUSED (Paramount Pictures 1988).

(85) MOORE, supra note 5, at 278 n.42.

(86) Moore, Causation and Responsibility, supra note 3, at 42.

(87) Most of my article Patrolling the Borders, supra note 3, deals with the doing/allowing distinction.

(88) See id. (manuscript at 44).

(89) Most famous is Regina v. Dudley & Stephens, (1884) 14 Q.B.D. 273, where the defendants killed a cabin boy who was about to die of natural causes anyway. I discuss the run of acceleration cases in Moore, Patrolling the Borders, supra note 3 (manuscript at 44-48).

(90) See Moore, Patrolling the Borders, supra note 3 (manuscript at 43-44) for a description of such interrogation techniques and an analysis of whether torture can be justified in the "war on terror."

(91) See id. (manuscript at 43).

(92) WILLIAM SHAKESPEARE, MACBETH act 2, sc. 2.

(93) Ireland v. United Kingdom, 25 Eur. Ct. H.R. (ser. A) at 67 (1976).

(94) Sister Diana Ortiz acknowledged this in her Keynote Address to the conference at the City University of New York Graduate Center, Torture After 9/11: The Legal and Ethical Implications of Torture in the New Age of Terrorism (Oct. 24, 2003).

(95) For those unfamiliar with the notation, P(E/C) means "the probability of E given C," whereas P(E) means "the probability of E."

(96) For an excellent summary of the different interpretations of probability, see DONALD GILLIES, PHILOSOPHICAL THEORIES OF PROBABILITY 1-13 (2000).

(97) An epiphenomenal fork is created when a single cause produces two distinct effects that have no causal relationship between themselves. To return to an earlier example, see supra text accompanying notes 25-26, at time [t.sub.1] I run in the morning with my dog; this causes (at [t.sub.2]) my foot to become sore. My running with my dog also causes (at t0 my dog to get tired. If I only run with my dog, he only gets tired when he runs with me, and my feet only get tired when I run, then my dog being tired at [t.sub.3] raises the probability of my feet being tired at [t.sub.2].

(98) For background information on the idea that causes raise the probability of their effects, see Christopher Hitchcock, Probabilistie Causation, in STANFORD ENCYCLOPEDIA OF PHILOSOPHY (Edward N. Zalta ed., Summer 2007), http://plato.stanford.edu/ archives/sum2007/entries/causation-probalistic/

(99) See id. [section] 6.2 for an explanation of theories that reach a contrary conclusion--that A did not cause V's death.

(100) A "hard way" case is where there are two causal routes possible to some result, one route being easier than another in the sense that success is much more probable with the easier route. For example, I could hit a birdie in golf with a direct shot to the hole; or once in a great while I could make the birdie by hitting a tree in just such a way that the ball ricochets into the hole. Making the birdie the "hard way" is to hit the tree. Hitting the tree admittedly caused me to make the birdie but hitting the tree also lowered the probability of the birdie compared to a direct shot. For a number of such counterexamples, see Wesley C. Salmon, Probabilistic Causality, 61 PAC. PHIL. Q. 50, 68-70 (1980).

(101) The example is discussed in Germund Hesslow, Two Notes on the Probabilistic Approach to Causality, 43 PHIL. SCI. 290, 291 (1976).

(102) See the summary in Hitchcock, supra note 98.

(103) I argue this throughout Moore, Causation and Responsibility, supra note 3.

(104) The leading case on the tort of "lost chance" is Herskovits v. Group Health Cooperative of Puget Sound, 664 P.2d 474 (Wash. 1983). Eric Johnson nicely shows how many putatively cause-based liability cases in criminal law are in reality lost chance cases (because of laxity in the proof deemed sufficient for causation). Eric A. Johnson, Criminal Liability for Loss of a Chance, 91 IOWA L. REV. 59, 69-76 (2005).

(105) E.g., State ex rel. Martin v. Tally, 15 So. 722 (Ala. 1894).

(106) An instance of the latter kind is Hyam v. Director of Public Prosecutions, [1975] A.C. 55, 78-79 (H.L.) (appeal taken from Eng.), in which the court determed that Mrs. Hyam did not intend to kill Mr. Booth, but she did intend to create a serious risk of death (in order to scare Mrs. Booth).

(107) There is considerable debate here. For arguments supporting the idea that risk is not in itself harm, see Heidi M. Hurd, The Deontology of Negligence, 76 B.U.L. REV. 249, 262-65 (1996); Stephen R. Perry, Risk, Harm, and Responsibility, in PHILOSOPHICAL FOUNDATIONS OF TORT LAW 321, 330-39 (David G. Owen ed., 1995). For an exploration of the idea that exposing someone to a risk is itself to harm her, see Claire Finkelstein, Is Risk a Harm?, 151 U. PA. L. REV. 963 (2003).

(108) Moore, The Independent Moral Significance of Wrongdoing, supra note 3, at 267-68.

(109) Tally, 15 So. at 738-39.

(110) Id. at 739.

(111) See, e.g., Perkins, supra note 4, at 618.

(112) See MODEL PENAL CODE [section] 5.01 (3) (1985) (making liable for attempt one who "aid[s] another to commit a crime that would establish his complicity under section 2.06 if the crime were committed by such other person").

(113) See, e.g., Perkins, supra note 4, at 600.

(114) MODEL PENAL CODE [section] 2.06 cmt. 6(c).

(115) Tally, 15 So. at 739.

(116) See 2 DAVID R. LEWIS, Postscripts to "Causation," in PHILOSOPHICAL PAPERS 172 (1986).

(117) Id. at 176.

(118) Id. at 178.

(119) Id.

(120) See Hitchcock, supra note 98, [section] 3.2.

(121) See Larkins v. Police, [1987] 2 N.Z.L.R. 282, 283-86 (H.C.).

(122) MODEL PENAL CODE [section] 2.06(3) (a) (ii).

(123) Id. [section] 5.01(3).

(124) People v. Rizzo, 158 N.E. 888, 889 (N.Y. 1927) (quoting Hyde v. United States, 225 U.S. 347, 388 (1911) (Holmes, J., dissenting)).

(125) Admittedly, it is difficult to discern any one policy intuition guiding the old common law defense of "legal impossibility." The requirement that there be a "near miss"--a significant raising of objective chance--is the best sense I can make of the cases.

(126) MODEL PENAL, CODE [section] 5.01 (2).

(127) See id. [section] 5.05(2).

(128) The "mental action" theorists saw this clearly. For a discussion, see MOORE, supra note 5, at 95-108.

(129) These are roughly the facts of Thacker v. Commonwealth, 114 S.E. 504, 505 (Va. 1922).

(130) Michael Zimmerman, Taking Luck Seriously, 99 J. PHIL. 553, 554-55 (2002). Christopher Kutz plainly feels the tug toward aretaic theories in his inchoate interpretation of complicity. See Kutz, supra note 10.

(131) But not as blameworthy, because she has done nothing wrong. In my lexicon, culpability is but one of two ingredients in overall blameworthiness, the other being wrongdoing. See Moore, The Independent Moral Significance of Wrongdoing, supra note 3, at 237-38.

(132) Pinkerton v. United States, 328 U.S. 640 (1946).

(133) Id. at 646.

(134) Id. at 651 (Rutledge, J., dissenting in part).

(135) MODEL PENAL CODE [section] 2.06(3) & cmt. 6(a) (1985).

(136) Joshua Dressier heads off in this direction, although with only two categories of accomplices, "causal accomplices" and "noncausal accomplices," rather than four. Dressier, supra note 9, at 124-30.

(137) On this notion of degrees of necessity, see Moore, Causation and Counterfactual Baselines, supra note 3, at 1204-13.

(138) For development of this suggestion, see Moore, Causation and Responsibility, supra note 3, at 19, 41-43.

(139) Regina v. Richards, [1974] Q.B. 776.

(140) Id, at 780.

MICHAEL S. MOORE, Walgreen University Chair, Professor of Law, Professor of Philosophy, Professor in the Center for Advanced Studies, Co-Director of the Program in Law and Philosophy, University of Illinois. This paper was given to the faculty in the Program in Criminal Law and Procedure of the University of Illinois, to the Faculty Workshop of the University of Pennsylvania Law School, and to GALA workshops of the Boalt Hall Faculty, University of California, Berkeley. Thanks go to the participants at each of these presentations for their many helpful comments. Special thanks go to my longtime friend and former colleague on the Boalt Hall faculty, Sanford Kadish, whose work on accomplice liability has had far more influence on the present work than the citations below would indicate.
COPYRIGHT 2007 University of Pennsylvania, Law School
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2007 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Moore, Michael S.
Publication:University of Pennsylvania Law Review
Date:Dec 1, 2007
Words:25468
Previous Article:Structuring judicial review of electoral mechanics: explanations and opportunities.
Next Article:Appended post-passage Senate Judiciary Committee report: unlikely "legislative history" for interpreting section 5 of the Reauthorized Voting Rights...
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |