I. INTRODUCTION
Michael Moore's sweeping, insightful, and startlingly learned book Causation and Responsibility Footnote 1 concerns, as the title suggests, a large number of questions—one might even say every question—that might be asked about the relation between causation and responsibility and the relevance of that relation to legal practices and their justification. One set of questions that Moore discusses in many different places in the book concerns the relationship between doing, or being active, and causing. The notion of activity is crucial to ordinary moral thought. Consider an example: in the November 1976 issue, then-Governor Jimmy Carter told Playboy magazine that although he had never been unfaithful to his wife, Rosalind, he had sinned in his heart in that respect. Rosalind Carter and the American public, who elected him president not long after, not to mention the readers of Playboy, seem to have had little trouble forgiving him. Sins of the heart just are not the same as sins of the body. But why not? What justifies the moral line we routinely draw between what we think and what we do?
On one appealing view, the distinction in responsibility between thoughts and actions is an instance of the more general distinction in responsibility between things with respect to which we are passive—things that just happen to us—and things with respect to which we are active—things we do. Carter's lustful musings, on this view, are not something for which he is to be held responsible precisely because they just happened to him, as they do to many a faithful but aching spouse. Had he acted on them, that would have been a different matter. This view has the added advantage of explaining why some thoughts, such as choices and plans, are less clearly, if at all, things for which we are not rightly held responsible. Those thoughts, in contrast to many, are, we might say, things that we do. We choose; our choices do not merely come upon us in the way that our desires do. And that is why choices, in contrast to our feelings of lust, are things for which we are rightly held responsible.
It is not just the distinction in responsibility between thoughts and actions that seems to rest on a distinction in responsibility between the passive and the active. Among other things, the entrenched maxim that “ought” implies “can” appears also to be linked to that distinction. Say the only way that the hydrophobe will take a bath is if he is subdued bodily and thrown into the water. So there is an extremely weak sense in which he “can” take a bath; it is perfectly possible that someone will tie him up or knock him unconscious and put him in the water. But this is obviously not the sense of “can” that gives “ought” place. Were the hydrophobe to fail to bathe, we could not take him to be blameworthy for this failure on the grounds that someone might have knocked him unconscious and thrown him in the bath. It is only the unrealized possibility of his performing the act of bathing that makes bathing something he ought to have done. Responsibility for failure requires the possibility of active compliance, and not merely passive.Footnote 2
It is extremely plausible on its face to think of human activity as opposed to passivity, of the sort that seems central to moral and legal responsibility, to be inextricably linked to causation. To do, it seems, is to bring about; and to bring about, it seems, is to cause. In fact, such a line of thought is so seductive that one might think that a book titled Causation and Responsibility might as well be titled Action and Responsibility. Surely, one might think, causation is crucial to responsibility because activity is. However, one of the more fascinating and, if true, profound claims that Moore is at pains to support in his book reverses the order of explanation. Moore holds that what is crucial for responsibility is causation. Action matters to responsibility, according to Moore, because where there is action, we typically also find the kind of causal contribution that is crucial for responsibility. But it is causation and not action that bears the normative weight.
This paper concerns Moore's grounds for this view and argues that they are unsatisfactory. In contrast to Moore's view, it is argued here that action matters to responsibility because it is action and not because it is an instance of causation (although it is). In rough form, the argument runs as follows. While being active is sometimes exhausted by an agent's causal role, it is not always exhausted by this. Activity with respect to some events requires an identification between the agent, on the one hand, and the protagonist of the event the agent causes, on the other; and it requires that the agent recognize and be guided by that identification. It requires guidance by an “I” thought, or a thought about oneself. The facts about identity make a difference to the facts about activity. On the assumption that activity matters to responsibility, it follows that its relevance to responsibility is not exhausted by causation's relevance.
Of course, causation may matter to responsibility for more than one reason. However, one of the reasons it matters, as is argued here, is because it matters to activity, and activity matters to responsibility for reasons that are not explained entirely by the way in which doing involves causing; the facts about identity and self-identification are crucial to the explanation as well. As shown below, one practical upshot of this is that contra Moore, it may be appropriate for some crimes to be defined in such a way that they can be committed only by those who are active with respect to a result and not by those who cause the result without being active with respect to it, even if their causal contribution to the result is quite large. This is a point that has implications for our understanding of accomplice liability, among other things. In particular, it supports the view that accomplice liability is, in a certain class of cases, a noncausal basis of liability.
Section II of this paper concerns Moore's discussion of a claim that both he and I accept, namely, that causing is a necessary condition of doing; if you are active with respect to an event, you are also a cause of the event. As emerges, in accommodating some putative counterexamples to this claim, it is necessary to hold that at least when conditions present in the putative counterexamples are met, causing is sufficient for doing. To defend the claim that activity with respect to an event requires causing the event, that is, one must hold that causing the event in certain conditions is sufficient for being active with respect to it.
Section III then turns to the claim that causing is sufficient for doing, a claim that Moore himself rejects, holding instead that only some causings are doings. However, it is argued that the class of counterexamples to the claim is wider than Moore thinks; some counterexamples cannot be accommodated in the way that Moore suggests. And further, reflection on such examples leads us to a conclusion that Moore will find uncongenial: some events that one causes are such that activity with respect to them requires satisfaction of entirely noncausal conditions, in particular conditions regarding identity between the agent who is the cause and the agent whose body is involved in the events that are caused.
The result, explored in the conclusion of the paper, is that in such cases, activity matters to responsibility for reasons that go beyond the reason that causing matters to responsibility. Doing itself, and not just causing, carries normative weight. The conclusion of the paper also considers the implications of this position for accomplice liability.
II. DO WE ALWAYS CAUSE THAT WITH RESPECT TO WHICH WE ARE ACTIVE?
A. Clarifying the Equivalence Thesis
In chapter 1 of his book, Moore discusses the claim that doing is both sufficient and necessary for causing, a claim that he formulates like so:
The Equivalence Thesis: X v-ed if and only if X caused a v-ing to exist.
Moore notes that the law seems to presuppose this claim. And Moore himself defends the left-to-right direction of the biconditional—the claim that every doing is a causing—in order to establish that the law in many areas inquires into what agents cause. If every doing is a causing, after all, then the law's evident interest in doings is ipso facto an interest in causings.
Under exactly what conditions is the left side of the Equivalence thesis true? Donald Davidson holds, and not without reason, that facts about what a person does are in part facts about language. He holds that an act is an event that is intentional under some description.Footnote 3 The man who flips the switch and inadvertently causes the burglar to be aware that someone else is in the home performs the act of alerting the burglar; alerting the burglar is among the things the man does because the event that the phrase “alerts the burglar” refers to (Davidson thinks that event is the man's bodily movement) is intentional under the description “flipping the switch,” among other descriptions. In that well-worn example, where v is “alert the burglar,” is the left side of the Equivalence thesis true? Perhaps. But if so, then the left side of the Equivalence thesis entails nothing about activity, in contrast to passivity, with respect to the event referred to on the right side of the biconditional as “a v-ing,” or in the example, “the burglar's becoming alerted.” Put in Davidson's terms, the event referred to on the left side of the biconditional is a different event from that referred to on the right side. The event referred to on the left is, Davidson thinks, the agent's bodily movement, while the event referred to on the right is the change in the burglar's state of awareness. It does not follow from the fact that the man in the example alerted the burglar—and so was active with respect to the bodily movement that the phrase “alerted the burglar” refers to—that he is active with respect to the change in the burglar's state of awareness; that seems to be something with respect to which he is entirely passive.
The point is that active verbs, especially transitive active verbs, often have place even where there is no activity with respect to events that one can infer took place from the fact of the applicability of those verbs. One can infer from the fact that the man alerted the burglar that there was a change in the burglar's state of awareness about his surroundings; the burglar changed from believing that no one was home to the opposite belief, and this fact is implied by saying that the man alerted the burglar. But this does not imply that the man is active with respect to that event in the sense that seems intuitively to be relevant to responsibility. It is not the case, for instance, that just prior to flipping the switch, the man could change the state of the burglar's awareness in the sense that would give “ought” place. Even if changing the burglar's state of awareness would be a good thing, it is not something that the man ought to do precisely because “ought” implies “can” in a sense of “can” that is not present in the case.
What this shows is that the relevant phenomenon from the point of view of responsibility is not, strictly speaking, doing, where that is understood as a language-dependent notion, but instead being active. What we want to know is what the relationship is between being active with respect to a particular event, on the one hand, and being the cause of it, on the other. What the agent does is equivalent to what can be truly said of him using active verbs such as “alert.” What the agent is active with respect to, however, is not a matter of language but a matter of metaphysics. And it is a matter about which our language can easily mislead us. Whenever an active verb has place, there is, to be sure, some event with respect to which the agent is active. There are also, typically, events that we can infer took place; but it is perilous to infer that the agent was active with respect to them. If the agent alerted the burglar, there is something that he is active with respect to (flipping the switch, moving his finger, etc.) but for many purposes we will want to know whether he is active with respect to the change in the burglar's state of awareness, or other remote consequences of the flip of the switch, and we can reach no conclusion about that question one way or another by noting that he alerted the burglar; in some cases he is and in some cases he is not (in Davidson's version of the example, he is not).
To avoid confusion, and to focus our attention on the normatively relevant phenomenon, it is helpful, then, to formulate the Equivalence thesis like so:
The Equivalence Thesis: X is active with respect to a v-ing if and only if X caused the v-ing to exist.
In this formulation, the event referred to on the left side of the biconditional as “a v-ing” is the very same event referred to on the right with that phrase. From here forward, it is this construal of the Equivalence thesis that occupies our attention. What we want to know is not precisely whether X burned Y's house if and only if X caused Y's house to burn; what we want to know is rather whether X is active with respect to Y's house burning if and only if X caused that event. The advocate of the Equivalence thesis holds that in nature we do not find two relations—the relations referred to, respectively, as “____ caused ____,” and “____ is active with respect to _____”—but only one relation, the relation of causation, which we happen to refer to sometimes with the language of activity.
Notice that the Equivalence thesis does not involve commitment to the claim that the tornado that chews a gas station to bits is active with respect to the event in which the gas station is destroyed. There are two reasons for this. First, the variable X in the Equivalence thesis need not be taken to range over the class of all causes; it can be limited to things like human beings and maybe other animals; the Equivalence thesis concerns only what it is for an agent to be active with respect to an event. Second, the Equivalence thesis does not stand alone but is coupled with a view of what it is for an agent to be a cause under which the right side of the biconditional is not true of the tornado. That is, to motivate the view that X ranges only over agents, the Equivalence thesis is best understood as coupled with a theory of what it is for an agent to be a cause, in the sense of relevance to the thesis, under which only things like human beings and maybe other animals can be causes.
There are various theories of the necessary and sufficient conditions under which an agent is the cause of an event. One possibility is agent-causalism—the view that there is a basic, irreducible causal relation that sometimes obtains between agents and events independently of any of the states of or events internal to agents. Add that only human beings and perhaps other animals are capable of being “unmoved movers” in this sense, and we reach the result that the right side of the Equivalence thesis is true only if X is an agent.Footnote 4
But there are other possibilities too. Causation by the agent might be analyzed event-causally and by appeal to events and states that as a matter of fact, belong only to agents. One might think, for instance, that the right side of the biconditional is satisfied only if the v-ing is caused by a complex of mental states of X, including intentions, desires, beliefs, and perhaps other mental states that only human beings and perhaps other animals are equipped to have. On this view, while air motions of the sort that we find in tornadoes do indeed have effects, tornadoes themselves do not. Talk of objects, persons, and the like as causes, the sense relevant to the Equivalence thesis, is to be understood as elliptical for talk of mental states and events as causes.
In what follows, I assume that some version of the latter, non-agent-causal, position is true. I assume, that is, that the phrase “X caused the v-ing” is to be understood as the claim that some appropriate set of mental states of X caused the v-ing. For the sake of convenience, let us use the term “springs of action” to refer to that set of mental states such that if they cause an event appropriately, then it is true that the agent is the cause of the event. The Equivalence thesis is the claim that one is active with respect to an event just in case one's springs of action cause it.
B. Activity with Respect to the Springs of Action
Some have rejected the left-to-right direction of the Equivalence thesis on the grounds that the springs of action—the mental states in virtue of which a person can be said to be the cause of an event—are things with respect to which agents are active. Hugh McCannFootnote 5, for instance, thinks that volitions are acts—they are things done and are described with active verbs, as when we say, “He chose that”—but they lack a component part that is caused by the agent. Thus they resist Wittgenstein's well-known question: “What is left over if I subtract the fact that my arm goes up from the fact that I raise my arm?”Footnote 6 The analogous question in the case of choice or volition would be something like: “What is left over if I subtract the fact that there is a choice of mine from the fact that I made a choice?” Unlike in the case of Wittgenstein's original question, the answer here is quite clear: nothing. There is no identifiable component of the springs of action that is left over when the nonactional component is “subtracted.” In the case of choices, having is sufficient for doing. Since having is insufficient for causing, it follows that there are events with respect to which one is active that one did not cause, namely volitions or choices (at least).
There are two ways to respond to this worry for the left-to-right direction of the Equivalence thesis, either of which is sufficient in this context to lead us to set the worry aside for now, although the second enjoys greater immunity from controversy. First, we might simply deny that the springs of action are things with respect to which we are active. On this view, our choices (if they are among the springs of action) just happen to us—we are not active with respect to them—but they have a remarkable power: when they cause an event, then the event is not something that just happens to us but is instead something with respect to which we are active. Such a view requires accepting the further claim that a person can be active with respect to an event despite being passive with respect to the causes of the event. While I myself do not find this claim problematic, many do.
However, there is another route to setting aside the worry on offer, and that is just to limit the variable v, as it appears in the Equivalence thesis, to events other than the springs of action. On this view, the Equivalence thesis concerns events such as trigger movements, injuries, and deaths and not events such as choices, volitions, and intentions. In fact, it is fairly easy to motivate this limitation in this context. After all, Moore is concerned with the general question of what role causation plays in the law and he is interested in the Equivalence thesis because under it, causation turns out to be endemic to all of the many areas of law—most notably but not limited to criminal and tort law—in which we have a concern with action or with that with respect to which human beings are active and not merely passive.
Under the voluntary-act requirement in the criminal law, for instance, we need to know not just whether the defendant's body moved in a certain way that caused a prohibited result; we also need to know if the defendant was active with respect to that bodily movement. And it is not just activity with respect to bodily movements that is important in the law. We sometimes need to know, for instance, whether a particular defendant was active with respect to the death that took place; we will not find him guilty of intentional homicide if he was not. Although it is undeniable that tort and especially criminal law have a concern with our choices, volitions, and intentions, it is also true that no doctrine in either area of the law inquires into the question of whether we are active with respect to those mental states. When we inquire, for instance, whether the defendant charged with arson intended to burn something or was instead merely reckless with regard to the burning that occurred, we do not inquire as to whether he was active with respect to that intention. The defendant is on the hook for intentional arson if he had the intention, even if that intention merely came upon him passively. The result is that for our purposes it does not matter whether the springs of action themselves are counterexamples to the left-to-right direction of the Equivalence thesis.
C. Activity with Respect to That Which Is Essentially Active
While this subsection concerns putative counterexamples to the left-to-right direction of the Equivalence thesis, counterexamples distinct from the springs of action themselves, what I have to say about them requires appreciation in outline of Moore's stance on the right-to-left direction, the claim that every causing is a doing. A casual reader of chapter 1 of Causation and Responsibility might leave with the impression that Moore accepts that claim. However, as we learn in later chapters, Moore rejects the right-to-left direction and makes the following claims instead: (i) many seeming counterexamples to the right-to-left direction are not in fact counterexamples to it; (ii) the right-to-left direction is true if “caused” is understood as “proximately caused” in the legal sense; but (iii) there is no adequate theory of causation under which it is properly equated with proximate causation in the legal sense, and (iv) the normatively crucial notion is not the notion of being active with respect to an event but the notion of making a substantial causal contribution to it; and further (v) causing an event, especially when one is a less-than-substantial causal contributor, may not be enough for being active with respect to it (which implies that the right-to-left direction of the Equivalence thesis is false).
In addition, Moore is somewhat cagey about exactly how he sees the relationship between being active with respect to an event and being a substantial causal contributor to it. Is being a substantial causal contributor sufficient for activity? His settled position seems to be that it does not matter one way or the other since, as indicated by (iv), the normatively crucial notion is not doing but substantial causal contribution. This conjunction of views needs to be kept in mind in what follows.
Now, we might think that there are at least some cases other than the springs of action themselves in which we are active with respect to something we did not cause. A line of objection recently advanced by John Gardner,Footnote 7 for instance, involves the claim that some events with respect to which we are active have that feature essentially—they can occur only if somebody is active with respect to them—and this is thought to undermine an effort, such as that enshrined in the Equivalence thesis, to analyze activity with respect to them through appeal to some extrinsic relation in which the agent stands to them, such as the relation of being the cause. One might think, for instance, that the event of suffering abuse is the event of being injured by an act of abuse. The injury is not a suffering of abuse, one might think, unless someone else is active with respect to it. (Additional conditions, too, may need to be satisfied; perhaps not every act of injuring infliction is an act of abuse.) This is why one can be abused by one's father but only injured by a storm, even if the physical effects of each are precisely the same. Storms cause injury, but they do not cause abuse, except perhaps in some metaphorical sense. If that is right, then the left-to-right direction of the Equivalence thesis would seem to imply the following claim:
(1) If X is active with respect to Y's suffering abuse, then X caused Y to be injured by abusing Y.
Assuming it is granted that “abusing Y” (in the sense that concerns us) at least entails “being active with respect to Y's suffering abuse,” then (1), together with the left-to-right direction of the Equivalence thesis, implies the following:
(2) If X is active with respect to Y's suffering abuse, then X caused Y to be injured by causing Y to be injured by abusing Y.
Since the term “abusing Y” appears in (2), as it does in (1), regress looms.
The regress here can seem problematic for the following reason. One might think that the regress indicates that the Equivalence thesis necessarily fails to explain what it is to be active with respect to some events, such as sufferings of abuse, since under it we need to appeal to the idea of being active with respect to those events in order even to explain what those events are, and so we must appeal to that notion in order even to assert that a person is the cause of such an event. It is as though we offered a recipe for chocolate chip cookies that had chocolate chip cookies (among other things, perhaps) on the ingredient list. If we want the recipe to tell us how to create chocolate chip cookies in a world that includes none (but does include, say, butter, flour, sugar, baking soda, and chocolate chips), such a recipe is all but useless. And similarly, if we want the Equivalence thesis to tell us how to build activity with respect to some event in a world that includes none (but does include, say, causal relations and nonactional events as well as persons with a variety of complex mental states), it would be of no use to learn that we are first to find activity with respect to an event out of which to build it.
Equivalently, if the man bent on abuse must cause something to engage in the act, what would we advise him to cause? If we tell him to cause an injury that is caused by an act of abuse, we are advising him first to be active with respect to an injury. When he asks what he must cause to do so, we tell him, again, first to be active with respect to an injury. But one might expect a claim such as the left-to-right direction of the Equivalence thesis to tell us how to give meaningful, useful advice to such a man. So the objection goes.
Here is what Moore says in response to this objection and in defense of the left-to-right direction of the Equivalence thesis:
To Gardner's ear, apparently, the state of being abused is not specifiable independently of an action of abusing; for Gardner, “abuse” is an action of which the state is necessarily derivative. Yet notice that even if this is true about the accidents of English, nothing follows about [the left-to-right direction of the Equivalence thesis]. Whether there is a state r of abuse—physical contact of various kinds, for instance—is not settled by our having or lacking a word (or a sense of a word) with which to name it. This would be to reinvent the metaphysical naivety of ordinary language philosophy. Even if it were true that the only state of physical injury referred to by “abuse” (as a noun) was a state caused by the agency of some person—a linguistic fact I doubt—that fact would go no distance towards showing that no such state exists, or that no causal relationship can exist with it as an effect.Footnote 8
As I understand him, Moore's response is to deny that the event referred to as “a suffering of abuse” is essentially one with respect to which someone is active. Rather, he grants for the sake of argument that the words, “a suffering of abuse,” apply to that event or pick that event out only if the event is one with respect to which someone is active. He then claims that one would have to make the mistake of assuming that the metaphysical structure of the world can be read off the structure of ordinary language in order to reach a conclusion about what properties of the relevant event are essential from the facts about the terms that we use to refer to those events. Analogously, the words “the event referred to on the front page of the New York Times” applies only if the Democrats’ loss of the House of Representatives is indeed referred to there. But of course it is not an essential property of that event that it was referred to there. That very event would still have happened, alas, even if the New York Times failed to report it.
Moore's response, however, is not sufficient. The reason is that the metaphysical claim—and not just the linguistic claim—that can be used to support the objection may indeed be true. It may be true that the injury and the suffering of abuse are different events, even if they share all of the same intrinsic properties. At least one of the relational properties of the suffering of abuse, in particular the relational property of X's being active with respect to it, may be essential to it. Now it is possible to have a metaphysics of events that entails that this is not so. If one thinks that all of the essential properties of events are intrinsic, for instance, then of course X's being active with respect to an event is never an essential property of the event. But it seems crazy to think that a plausible claim like the left-to-right direction of the Equivalence thesis can be defended only by those who accept a metaphysics of events with this implication. And crazy or not, it would at least be preferable to be able to defend that claim without appeal to such fancy and controversial metaphysical claims.
However, there is a different and I believe fatal problem for the objection that Moore attributes to Gardner. The problem is that the offered objection to the left-to-right direction of the Equivalence thesis succeeds only if all relevant instances of the right-to-left direction are false. To see this, assume that the right-to-left direction is true across the board: If X causes a v-ing to exist, then X is active with respect to that v-ing. It follows that any injuring that X causes is one with respect to which he is active. Now, say we grant that an essential property of the event of suffering abuse is that the relevant injury is one with respect to which someone is active. Put these things together, and we get the result that any injury that X causes is potentially a suffering of abuse, since it has one of such an event's essential properties, namely, that someone, namely X, is active with respect to it.
Put another way, the phrase “by abusing Y” that appears in (1) and (2) is entailed by what precedes it, namely, that X caused the injury. Hence its appearance there is not problematic if the right-to-left direction of the Equivalence thesis is true. (1) is of the form “If p then q & r.” But if the right-to-left direction of the Equivalence thesis is true, then q entails r. Hence: (If p then q & r) if and only if (If p then q). The two conditionals are logically equivalent. Since “If p then q” does not seem problematic—it merely amounts to the claim that someone who abuses another causes an injury to him—then neither is the logically equivalent “If p then q & r.” (Imagine insisting that the claim that all crows are birds is metaphysically unilluminating since it is also true that all crows are both birds and vertebrates.) The result is that even if we grant that no event is a suffering of abuse unless someone is active with respect to it, the left-to-right direction of the Equivalence thesis implies something perfectly sensible and nonproblematic.
Now, as mentioned above, Moore does not accept the right-to-left direction of the Equivalence thesis. Although he is cagey on the point at best, he accepts only the following weaker claim: if X is a substantial causal contributor to a v-ing, then X is active with respect to the v-ing. This falls short of the right-to-left direction of the Equivalence thesis, since small causal contributors are also causes. But Moore can still accept the response to the objection just offered, since the appealing claim is that a suffering of abuse is essentially caused by an act of abuse. That is, what seems essential to the suffering of abuse is not just that someone is the cause of it but that someone is the cause of it in the way or to the degree that is sufficient for being active with respect to it. It is consistent with Moore's position to hold that this is so when one is a substantial causal contributor, which is itself sufficient although not necessary for being a cause.
In fact, all that is needed to accommodate activity with respect to events such as sufferings of abuse is to hold the following conjunction of positions:
If X causes a v-ing & C, then X is active with respect to the v-ing, & If X causes a v-ing, then C.
Consistent with Moore's position, we can take C to be “X is a substantial causal contributor to the v-ing.” But since this condition is entailed when v-ing is suffering abuse, the objection does not have traction. In fact, since an injury counts as a suffering of abuse only if caused by an act of abuse, the objection will fail no matter what is involved in condition C. Any account of the additional conditions that when added to a causing, are enough to generate activity will succeed in responding to the objection and thus defending the claim that one is active with respect to an event only if one caused it.
So the simple truth contained in the left-to-right direction of the Equivalence thesis, when applied to the case of being active with respect to another's suffering of abuse, is just this: a person is active with respect to another's suffering of abuse only if he is the cause of the other's injury. But to say this is to leave open the possibility that there is more to being active with respect to another's suffering of abuse than causing his injury; additional conditions may need to be satisfied—perhaps, as Moore thinks, you must in addition be a substantial causal contributor to his injury, and perhaps more besides. But this is just to say that establishing the left-to-right direction of the Equivalence thesis leaves open the possibility that the right-to-left direction is false; responding to the objection to the left-to-right direction requires only acceptance of the claim that causing in certain conditions is sufficient for activity, not that it always is. And so, it is to the claim that causing an event is sufficient for being active with respect to it that we now turn.
III. ARE WE ALWAYS ACTIVE WITH RESPECT TO THAT WHICH WE CAUSE?
In his discussion of the right-to-left direction of the Equivalence thesis, Moore considers several different kinds of possible counterexamples to the claim. For our purposes in this section, the most important are the so-called “means restrictions.” In examples of this kind, an essential property of an act is that some result that is involved in it was brought about through the adoption of certain specific means or with the use of certain specific instruments. So, for instance, if X kisses Y, then necessarily that which Y undergoes (lip contact with his cheek, for instance) is caused by X's use of his lips. Touching, crushing, fondling, shooting, chopping, hammering, and many other acts are similar in this respect. If you hammer a nail, then you necessarily whack it with something, and usually a hammer; the man who can make nails flush by exercising his telekinetic powers does not hammer them.
A particularly important subset of the means restrictions, for our purposes, are those in which the restriction involves identity between the agent and a person involved in one of the restrictive conditions. Kissing is a case in point. X kisses Y only if contact is made between a pair of lips and some part of Y's body. But in addition, the lips in question must be X's. Even if X arranges for unconscious Z's lips to be in contact with Y in just the right way, X has not thereby kissed Y. Sandy Kadish coined the term “nonproxyable crimes” to refer to crimes involving acts in this class.Footnote 9 Kadish places rape, or at least certain forms of it, such as those that are defined to involve specific forms of penetration, in this class, which is plausible. Following Kadish, let us use the term “nonproxyable means restrictions” to refer to this subclass of cases. The remaining cases of means restrictions are referred to as “proxyable means restrictions.”
As just described, means restrictions concern actions rather than the relation of being active with respect to an event. But the point that these examples raises can be put in those terms. In means restrictions, the agent is active with respect to some particular event only if certain means were adopted or certain instruments used in causing the event. In nonproxyable means restrictions, identity between the agent who is the cause and some person involved in some required means or instrument is also required for activity with respect to the event caused.
Given this characterization of means restrictions, it is not hard to see why they would seem at least to pose a problem for the right-to-left direction of the Equivalence thesis. If activity with respect to the event in question is secured only if the event is caused with certain instruments or through certain means, then causation itself is insufficient for activity. Cause the same event with different instruments or different means, and you will not be active with respect to it. So causation is not enough for activity. At least, so it seems.
However, Moore does not think that means restrictions pose a problem for the right-to-left direction of the Equivalence thesis. Moore first notes, correctly, that many cases that can seem properly classified as means restrictions may not be. He considers whether removing someone's appendix is a case of means restriction, noting that we would not ordinarily describe this to have been what took place when someone (a doctor even) hits another (a patient even) with his Lincoln Continental and happens thereby to knock his appendix from his body. In this case, Moore thinks, when we say “the doctor removed the patient's appendix,” we conversationally imply that the event was caused by scalpel and not by Lincoln. But activity with respect to the appendix's change of location does not require that one means rather than another was adopted; if the doctor is trying out a new method for automotive appendix removal, then even in this case he is active with respect to the appendix's change of location. Just because our language intimates that an event came about in the normal way and not in some weird way, it does not follow that activity with respect to the event necessarily requires that it came to pass normally.
However, it is one thing to say that cases of means restriction are less common than our language might suggest, quite another to suggest that there are no such cases. And in fact, in the cases listed above, and others—touching, crushing, fondling, shooting, chopping, hammering, kissing, and so on—it would be merely stubborn to insist that they are not cases of means restriction; as though it were merely an overrideable assumption of the normality of the conditions that leads us to conclude that the lips of the man who kissed a child were in contact with the child. Moore acknowledges this point but claims that even cases such as these are not counterexamples to the right-to-left direction of the Equivalence thesis:
[I]t may seem that “kicking,” “kissing,” etc involve specific body motions in the way that “removing” and “felling” do not. Yet this is an illusion brought about by a failure to attend carefully to what must be caused in order for there to be a kicking, a kissing, etc. The actor's specific body parts are involved in such actions, but not as necessary means. Rather, some relation to such parts is required as the state of affairs caused. The state of affairs involved in kissing is not just contact with the lips (or elsewhere) of another; it is contact between the lips of the actor and the lips or other body part of another. The state of affairs involved in kicking in the shins is not just contact with the shins of another; it is contact between the foot of the actor and the shins of another, etc. There is no requirement that the actor move his lips in order to kiss (it may well be a bad kiss if he does not move his lips, but a kiss nonetheless); the only requirement is that the actor move his body in some way so as to cause lip contact.Footnote 10
As I see it, Moore's point here (despite reference to the “illusion” at the beginning of this remark) is not to deny that cases of kicking and kissing are cases of means restrictions (and nonproxyable means restrictions, at that) but instead to claim that true cases of means restrictions are not counterexamples to the claim that causing is sufficient for activity. In the case of kissing, for instance, one causes one's lips to be in contact with another and is thereby active with respect to that event. There is no restriction on what means one adopts to bring about that event; the restriction of means is folded into the event one causes rather than a condition on the way in which one causes it.
In fact, Moore makes the same move much later in the book in discussion of the narrower case of nonproxyable means restrictions. Moore is considering an example of Kadish's of a person who deceives another into believing that he is unmarried thereby causing him to marry, in ignorance, a third person. It is clear that the deceiver is not guilty of bigamy on these grounds; he might not be married to anyone himself, much less two people. Here is Moore:
Redrafting such statutes [such as bigamy 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 he is not causing himself to get married (anymore than he is doing any marrying). This is because the event such non-proxyable crimes seek to prohibit essentially involves reference to the actor being charged: it is not criminal to cause another to get married, only to cause oneself to get married.Footnote 11
Moore's central point here concerns the proper boundaries of liability rather than the proper construal of the relation between causation and activity (the point about liability is discussed in the conclusion of this paper). However, here, as before, Moore's view is that even nonproxyable means restrictions are not counterexamples to the right-to-left direction of the Equivalence thesis. If X causes the event in which he himself becomes doubly married, then he is active with respect to that event. The restriction that the causer is identical to the marrier is to be folded into the event caused rather than seen as a limitation on the way in which that event is caused.
Moore's redescriptive move does not serve to defend the view that nonproxyable means restrictions are not counterexamples to the claim that causation is sufficient for activity. To see this, start with a more careful characterization of the objection to which Moore is responding. The objector reasons like so, where E is the event in which Z's lips make contact with Y's body:
(i) If X causes E & X ≠ Z, then X did not kiss Y.
(ii) If X did not kiss Y, then X is not active with respect to E.
(iii) ∴ It is not the case that: If X causes E, then X is active with respect to E.
Moore responds by rejecting (ii): he denies that everything that we are active with respect to figures in a description of something we did. Moore's move is to note that the following is true:
(iv) If X causes E & X = Z, then X kissed Y.
He then notes that X causing Z's lips to be in contact with Y's body when X = Z is just the same as X causing his own lips to be in contact with Y's body. Hence (iv) is an instance of the right-to-left direction of the Equivalence thesis. Notice, however, that at this point no reason has been given to reject (ii). The fact that in standard cases of kissing—where there is identity between the cause and the person's whose lips make contact—there is no reason to think that there is causation without activity says nothing about nonstandard cases, such as those in which X causes lip contact with Y but not with his own lips. The problem posed by nonproxyable means restrictions is not that they themselves are counterexamples to the right-to-left direction of the Equivalence thesis; the problem is that in each case we seem to be able to duplicate the causal relation between an agent and the relevant event without there being activity with respect to the event. The way we do it is by subtracting identity.
But to stop here would be to miss Moore's point. Moore's point is that by adding identity to a case of causation, we make the verb of action applicable but without changing the facts about activity. The idea is that the relation between X and the lip contact with Y's body is exactly the same whether or not the lips are X's. Thus, if in the case where the lips are X's, X is active with respect to the event, we should say the same thing about the case in which they are not X's lips at all. We therefore learn from the case where there is identity that we have in the world a relation, the relation of activity, even in the case in which we lack identity. Our words are apt to mislead us about this—when X pays Z to kiss Y, X does not kiss Y—but we should not let them. We should recognize instead that even in such a case we have the same relation between X and the relevant event as we do when X kisses Y himself. At least, so Moore believes.
The problem with Moore's response to the objection is that there is no reason to believe that we are holding fixed the facts about activity when we alter the facts about identity. Moore's response to the objection fails if we cannot infer from the fact that there is activity when X causes his own lips to be in contact with Y that there is also activity when X makes an equivalently large causal contribution to someone else's lips making contact with Y. If the facts about identity make a difference to the facts about activity, then this is not a safe inference. And in fact there is reason to think that the facts about identity do make a difference to activity in nonproxyable means-restriction cases. Moore accepts that both (i) and (iv) are true: in nonproxyable means restrictions, whether a person is correctly characterized as having acted will turn on the question of identity independently of the facts about causation. But our willingness to apply the relevant verb of action in the one case and not the other—the willingness to describe X as having kissed Y only if it is his lips that contacted Y—is significant: it tells us that we are sensing some different and significant relation between X and the event in the one case as opposed to other; it indicates that we sense a difference in activity in the two cases.
In fact, linguistic intuition can provide some kind of evidence that we sense a difference in activity in such cases even when there is no special verb of action that applies in the one case and not in the other. Joel Feinberg notes that one can make another laugh—can be active with respect to the event in which the other laughs—despite the fact that it is not true, in fact it is even nonsensical, to say that one laughs the other.Footnote 12 What Feinberg does not note is that sometimes we use the verbs “caused” or “made” to indicate activity with respect to an event when we lack a special verb for picking out activity with respect to such events. This is what we mean when we say, “He made you laugh” in contrast to “You laughed at him.” The difference between making someone laugh and being laughed at is a difference in activity. We find ways to use words to indicate such differences even when we lack appropriate verbs. I cannot think of a case of nonproxyable means restriction in which there is no verb of action that distinguishes the case in which there is both causation and identity from the case in which there is causation without identity. But even if there were such a case, we would find a way in language to register the difference in activity.
As we see above, Moore is right to criticize those who reflexively infer facts about metaphysics from facts about language. It is indeed a naïve error. There is plenty of reason to think that there are distinctions in language where there are no distinctions in reality and implications in language about what exists where there are no such existent things. But still, linguistic evidence is evidence about how things are and what the world includes—defeasible of course, but evidence still. The fact that we register a difference in language between causing where there is identity with the person involved in the event caused and causing where there is no such identity and the fact that the difference is registered by intimating a difference in activity in the two cases give us some reason, even if not decisive, to think that the facts about identity make a difference to the facts about activity.
But if this does not convince, there is additional evidence that is not linguistic. Return first to the assumed analysis of causation by an agent, the analysis according to which causation by the agent consists in causation by certain crucial mental states of the agent. These mental states necessarily include representations of the event to be caused. To be the cause of E, for instance, one must intend to E, or must desire E, or believe that E will come to pass, or something of this nature. Any plausible, specific proposal about the necessary and sufficient conditions under which an agent is the cause of an event in the sense of relevance to the Equivalence thesis must appeal to such mental states. But in the cases of nonproxyable means restrictions, such mental states necessarily represent one of the participants in that future event as identical to the subject of the representation. X did not cause his lips to be in contact with Y unless among the causes of the event was a mental representation of X's of himself making such contact with Y.
John Perry discusses Ernst Mach's experience of seeing a person some distance from him at the other end of a bus and thinking, “What a shabby pedagogue!” before realizing that he is looking at his own reflection in a mirror.Footnote 13 The discovery that he is looking in a mirror prompts Mach to the discovery that the man in the mirror is himself. On making this discovery, Mach alters his mental representation of the man in the mirror; that representation comes to include a representation of the relational property of the man of being identical to the subject of the mental representation, namely Mach. Mental representations of the kind that Mach thereby achieves—self-representations—are necessarily among the springs of action in cases of nonproxyable means restriction. Add that when there is causation by the agent, the world comes to match the content of the agent's mental representations of it, and it appears that a necessary condition of causation by the agent in cases of nonproxyable means restriction is identity between the agent who causes and the agent who is involved in the event caused. Add that there is no activity without causation by the agent—the claim discussed in Section II of this paper—and it starts to appear that the facts about identity in cases of nonproxyable means restriction make a difference to the facts about activity.
One more potential piece of evidence for the claim that the facts about identity are part of what determines the facts about activity is worthy of note. This piece of evidence derives from the solution to a well-known problem about personal identity, namely Bishop Butler's circularity problem for memory theories of personal identity.Footnote 14 According to the memory theorist, a present memory of a past experience is sufficient for identity between the subject of the memory and the subject of the past experience. But under what conditions is a present representation of a past experience a memory as opposed to a delusion? If the answer is that the present representation is a memory only if the subject of the representation is identical to the subject of the past experience, then there is vicious circularity in the proffered theory of personal identity.
An alternative, noncircular answer is that a present representation counts as a memory of the sort that suffices for identity only if it is caused in the right way by the experience which it represents.Footnote 15 And we can make the same move to avoid circularity under a view of personal identity according to which present execution of a past intention suffices for identity between the actor and the intender only if the intention causes later conduct in the right way. Not unrelatedly, any adequate account of what it is for an agent to cause an event must draw also on some notion of being caused “in the right way.” The agent causes the event only if the springs of action cause the event in the right way, and not through some peculiar roundabout route.
In cases of nonproxyable means restriction, then, we have causation in the right way by the springs of action of just the sort that is sufficient for identity with the subject of those mental states. If there is no identity, it could be because there is no causation of the right sort; that in turn would undermine causation by the agent, which would in turn undermine activity. So those who hold a psychological-continuity theory of personal identity and defend it against the circularity charge through appeal to causation in the right way ought also to hold that the facts about identity matter to the facts about activity in nonproxyable means restriction cases.
None of this is decisive. Moore could reject the veridicality of the linguistic evidence. He could reject the claim that causation by the agent in nonproxyable means-restrictions cases necessarily involves causation of an event by a self-representation that it matches. He could reject psychological-continuity theories of personal identity or find an alternative solution to the circularity worry that these theories face. Or he could insist that causation “in the right way” means something different in theories of personal identity from what it means in the analysis of causation of an event by an agent. There are ways out for Moore. But it seems to me they have diminishing returns. Why not just accept that nonproxyable means restrictions provide counterexamples to the claim that causing is sufficient for activity? If we are to have activity in such cases, in addition to causation we need identity.
IV. CONCLUSION: THE NONCAUSAL BASIS OF ACCOMPLICE LIABILITY
Kadish thinks that the existence of morally important cases of nonproxyable means restriction, such as bigamy and rape, shows that the relationship between accomplices and the objectionable events for which they are to be held criminally liable is importantly different from the relationship between principals and those events.Footnote 16 If X holds Y while Z rapes her, X has necessarily fallen short of raping Y; the crime, after all, is one of nonproxyable means restriction. But X is nonetheless to be held liable for the rape of Y, and so we need a basis of liability that is importantly different from that on which we found Z's.
Moore thinks this line of thought is flawed. After arguing that many crimes such as rape that are in fact defined so as to require identity between the defendant and the person involved in the crucial objectionable event should not be so defined, he makes the following remark:
Non-proxyable crimes . . . 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 non-proxyable—battery, homicide, and rape coming to mind. But for those crimes that are rightly so-defined, such as (perhaps) bigamy and escape, when the evil prohibited has not occurred—because the perpetrator has not caused himself to be in a certain state—then no form of liability should attach, accomplice or otherwise. 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 non-proxyable 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 non-proxyable crimes goes no distance in showing a need for accomplice liability.Footnote 17
Moore here commits himself to the claim that on the assumption that a crime such as bigamy is correctly defined so as to be a case of nonproxyable means restriction—the crime is committed only if the person who causes someone to be doubly married is the same as the one who says “I do” and is thereby doubly married—those who aid bigamy are not to be held criminally liable. His idea seems to be that in such a case, the court is trying to determine if the defendant caused the event of himself saying “I do” and thus becoming doubly married. If the defendant merely aided another to be involved in such events—even if his aid was a substantial causal contributor to those events—then he did not cause such events and so cannot be held criminally liable. As Moore puts it, without identity, the “harm has not occurred at all” because “the harm” in this case would be the defendant becoming doubly married, which is not what happened if he merely aided another.Footnote 18
If we hold, as I think we should, that this is the normatively incorrect result—aiding bigamy is no less worthy of liability than aiding theft, say—Moore has a ready response: bigamy should not be defined in such a way as to make it a case of nonproxyable means restriction. Presumably, on the alternative, proxyable definition of bigamy, a person is guilty of the crime only if he causes someone (maybe himself, maybe someone else) to be doubly married. Under this definition, one can be guilty of the crime of bigamy even if one never says “I do” and is never married to anyone. If the argument of Section III succeeds, then such a statute allows for criminal liability in the absence of activity. After all, if there really are cases of nonproxyable means restriction, then there are events that one cannot be active with respect to unless one is the very person whose body is involved in those events. No linguistic definition can change the metaphysical facts, any more than we can instantly decode the metaphysical facts from our language.
So, in addition to causation, activity with respect to the relevant events requires identity with the person who says “I do” and is doubly married. But any effort to impose liability in the absence of activity, which a nonproxyable definition of bigamy would do, assuming that bigamy is indeed a case of nonproxyable means restriction, requires justification: when causation is insufficient for activity, why should we hold the party criminally liable? Someone who takes Kadish's line has an answer: we should not unless the social relations between the defendant and the perpetrator of the crime are such as to warrant liability in the absence of activity. Liability in such a case has a source that is quite distinct from its usual source. But this is to appeal to accomplice liability as an independent ground. It amounts, in the end, to abandoning the approach that Moore supports.
The point just made extends also to the case of rape, a crime that Moore specifically claims should be defined in some proxyable way. The problem is that if the crime is defined as, for instance, causing another's body to be penetrated by someone's, leaving it open as to whether the someone in question is also the person who causes the penetration, the definition allows for liability without activity. Here, too, activity requires identity, but the definition of the crime does not. If Moore is to explain and justify such an imposition of liability, he cannot do so by appeal to special social relations between the defendant and the person whose body does the penetrating, the sort of relations that one might have thought to be present whenever one person helps another. To make such an appeal is to accept what Moore denies, namely that complicity is an independent and noncausal basis of liability.
Why should we think that social relations between aiders and those they aid can take up the needed slack in cases of nonproxyable means restriction and so provide us with a ground for justifying the imposition of criminal liability even in the absence of activity? Recall that in such cases, what prevents the aider from being active with respect to the relevant events is the absence of identity between the aider and the person whose body plays the crucial role. But there is still a relation between the aider and the person whose body does play the crucial role: there is the social relation constituted by lending a helping hand. In rape, it is not the aiding defendant's body that does the penetrating of the victim, but still, the defendant bears a more intimate relation to the person whose body it is than do those who play no role in the event; social bonds, although not the special activity-relevant bond of identity, tie him to the person who actually does commit the crime. Although I cannot argue the point here, it seems to me that social bonds are morally significant in some of the same ways in which identity is morally significant—some but not all. Social bonds are insufficient for activity in nonproxyable means-restriction cases; identity is a necessary condition there, and a social tie is not enough for identity; a friend is not literally another self. But social bonds provide us with a ground for liability in the absence of activity precisely because there is moral similarity between identity and the kinds of social ties that are constituted by the lending of aid.
The problem that Moore faces in trying to replace accomplice liability with causation is really at its heart a conflict with moral intuition. Intuitively, passivity with respect to an event undermines responsibility for it. To cite the intuitions with which we began, Carter is not to be held liable for his lustful thoughts because they merely happened to him. Ought has place only where the possibility of active compliance is also present. Since in some cases activity is secured only if there is identity between the person who causes the event and the protagonist of the event, it follows that causation is sometimes insufficient for activity. To insist, as Moore does, that we can be responsible for what we cause even in such cases is to run counter to the intuitive thought that that with respect to which we are passive, even if it is something we cause, is not something for which we are rightly held liable. We can answer the concern by pointing to additional noncausal conditions—such as the social ties between accomplice and principal—that extend the circle of liability beyond that with respect to which we are active. But Moore cannot help himself to such resources without giving up his claim that accomplice liability is superfluous.
To the degree, then, to which we hope for a law that conforms with and is guided by the contours of our moral thought, we should hope also for a law that gives pride of place to the facts about activity. Causation matters to responsibility, but only because it is necessary for what really matters, namely, activity.
