A. Causation versus Double Prevention—“The Beheading Problem”

In the Talmud it is written that “Pseek raisha ve-lo yamut”—“you can't cut off the head of a chicken and then say you are not responsible for its death.”Footnote ³ Schaffer would have his readers believe that my view commits me to denying this. For he interprets me (correctly) to think that absences can neither be causes nor be effects; that therefore omissions and preventions are not causal in nature; and that therefore double preventions cannot be causal either. From this Schaffer would have me conclude that there can be no cause-based responsibility for the death of a chicken (or a person) if you decapitate her (because decapitation only prevents what would otherwise prevent death—blood flow to the brain—and therefore is doubly preventative and not causal in nature).

This is an absurd conclusion, and if I were committed to it, I would be in trouble. In his review, Schaffer recapitulates his earlier work at some length.Footnote ⁴ This work showed how many instances of what we take to be causal relations are in reality double-preventative in nature if we but look closely at the details of what is going on. Schaffer then depicts a series of conclusions that he says must follow if one accepts that double preventions are noncausal in nature, such as the demise of the causal theory of action.Footnote ⁵ Any one of these would be hard to swallow.

What Schaffer downplays is that in my book I accept rather than deny the Talmudic conclusion: I affirm that decapitations are both causings of death and killings and that we are responsible for such deaths accordingly. And I do this while indeed affirming the thesis that Schaffer calls “No Absences,” namely, that absences cannot serve as causal relata, neither as causes nor as effects.

I perform this hat trick by distinguishing two kinds of double-prevention cases that I variously call “macro” versus “micro,” or “nonclose” versus “close.”Footnote ⁶ Most of Schaffer's examples (such as that of decapitation) I class as “close,” but perhaps some are “nonclose.” In any case, there is responsibility enough to go around: for nonclose cases, the responsibility is based on the relation of counterfactual dependence, which I defend in the book as a desert base independent of causation;Footnote ⁷ and for the close cases, the responsibility is based on the causal relation, not that of counterfactual dependence.Footnote ⁸

So why the fuss by Schaffer about these cases? For two reasons, apparently, one for each disjunct of the thesis of mine that Schaffer calls “Disjunctivism.” One, Schaffer does not believe that my view can allow for a cause-based responsibility in the close cases; and two, Schaffer does not believe I can sustain counterfactual dependence as an independent desert base for nonclose cases, since such dependence seems too lax in its requirements for responsibility. As Schaffer divides his discussion of these two worries between two separate sections of his article, so do I, dealing with the first here and the second in the succeeding section.

Schaffer himself affirms that there is both causation and a cause-based responsibility in cases of close double preventions such as decapitation killings. What he denies is that I can (consistently with my No Absences thesis) join him in affirming these two conclusions. Schaffer's denial here is, he thinks, based on three reasons: first, the distinction between “close” and “not-close” is “unclear”: Schaffer indeed denies that my distinction “is sufficiently articulated yet to take seriously” (411) (ouch!); second, my accepting that there is a cause-based responsibility in cases of close double preventions “is just a surreptitious way to abandon No Absences,” and that “would be surrender” by me (412); third, I am allegedly inconsistent in that I admit that “the beheader ‘did not literally’ cause death” while I hold that the beheader has a cause-based responsibility for that death (413).

Schaffer's first point is the one requiring the most extensive discussion, so let me begin with it. Schaffer is unclear how my “macro/micro” distinction relates to my “nonclose/close” distinction; so let me clear that up first. My intent in the book is simply to have “micro” be the name for looking closely at the details of a case, which then allows one often to conclude that in detail it was a case of double prevention. If one looks in detail (i.e., “micro”) at a case that had figured prominently in an earlier version of Schaffer's present paper, a case where a terrorist blows up a dam, floods a city, and kills its inhabitants, one sees the following: Terrorist's act of setting off the explosion causes the dam to crumble and fall to pieces; the water rushing out of the reservoir causes both the flooding of the city and, through that event, the death of the city's inhabitants. Finding intuitive (as do I) the conclusion that Terrorist has a cause-based responsibility, Schaffer in his former draft queried: “in what possible sense is the collapse of the dam ‘close’ to the thousands of scattered deaths from the ensuing flood?” .

Schaffer here ignored what I explicitly said in the book: in cases of double prevention, closeness is judged between the event/state of affairsFootnote ⁹ caused by Terrorist's act and either the harm for which responsibility is to be assessed or the cause of that harm which is unleashed by defendant's act.Footnote ¹⁰ I agree with Schaffer that the deaths of the city's inhabitants are not close to the collapse of the dam; but my question was and is, is not the collapse of the dam close to the water right behind the dam rushing out of the reservoir immediately upon the dam's collapse?

Schaffer has now substituted another hypothetical case for the dam case just discussed (412). The new hypothetical involves the terrorist who blows up an air traffic control tower so as to prevent that tower from sending needed signals to pilots, with the result that there is a mid-air collision and massive loss of life of those on the colliding airplanes. This time Schaffer has succeeded in producing a regular (that is, non-close) instance of double prevention. The act of blowing up the tower caused it to fall, but its fall is close neither to the deaths of the passengers nor to what immediately caused those deaths, the collision of the aircraft.

Such closeness judgments will seem “ad hoc” (413) and unmotivated until we broaden our vision somewhat, something we need to do anyway to address Schaffer's more general charge of “unclarity-so-serious-as-not-yet-to-be-worthy-of-consideration.” Let us thus ask the “Why closeness?” question as a preliminary to asking the “What is closeness?” question.

In the book I advert explicitly to another “closeness” beheading problem,Footnote ¹¹ one that I and other philosophers have dealt with for more than fifty years.Footnote ¹² My forty-year-old example of the problem is this one: Can Herod truthfully deny responsibility for intending the death of (intending to kill) John the Baptist because, while Herod intended to decapitate John (in order to please Salome with John's head on a platter), Herod only believed that John could not live without his head?Footnote ¹³

We use the intention/foresight distinction to distinguish both degrees of moral culpability and degrees of permissibility/impermissibility of consequential justification of actions.Footnote ¹⁴ Intending that one's acts cause a bad event (such as John's death) makes one more culpable than does merely foreseeing that one's act will cause that event; and intending that one's acts cause a bad event precludes consequential justification of those acts when foreseeing that one's acts will have those bad effects does not. In this, the moral uses of the intention/foresight distinction are very much like what I claim in my book to be true about the cause/counterfactual dependence distinction; for the latter distinction, too, is taken to mark both degree of blameworthiness in cases where no justification is in sightFootnote ¹⁵ and degrees of permissibility when consequential justifications are (because good consequences are in the offing) possible.Footnote ¹⁶

Yet cases like that of beheading present a challenge to the intention/foresight distinction, just as they do to the causation/counterfactual dependence distinction. What seems intuitive is to say that Herod both intended to kill John the Baptist and that he did kill (i.e., caused the death of) John. Yet in the way of both judgments is the “micro” insight that Schaffer presses: what really killed John (by my No Absences thesis) were the processes of decay of brain cells when not supplied with oxygen and sucrose in the blood; Herod (again, by my lights as Schaffer sees them) only caused or intended the severance of John's head from his body. Herod neither caused nor intended John's death, even though John's death counterfactually depended on Herod's act and even though Herod foresaw with certainty that John could not live without his head. This, because counterfactual dependence is to be distinguished from causation, just as intent is to be distinguished from foresight (or predictive belief).

In the literature on the intention/foresight distinction, there are three general responses that philosophers have adopted over the last fifty years to Schaffer's kind of “micro” insight. One is the response with which Schaffer wants to stick me on the cause/counterfactual distinction: admit that Herod really neither caused nor intended John's death; he caused and intended only John's decapitation. This view lets the (moral responsibility) chips fall as they may, which is, of course, very counterintuitive.Footnote ¹⁷

The second response is just as bad, but in the opposite direction. Rather than restricting what we intend, this response broadens what we intend to include all the things that we foresee with (relative) certainty.Footnote ¹⁸ This is the kind of view Schaffer holds about the cause/counterfactual distinction: let there be a causal connection in all double-prevention cases, because there is counterfactual dependence in such cases.

The third response is the “closeness” response: admit that Herod has both an intention-based and a cause-based responsibility for John's death but deny that this is because John's death was either foreseen by Herod or counterfactually depended on Herod's actions, respectively. Rather, Herod has such an intention and cause based responsibility because Herod intended and caused something (decapitation) that is close either to John's death or to what immediately caused John's death.

The motive for this third response (which is the one I adopt explicitly in my book for both intention and causation) is simple. We cannot easily abandon the distinctions between intending/foreseeing and causing/being counterfactually necessary for (which the second response does), nor can we easily accept the counterintuitive conclusions of the micro view of these matters (which the first response does). Therefore we must use commonsense consolidations of “close” events (or representations of these, in the case of intentions) in our assessments of cause and intention-based responsibility.

This repair to the event consolidation of common sense extends beyond the intent/foresight and cause/counterfactual contexts. This, too, I point out in my book,Footnote ¹⁹ giving an extended example from the law of insurance contracts. As I recognize in the long appendix to my book,Footnote ²⁰ when it comes to the sizing of events for determining how many occurrences there are of harm or harm-causing events under all-risk insurance policies, the true metaphysics of event-individuation has a limited role to play. More important are the macro-sized events of ordinary thought; this on the supposition that such macro-sized events are normatively relevant in the insurance-contract context because they are much more likely to figure in the intentions and expectations of contracting parties than are the micro-sized events that are the subject of some metaphysical theories of event-individuation.

Even closer to home is the fourth context in which some event/state of affairs consolidation plausibly plays a role, a context to which reference is also made in my book.Footnote ²¹ This is the context where we measure the degree of moral blameworthiness by the “unit of offense” determinations of double jeopardy.Footnote ²² In keeping blame and punishment proportional to desert, one must assess not only what wrong some culpable actor did but how many times he did it. While volume discounts are possible,Footnote ²³ usually robbing one bank on Monday, another on Tuesday, and a third on Wednesday merits more blame and more punishment than simply robbing one bank one time (holding constant all other desert-determining features of such robberies, such as amount of money taken, force used, etc.). Here, too, some event consolidation is needed to render the metaphysics of act-token individuation usable for these normative purposes.

Other moral, legal, and commonsense contexts call for other forms of event consolidation. Where we must find the spatiotemporal location of events for various legal purposes, for example, often here, too, metaphysically distinct consequences of events are included as part of the event itself in order to reach normatively plausible, commonsense conclusions.Footnote ²⁴ The motive for such event consolidation in all such cases, including the one presently at issue with Schaffer, is to keep distinctions we can ill afford to abandon in line both with common sense and with normatively plausible conclusions.

Despite these good reasons to adopt the third, “closeness” strategy distinguished above, it is still on point for Schaffer to ask (as he does) the “What?” question about closeness: In the context we are discussing, what does “closeness” mean? (1) Is it spatial closeness, Schaffer asks (411) (2) Temporal closeness?Footnote ²⁵ (3) Spatiotemporal closeness?Footnote ²⁶ (4) Some sort of logical connection?Footnote ²⁷ (5) Some sort of mereological or constitutive relation?Footnote ²⁸ (6) Strong or invariant causal connection?Footnote ²⁹ (7) A moral connection of one type or another?Footnote ³⁰ (8) A contingent, case-by-case connection, whereby it just turns out that to foresee a close consequence will (in almost all cases) also be to intend that consequence?Footnote ³¹ (9) A subjective criterion of closeness, whereby an actor who thinks that to intend x is to intend y will be held to intend y if he intends x?Footnote ³² My younger selfFootnote ³³ sounded a lot like the present Schaffer (who, after all, is still young, comparatively speaking) in rejecting all of these in the context of drawing the intention/foresight distinction, for they all either beg the question (as do some versions of (7)) or draw the lines in the wrong places.

But consider the following variations on a tenth possibility, one that I shall call the “common sense” possibility.

a. Herbert Hart. “When a foreseen outcome is so immediately and invariably connected with the action done that the suggestion that the action might not have that outcome would by ordinary standards be regarded as absurd, or such as only a mentally abnormal person would seriously entertain,” then to intend the action is to intend the outcome.Footnote ³⁴

b. Philippa Foot. “What is the criterion of ‘closeness’? If we say that anything very close to what we are literally aiming at counts as if part of our aim?” Foot eschews giving such a criterion, answering instead by means of two paradigm examples:

1. (i) The crushing of a child's skull and that child's death must be identified as one for purposes of asking if the defendant who intended the first must be construed to have intended the second, else it would “make nonsense of it [the distinction] from the very beginning.” These two events are “obviously much too close for an application of the” distinction.

2. (ii) The blowing to bits of a man by dynamite and his death may be identified as one for purposes of asking if the actor intended the second if he intended the first, else the distinction between intent and foresight would be shown to be “ridiculous.”Footnote ³⁵

c. Jonathan Bennett. When one event is “tightly bound” to another event, to intend the one is to intend the other. Bennett defines “tight binding” as:

d. Neil Delaney. We should use “the notion of ‘felt closeness’ between effects in determining what an agent (directly) intends rather than merely foresees.” Delaney gives two readings to “felt closeness”:

1. (i) Where the two events are proper parts of the object of an overall complex intention; and

2. (ii) Where “a rational agent cannot have the former intention without at the same time having the latter.”Footnote ³⁷

I take all of these to be variations of a conventionalist criterion of “closeness.” Generally speaking, two events are too close to be separated (for purposes of ascribing either causation or intentionality) if ordinary folk would feel them too close to be separated; would regard it as absurd or ridiculous to separate them; would not have the faintest idea about how they could be separated; and so on. Such a test seems to allocate Schaffer's kinds of examples the way I in fact allocate them in my book: decapitations are much too close to death for one to maintain with a straight face that although one caused or intended the first, he did not cause or intend the second. The Talmud has it right about that chicken.

To be sure, no one should be exceedingly pleased with this commonsense criterion of “closeness.” As Bennett concedes, it is “rather loose,”Footnote ³⁸ or as Delaney, too, laments, giving “closeness” only two cheers, “sadly, the key notion (closeness) is vague and the salient principle (of intent consolidation) is a little ‘hinky.’”Footnote ³⁹ Still, that “closeness” is vague makes it not much worse than other, perfectly serviceable concepts.Footnote ⁴⁰ Indeed, is “closeness” in the present context any fuzzier than the close possible worlds that David Lewis (and Lewisians like Schaffer) tells us are the worlds we are to consider when we ask after the truth value of the consequent clauses of counterfactuals?Footnote ⁴¹ Moreover, the alternatives (to retaining the closeness criterion) are both worse: conceding that decapitations are not killings; or saying that all counterfactual dependings are causings. I have long agreed with Schaffer's generally holistic approach to epistemic justification, according to which we should assess the plausibility of one item (like closeness) by the plausibility of what must be given up if that is given up. Unlike Schaffer, I find the least costly equilibrium of beliefs to be one affirming the closeness distinction, vague as it admittedly is.Footnote ⁴²

I am now in a position to spell out (much more quickly) my replies to Schaffer's other two objections. Recall that Schaffer's second objection is that by pursuing the closeness strategy, I am surreptitiously abandoning my No Absences thesis and that this is a “surrender” by me. Not so. To say that the beheader causes the death of his victim because he causes a state of affairs—decapitation—that is close enough to death is not to say that absences can be causes. The idea of closeness nowhere relies on causing by absences (omissions), causing of absences (preventions), or causing through absences (double preventions). Absences (in any of these roles) are absent in the closeness account.

Schaffer's third objection is that I have not dealt with the intuition that the beheader caused the death of his victim and therefore killed him. Indeed, I am supposed to be flatly opposed to this conclusion because I admit that the beheader did not literally cause death. In truth, my closeness doctrine allows me to say two otherwise contradictory things: “the beheader did not cause the death”; and “the beheader did cause the death.” The first statement is true on the fine-grained (or “micro”) view of events; the second statement is true on the coarse-grained (or “macro”) view of events.

One might think that I am equivocating on the meaning of “cause,” but that, too, is not so. I do not stipulate some secondary sense of “x causes y”Footnote ⁴³ with the closeness maneuver; no more than one stipulates a secondary sense of “x intends y,” with the analogous closeness maneuver. In both cases, what shifts is what is caused or intended; not what causings or intendings themselves are.

One might also think that I am giving up on the thesis Schaffer calls “Unity,” namely, the thesis that law, science, morality, and common sense share a single core notion of causation. If by “notion of causation” one means things related (relata) as well as relation, then this is true to a certain extent. For although the causal relation is asserted to be the same in these four domains, the relata used in commonsense explanations, evaluations, and legal ascriptions would be in part conventional and thus somewhat divergent from the true metaphysics of these matters. Yet my book refuses to defend that version of “Unity” anyway. As I recognize in the chapters on causal relata, in law (as well as in morality and in commonsense causal explanations) we use a constructed notion of relata, close to Davidson's view of whole events, whereas the correct metaphysical relata of the causal relation are the much finer-grained states of affairs.Footnote ⁴⁴ So further event consolidation via the notion of “closeness” is just more constructivism about relata, not a sudden loss of metaphysical purity.

This has been a long and complicated reply to Schaffer's dominant criticism of my book. This should not be surprising in light of the nature of the objection replied to, which is essentially, “you have not said enough about closeness.” So I have said more. But in closing, notice that there is a much shorter route to answering Schaffer's criticism, one that does not depend on closeness. This is the route that sees the causal relation as being such that there is no bar to “action at a distance,” perhaps some form of primitivist singularism, some kind of primitive powers view, or some form of reduction-to-laws view of the causal relation—indeed, any view of causation that can handle spatiotemporally remote causal relations. Then one could hold that acts of beheaders cause deaths simplicitas—not through any absence, not through any existent states of affairs such as head-severings, but directly. Such causings of death nowhere take absences as relata, for the moving of an axe and the victim's death are both existent events. The result of the strategy would be to regard double preventions as causal while preserving the No Absences thesis and thus preserving the view that omissions and preventions are noncausal.

Attractive as this last strategy is, it, too, comes at some moral cost (in addition to whatever metaphysical costs are attached to allowing for causation at a distance). One such cost is the need generally for double preventions to be noncausal. The distinction between causal and counterfactual relations is the metaphysical difference that can make sense of the moral differences between double preventions and causings, moral differences in blameworthiness and in permissibility. But more crucially, the double prevention/causation distinction is a seemingly critical part of the thousand-year-old distinction between doings and allowings.Footnote ⁴⁵ An allowing is a case of double prevention where the actor's preventative act restores a status quo that she would have done no wrong in not disturbing to begin with. Schaffer nowhere discusses these cases,Footnote ⁴⁶ yet giving them up would be an important cost to this last proposed version of how we should look at things.

B. The Alleged Overinclusiveness of Counterfactual Dependence Unaccompanied by Causation: “The Problem of Correlates”

As previewed in the prior section, Schaffer also questions the other half of my disjunctive thesis about responsibility, namely, the half that attributes a distinct (if lesser) responsibility to counterfactual dependence when unaccompanied by causation. Like Chris HitchcockFootnote ⁴⁷ and Kim FerzanFootnote ⁴⁸ in their critiques of my book in another symposium, Schaffer worries about the possible overbreadth of a responsibility predicated on counterfactual dependence alone. Specifically, since in my book I assert (unlike Lewis) that counterfactual dependence exists, both backtracking (in time) from effect to cause, and across the horns of some epiphenomenal forks,Footnote ⁴⁹ Schaffer wonders how I can rule such cases out from generating counterintuitive instances of moral responsibility.

Schaffer believes that what I must have “is an in-principle argument to the effect that no case of mere correlation or effectuation could possibly satisfy” my conditions for a counterfactual-based responsibility (415). I deny this, but since I have given detailed examples of both backtracking and epiphenomenal responsibility in my replies to Hitchcock and Ferzan,Footnote ⁵⁰ I will not lengthen this response by recapitulating those examples and argument. The interested reader is referred to my earlier replies.

But Schaffer is correct this far: most cases of counterfactual dependence operating backward through time and across epiphenomenal forks do not result in responsibility for the bad states of affairs so depending on defendant's action. And I do need principles to rule such cases out. Among the conditions of responsibility that Schaffer adds to (1) mere counterfactual dependence are: (2) breach of duty; (3) ability to prevent (in cases of omission); and (4) intention (414–415). Yet these, he notes, are insufficient to rule out cases when intuitively there is no responsibility despite satisfaction of his versions of conditions (1) through (4). Needed, he thinks, is: (5) defendant's act must cause the bad state of affairs. But this, of course, would be to surrender the disjunctive thesis about responsibility.

Of course I will eat no such poisoned fruit as that tendered by Schaffer as (5). Rather, if we frame (3) and (4) carefully, I have all the limitations my view needs or wants. Condition (3) should be framed as follows (for backtracking/epiphenomenal cases): the actor must have the ability (in a compatibilist sense of that word)Footnote ⁵¹ to act in such a way that the earlier or epiphenomenal events for which he is to be held responsible did not happen. And condition (4) for such cases is: defendant must know he has such ability as he acts, and he intends to utilize such ability because he wants the earlier/epiphenomenal event to have occurred (or some effect of that earlier epiphenomenal event to occur).

This is pretty abstract. Apply it to this case, which I call Paralyzed Patriot:Footnote ⁵² Patriot wants to warn the rebels that the British are coming by land and not by sea. By prearrangement, he can do this by flashing one light rather than two in the tower of the Old North Church in Boston. Unfortunately he is completely paralyzed and cannot move his limbs. Fortunately, John-Dylan Haynes and the associates of the late Benjamin Libet have hooked him up with a mind/brain interface machine that reads both the readiness potential (a slow negative shift in electric potential) in the supplementary motor area of the brain, and the blood-flow patterns in the precuneus region of the parietal cortex.

They explain to him that if he wills the movement of his right index finger, the finger will, of course, not move to light the light (he is paralyzed, remember). But if he merely wills the movement, his aforementioned brain processes will have occurred one half to one second prior (in the case of the readiness potential) and seven to ten seconds prior (in the case of the parietal cortex blood-flow patterns) to his conscious willing. Further, they have hooked up readers of these processes, put an amplifier on the readers, and hooked that up to the light on the Old North Church. Will one movement, and the light will go off once; will two movements, and the light will go off twice. Patriot believes all this; it is all true; Patriot intends to warn his fellow rebels by willing the single movement of his paralyzed finger; he wills such single movement for that reason. By my lights, Patriot is responsible for warning the rebels and deserves a reward (if the rebels win) or to be hung (if the British win) for it.

Schaffer's Neighbor-Kingpin-Thug scenario (413–415) fails to generate responsibility because it does not satisfy my version of conditions (1) to (4). Specifically, with regard to condition (1), Schaffer accurately notes that I do not think there is always backtracking counterfactual dependence whenever there is (forward in time, of course) causation. In my book I limit such cases to those where the cause is sufficient Footnote ⁵³ in some suitably nuanced sense of that context-sensitive word. Yet even here, cases ruled out include those of concurrent overdetermination, namely, Kingpin's phone call was sufficient for Neighbor to grumble where, when, and in the way he did, but so was Neighbor's independent hatred of being up that early in the morning; since neither the phone call nor the early wake-up were necessary to the grumble, the latter was sufficient neither for the phone call nor for the early wake-up.

With regard to condition (3), it is not very intuitive to think that Neighbor has the ability to have it be the case that Kingpin makes his phone call, as another of my commentators in this volume, Gideon Yaffe, has observed.Footnote ⁵⁴ But as Yaffe also notes, how to distinguish cases like Schaffer's Neighbor from cases like my Patriot is a subtle matter requiring further exploration.

With regard to condition (4), Neighbor does not believe he has the ability in question, even if in fact he does, nor does Neighbor grumble with the intention of using any such known ability. Thus, on my versions of conditions (1) to (4), there is no responsibility for Neighbor, whereas there is responsibility for Patriot.

Schaffer anticipates that I might squeeze around his intended counterexample(s) with my version of the principles of a counterfactualist responsibility, but he nonetheless worries that even if I were to do so, I would not have made it intelligible why counterfactual dependence without causation can be an independent desert basis (415). My response here is simple: we all, Schaffer included, need to be careful in what we can fairly demand here, for it turns out that respectable philosophical arguments for why some natural property is or is not a basis for moral blame are few and far between. Take causation itself. As I detail in both Chapter 2 of Causation and Responsibility and Chapter 5 of Placing Blame, it is at once astonishing and disappointing how many terrible arguments for and against have been advanced by otherwise good philosophers on the issue of whether causing a harm increases one's blameworthiness, over and against merely trying to cause/risking the causing of such a harm.Footnote ⁵⁵ The honest philosopher makes do with very short argument chains here, perhaps as short as, “this is how it seems to me.”

Schaffer is handicapped in his ability to intuit how responsibility could flow out of mere counterfactual dependence by his acceptance of the counterfactual theory of causation. He thus does not have responsibility for omissions, for example, as an intuitive instance of counterfactualist responsibility, for omissions are all causal in nature for him. If he could get his head out of his merely possible world about causation into the actual world described in my book—if he would adopt the right theory of causation, in other words—I predict that counterfactualist-based responsibility would look quite intelligible to him.

C. Schaffer Needs to Explain Various Items if Absences Can Be Causes

As Schaffer notes, I offer three arguments for why there is no causation in omission, prevention, and double-prevention cases, arguments that I rely on in addition to relying on the ontological argument that absences are nothing and “nothing comes from nothing.”Footnote ⁵⁶ Schaffer makes two untoward moves in his dealing with these arguments in Section V of his article (429). The first is his stubborn refusal to concede, even arguendo, the correctness of my closeness argument discussed above. Thus he dismisses all of my examples as not really contrasting causation with omissions, double preventions, and so on (431). Yet this makes his reply here hostage to the correctness of his criticisms of the closeness maneuver. If he is wrong on the latter point, he apparently has nothing to say in response to my other three arguments. (And by “nothing to say,” let me stipulate that I mean to refer to a true absence—see below.) So I ignore Schaffer's first argument here on grounds of repetition. If I am justified in treating decapitations (and drownings too—same point) as causings both in commonsense ascriptions of causation and in ascribing moral responsibility, as I argue above that I am, then Schaffer's first argument here disappears.

Schaffer's second misstep relates to his second argument against my three arguments for my No Absence thesis. He prefaces his move here by claiming that it is “optional” for him to advance any explanation for the metaphysical and moral differences that I advert to in my three arguments (432). Yet it is not optional, on Schaffer's own epistemological holism. Schaffer rightly sees that when one faces a set of premises leading to a very counterintuitive conclusion, one either swallows the conclusion or gives up one of the premises; in doing this, rationality dictates giving up the least supportable statement, whether it be premise or conclusion. This is the overall structure of Schaffer's argument against me in the body of his paper. Schaffer therefore needs to show why his view (of absence causation) does not lead to the obliteration of distinctions that we cannot afford to see obliterated. This means that he is not free to eschew giving explanations alternative to mine of the phenomena to which my three arguments advert.

Those alternative explanations that he does proffer, Schaffer confines to explanation of only one of the three phenomena to which I advert: the moral phenomenon.Footnote ⁵⁷ The moral phenomenon to be explained is the differential stringency in the duty not to cause some evil, on the one hand, contrasted with the duties to prevent such an evil (omissions), not to prevent the absence of such an evil (preventions), or not to prevent a preventer from preventing such an evil (double preventions), on the other hand. The former duty seems more stringent than the latter duties (holding the evil in each of them constant, of course). The question is: Why should this be so?

I explain the moral difference in terms of the metaphysical difference in these pairs of cases: in the first, there is a cause-based responsibility, but in the second, only a counterfactual-based responsibility; and the thesis is that counterfactual dependence generates less blameworthiness than does causation.

Schaffer by contrast urges two explanations for this moral phenomenon. One is based on my idea that causal relations are scalar, an idea that has attracted much attention in the secondary literature on my book.Footnote ⁵⁸ Schaffer hypothesizes that perhaps cases of omission, prevention, and double prevention, while still causal, are across the board less strongly causal of some evil than normal, connectionist causation would be.

While I find Schaffer's suggestions here ingenuous, I doubt that however scalarity of causation is cashed out—and I confess that my book does not go very far in cashing the notion outFootnote ⁵⁹—it will turn out that omissions, and so on will be lesser causes. Rather, they will be zero causes, that is, noncauses on the causal scale. Certainly that is true for transmission versions of physically reductionist, singularist views of causation, where degrees of causation correspond with degrees of force, energy, and so on; omissions and so on will have zero of such force, energy, and so on, transmitted. While conceptions of scalarity might have greater flexibility under primitivist singularist theories, the one primitivist theory with which I am most familiar—Stephen Mumford's nonreductionist powers theory—explicitly affirms scalarity of causation while denying that omissions, and so on, register anywhere but zero on the causal scale.Footnote ⁶⁰

So I find Schaffer's first proffered explanation not sufficiently articulated yet to take seriously. In fairness, he himself conceded in an earlier draft of his present paper that he has “offered only brief and inconclusive speculations” here. But he feels free to do that only because he regards this entire (alternative explanation) exercise to be optional for him.

I am even less clear about what Schaffer's second suggestion here is intended to be. Schaffer urges that morality could consist of duties not to do certain act types and that the stringency of such duties could vary by these act types; further, that it could just happen to turn out that: (1) the act types that are the subject of duties of lesser stringency are all and only the ones I call omissions, preventions, and double preventions and (2) the act types that are the subject of duties of greater stringency are the ones I call causal (433).

What I find puzzling about Schaffer's second suggestion is its apparent incompleteness. For I could accept all of what is said and then explain this pattern (of comparative stringency of moral duties by act types) in terms of causation versus mere counterfactual dependency. To frame an explanation truly competing with mine, therefore, Schaffer needs to proffer his own natural property that the stringent (duty) act types possess but the less stringent (duty) act types lack. Or he needs to make plausible that there is no such property. I do not see either of such showings in the works, so this is also too poorly developed to assess further.Footnote ⁶¹

It is a pity that Schaffer nowhere includes (in his discussion of the moral phenomena to be explained) any mention of that subclass of double-prevention cases that I call allowings. Take a double-prevention case that Schaffer does mention, the Tally case.Footnote ⁶² Judge Tally did not kill the victim, Ross; rather, he sent a telegram countermanding a warning telegram sent by others to Ross, thus preventing a preventer (the warning telegram) from preventing the killing of Ross by the Skelton boys. To make the case a true allowing, suppose (contrary to fact) that it was Tally himself who had sent the warning telegram to Ross but then thought better of it and sent a second telegram countermanding the delivery of the warning telegram to Ross. Then Tally allowed Ross to die exactly the death he would have died anyway without Tally's (self-canceling) actions.

Most people have the strong sense that cases like my reimagined Tally case instance little or no moral blameworthiness on Tally's part, certainly no more than for pure omissions.Footnote ⁶³ (Indeed, a thousand-year-old tradition in ethics so distinguishes such doings from such allowings, such killings from such lettings-die.) A necessary part of my explanation for this is that Tally is only a double-preventer, not a causer of Ross's death. Schaffer needs an alternative explanation for this moral difference as much as he does for simpler cases of omission.

It is also a pity that Schaffer does not take seriously the other two phenomena on which I base my arguments. I would, in particular, have been pleased to hear his alternative account of the overdetermination phenomenon. That phenomenon is that in cases of ordinary causation, we have clear-cut intuitions separating cases of concurrent overdetermination from cases of preemptive “overdetermination.” Yet in cases of supposed “omissive causation,” we have no such intuitions. Indeed, we do not have a clue how to draw such a line, as I detail in my debate with Richard Wright.Footnote ⁶⁴ To me, this suggests that omissions and so on are noncausal in nature. What other explanation is there?

D. The Semantics and the Ontology of Omissions: “The Problem of Nothingness”

Schaffer is careful and illuminating in his diagnosis of just what it is that we disagree about when it comes to the semantics of “absence talk” and the ontology of absences. Yet despite the carefulness and the sophistication of his treatment here, I think that he overstates our disagreement, misunderstands its significance, and mischaracterizes what disagreement there is. Consider a standard example of absence talk, namely, omission talk (OT):

My semantic premise, as Schaffer calls it, is that the omission talk in statements like OT (“Jones failed to water the plants”) is usually to be interpreted as externally negated, existentially quantified statements:

In English (sort of): it is not the case that there was something such that it was both a watering of the plants and that Jones did it.Footnote ⁶⁵

Schaffer, by contrast, would sometimes interpret such seeming absence talk in one of two ways (418), both of which move the negation inside the scope of the existential quantifier:

There was something such that it either was not a watering of the plant, or it was not done by Jones. Alternatively:

There was something such that, although it was done by Jones, it was not a watering.

The semantic disagreement between us, as I understand it, is how often statements like OT use omission talk my way (“external negation”) versus how often such statements use omission talk Schaffer's way (“internal negation,” wide or narrow scope). Neither of us deny that omission talk can be used in any of these ways in the appropriate contexts. Schaffer's attempt to construe me to deny this last (420) is contrary to what I thought I said.Footnote ⁶⁶ He thus here overstates our disagreement.

He also mischaracterizes it. At one point he says that “Moore and I disagree about what absences actually are” (426). I think Shaffer should rather have said that he and I disagree about what omissions are, or even better, we disagree about how often omission talk really is talk about absences (rather than disguised talk about existent particulars).

Third, the significance of our semantic disagreement is less than may meet the eye in the context of the present debate. To the extent that Schaffer is correct (that seemingly negative statements actually pick out existing particulars), I have no ontological problem in countenancing the existence of such particulars—why would I? I also have no ontological problem in fitting them in as causal relata in the appropriate contexts. It would just turn out that in the instances Schaffer is correct about, there were no omissions; common speech just made it look as though there were. Now consider the cases where my interpretation is correct (that negative statements like that contained in OT refer to no particulars but are instead negative existential statements about the absence of any instance of types): Schaffer's general ontological view is surely that nothing is not something (outside of bad jokes and equally bad existentialist philosophy), and his apparently preferred view about causal relata is that true absences (absences of any instance of a type) cannot play these roles (417, 428–429). So for him, true omissions are not causes either.

So Schaffer and I in fact have only a wishy-washy, “how often” kind of semantic disagreement; it concerns only the intended referent of omission talk, not the question of what absences are; neither of us defends a categorical view of what such referent must always be; and our disagreement here has no significance as to the question of whether absences can enter singular causal relations but only significance in determining how often our omission talk is actually disguised positive-presence causal talk. Now, does anyone care who is right in this semantic dispute? Well, in truth, I do. That is because I want my No Absences thesis to have significance for most of our omission talk, for that makes most of our omission talk about something other than causation. It also does not require us to revolutionize our ethics of what are positive versus what are negative duties in morality—that is, we need not reclassify what we now think of as positive duties requiring us not to omit to do some action into negative duties enjoining us not to do some active causing.

So return to OT. How often do we utter OT meaning either of the things Schaffer proposes? As to the internally negated wide-scope variation, does anyone ever mean this? It is always true (unless the world went totally inert at the time interval in question) that something was going on somewhere, involving someone other than Jones, and that this something was not a watering of the plant in question. Even the most platitudinous of us rarely deliver up such vacuous truths. Schaffer assures us that context of utterance could narrow the range of possible things going on and thus decrease the vacuousness of the statement (422), but he gives no examples.

How about Schaffer's internally negated narrow-scope version? How often does an utterer of OT mean that at the time interval supplied by context, Jones did something, but what he did was not a watering of the plant? I think the correct answer is: “almost never.” For notice that the counterfactual dependence of the plant's death on its nonwatering holds irrespective of whether Jones danced a jig, fed his pigs, or did nothing at all. And speakers typically appreciate this fact. In explanatory contexts where speakers often want to convey and listeners to learn what would have made a difference, why would perspicuous speakers pick out the way in which Jones did not water the plant (he was dancing a jig instead) rather than the generic fact that whatever Jones was doing, there was nothing that he did that was a watering of that plant? Or suppose we are not just explaining with OT why the plant died; suppose we are uttering OT to blame Jones for the plant's death. Do we blame Jones for dancing a jig? Hardly—he might be a terrific dancing waterer! We blame him for not watering the plant, not for anything he was doing at the time.

Once in a while we might use OT to pick out what Jones was doing at the relevant time. Suppose Jones was conversing with Schaffer; Jones was so puzzled by Schaffer's curious arguments about omissions that he entirely forgot that he should now water the plant, and the plant died for lack of water. A speaker who knows all of this might utter OT, meaning Schaffer's internal-negation narrow-scope version. Because now the conversing is explanatory, because now the plant's death counterfactually depends on the conversing (it is my hypothetical, so I can tell you this with certainty—without the distraction, Jones would have remembered to water the plant); and Jones can be blamed for conversing with Schaffer, a known Socrates-like seducer of curious gardeners.

Yet even here (in what is surely a rare and unusual case), OT might more often be uttered with my meaning: what explains why the plant died and what Jones is to blame for is: not watering the plant. When the speaker wants to explain further why Jones failed to water the plant or condemn Jones further for not adopting precautions such as a string on the finger, he might use OT Schaffer's way. But even in this last scenario, the speaker would do better to refer to the conversing with Schaffer with more perspicuous and less ambiguous wording, namely, “conversing with Schaffer.” In that case, instead of uttering OT, he would utter action talk (AT):

Let me now turn to Schaffer's and my supposed metaphysical disagreements about absences. As I say above, I do not think we disagree about what absences—true absences, not negatively described presences—are. They are not particulars but are, as Lewis put it, “bogus entities”—that is, nothing at all. The absence of an act of saving is no more an existent particular than the absence of an elephant is a kind of ghostly elephant. Nor, as seen above, do Schaffer and I disagree about whether our statements sometimes are not about particulars but indeed are “about” absences. Nor does Schaffer (in his most preferred view, at least) disagree with what he calls my “metaphysical premise,” namely, that true absences (the “things” captured by negative existential statements) cannot be causes or effects.

So on what do we disagree? Schaffer thinks it is about whether there is always a “robust metaphysical distinction” within pairs such as dying versus surviving or telling a secret versus keeping a secret (424). I am supposedly committed to the view that one of these must refer to an existent, while the other involves only an absence. And Schaffer finds my supposed “claim of a robust metaphysical distinction implausible” (424). Again, I think that Schaffer overstates our disagreement and mischaracterizes its significance. When I used these and other such pairs, I thought I was arguing against the skeptic who denies that we can ever separate presences from absences.Footnote ⁶⁷ I thought (and think) that this is a daffy view, and so, I gather, does Schaffer. I tried to show how we can easily be confused as to whether a speaker intends to refer to a presence or an absence by a given utterance; but that cuts no ice toward establishing that there is no difference between presences and absences.

For some pairs, my intuition is that J.L. Austin was right, that one of such pairs “wears the trousers” (in Austin's not so politically correct phrase) in the sense that it has sense and reference while the other of the pair takes its entire meaning by mere negation of the first.Footnote ⁶⁸ Health is the absence of illness, or perhaps illness is the absence of health. If there are such contradictory pairs, then for them, Schaffer can sign me up for his “robust metaphysical thesis.” But perhaps there are not, or perhaps, if there are, “death/surviving,” and “telling/keeping a secret” are not among them. In which case why not say just what Schaffer says, namely, each of such pairs refers to some real existent?

Schaffer also wants to find something about which we can disagree by making light of my appeals to science to determine whether dying, surviving, neither, or both, are real existents to be referred to by speakers using these words (425). Yet Schaffer himself is willing “absolutely [to] grant that there is a distinction between actual objects and absent ones, and that there is a distinction between actual events and absent ones. There is a world out there” (425–426). So let us ask Schaffer: how do you tell whether an elephant is present or absent? Or whether a killing took place or did not, so that there was an absence of any killing? How do you decide whether deaths are events but nondeaths (“survivings?”) are not? Do you have some access to these facts other than by the observational methods of science, broadly construed?

So again, there is not much on which Schaffer and I disagree here. Nor do I see much significance to whatever disagreement we may have on these matters. For if both “dying” and “surviving” do each refer to existents, then each is eligible to serve in true causal statements. Such an eventuality does not affect any of the theses or premises that Schaffer accurately enough depicts me as caring about.

B. Yaffe's Substituted Equivalence Thesis

Early on his commentary, Yaffe makes a move that will surely puzzle many of his readers as it puzzles me. He substitutes his own Equivalence thesis for the thesis I examine in my book, which is (1). Yaffe replaces my action verb V on the left-hand side of biconditional (1) above with his “active with respect to a V-ing.” This yields: (2) x is active with respect to a V-ing iff x caused the V-ing to exist. This of course yields Yaffe's distinct versions of my two conditionals (1a) and (1b) above. Respectively: (2a) asserts that causing is necessary to activity-with-respect-to; (2b) asserts that causation is sufficient for such activity.

What is puzzling about this is that Yaffe has literally changed the subject. “Activity with respect to” is his subject, not mine; I was (and am) interested in the relations between actions and causation, not something else. So in an important sense everything Yaffe has to say in the body of his paper passes me by like the proverbial ship in the night.

Yaffe gives two reasons for his change of topic. One is “to avoid confusion.”Footnote ⁷¹ Yet what is the confusion in (1)? It cannot be lack of clarity, at least in comparison to (2) with its opaque phrase, “active with respect to.” The latter phrase is paraphrased by Yaffe at one point into “agent who is involved in the event caused (454).” Does that help? Both phrasings remind me of the late Warren Quinn's attempt to capture the difference in agency between merely foreseeing a result of one's actions and intending it, with the requirement that in the latter case, where harm comes to some victims, there will have been “the agent's deliberately involving them in something in order to further his purpose precisely by way of their being so involved.”Footnote ⁷² In Quinn's case, at least his excuse for involving the obscure notion of “involving” was to replace the somewhat obscure notion of closeness (between harms intended and further harms merely foreseen) that I examine above. Yaffe has no such excuse. Actions are comparatively well-understood items, or at least, so some of us think.Footnote ⁷³

Yaffe's avoidance of confusion motivation therefore cannot be a quest for clarity, substituting as it does an unclear notion for one that is clear enough. Another possibility, one that Yaffe appears to accept, is that actions such as killing, burning, hitting, raping, and so on are in some sense language dependent, while “what the agent is active with respect to . . . is not a matter of language but a matter of metaphysics” (441). Activity is thus a less “confused” concept than action. But not so fast. The only reason Yaffe has (or at least gives) for thinking that actions are language dependent is the famous old statement of Donald Davidson's that Yaffe quotes, to the effect that an act is an event that is intentional under some description (440). This indeed sounds linguistic, in the same way that Elizabeth's Anscombe's well-known thesis that “the term ‘intentional’ has reference to a form of description of events” is linguistic.Footnote ⁷⁴

Yet I thought we all understood that Davidson's remark was to be freed of these holdover influences of ordinary language philosophy. In my action book, for example, I interpret Davidson's remark to mean no more than: an event is an act when but only when that event consists of an intention-causing-bodily movement, when that bodily movement instantiates the type of bodily movement represented in the object of that intention.Footnote ⁷⁵ What Davidson was getting at was that intention is at the root of agency and action, even if he put the point in the locutions of an ordinary language philosophy that he more than anyone else destroyed. So I see no metaphysical leg up for Yaffe's activity over action. Both are equally in the world, ready to do service in explanations and evaluations.

Yaffe's second reason for preferring the “activity” phrasing in (2) to my action phrasing in (1) is that using this formulation will “focus our attention on the normatively relevant phenomenon” (441). Take Herb Morris's old example (picked up by Gil Harman and Yaffe), the verb “alert.”Footnote ⁷⁶ Yaffe correctly recognizes that one can alert the burglar without being responsible for the state of affairs that makes what one is doing an alerting: one could shoot the snake crawling toward one without knowing, intending, or even unreasonably risking the alerting of the burglar; still, it would be an alerting of the burglar. Therefore, Yaffe thinks, the actional alerting is not relevant to responsibility, and we need some richer notion that is (439). Enter Yaffe's “active with respect to.”

Yet cannot something be relevant to responsibility without being sufficient, all by itself, for responsibility? That x acted in a way that alerted the burglar is relevant to responsibility; it shows that x did the wrong thing, that is, he was guilty of wrongdoing, at least prima facie. For x to be responsible and blameworthy, we need to know in addition whether x was culpable in doing the wrong thing.Footnote ⁷⁷ Did he intend to alert the burglar? Know that he would? Unreasonably risk that he would?

Yaffe seemingly wants to build these aspects of culpability into his notion of “active with respect to.” Certainly all of his examples throughout his paper add an intention that represents the result that one is then said to be active with respect to. Moreover, Yaffe needs such activity to require more than action (or else his (2) is no different from (1)); and what else could be added to action but intention or other mental states of culpability that would make Yaffe's activity “normatively relevant,” in his sense of that phrase?

If this is what Yaffe is up to—and I give alternative interpretations of activity below—then we are entitled to ask whether his mode of proceeding does not run roughshod over the important moral distinction between wrongdoing and culpability. What is the gain in lumping both under some more general rubric, such as “activity with respect to?” As the saying goes, when you lump quite different things together, the result is bound to be a bit lumpy. I prefer my desert bases separate: to be responsible for some harm H, one must have wrongfully H-ed, and one must have been culpable (intent, foresight, risk) with respect to H.

So (2) is not only not my thesis; it is also not a thesis that is more worth discussing than my (1). Still, I follow Yaffe in pursuing (2), because at the end of the day he arrives at something that is worth discussing and about which he and I significantly disagree. Thus, think of (2) as a good heuristic—not of interest itself, but it can lead to something interesting. Heuristics, of course, know no criterion but success. (2) works for Yaffe. A cold shower or listening to Celtic Woman would have worked as well for me.

C. Does Action (and Activity) Require Causation?

Let us separate (2) into its two conjuncts, as does Yaffe in his discussion. Take (2a) first: the thesis that one is active with respect to some event e only if one's acts caused e. Here Yaffe and I can join hands, because as he recognizes, my (1a) is true just as his (2a) is true. Still, Yaffe seeks to find disagreement between us in the reasons for which I believe (1a) versus the reasons for which he believes (2a).

Take the case John Gardner raises in his criticism of my Chapter 13 when it was still in article form.Footnote ⁷⁸ Gardner urges that (1a) is nonsensical in cases like that involving the verb “abuse.” (1a) says that x abuses y only if x causes an abusing of y. Gardner objects that there is no “abusing” in the intransitive sense, that is, the sense leaving open whether the injury in question was inflicted by a person. Necessarily, Gardner thought, an abusing of y requires an abuser such as x; for Gardner, injuries produced by storms, accidents, and the like conceptually cannot be abusing—in which case (1a) is trivial: abusings of y are required for actions of abuse to y only because it is analytic of the concept of abusing that abusings require actions of abuse.

My response to Gardner is pretty much as Yaffe describes it.Footnote ⁷⁹ It is twofold. First, I doubt that it is analytic that suffering abuse (“abusing”) requires a human actor in the role of abuser. “I suffered much abuse from the elements while in the lifeboat without food or water for days” strikes me as OK. But second and more important, leave the language for the underlying metaphysics. (1a)—and (2a), for that matter—do not depend on every verb V in English having an intransitive sense that can be used in “V-ing.” What is required is that there be some state of affairs—say, of injury, as Yaffe supposes—that is caused by the agent and without which there can be no act by x whereby x V-ed. “X abused y only if x caused injury to y” sounds pretty good; and if there is anything wrong with it, it is because “injury” is an inapt choice. If “injury” does not stand to “abuse” (verb) as “death” stands to “kill,” then the repair is to find a better analyses of the example, not to throw out (1a).

Yaffe accuses my second answer to Gardner of depending on “fancy and controversial metaphysical claims” (447). Just what these are is not easy to pinpoint. Apparently my second answer commits me to denying that “the relational property of x's being active with respect to it [suffering of abuse] may be essential to it” (447). And I am committed to this denial, Yaffe says, because I must think “that all of the essential properties of events are intrinsic” and that x's being active with respect to the suffering of abuse is not intrinsic (447).

I never raised the question of whether x's being active with respect to the suffering of y is an essential property of y's suffering abuse. My book is about actions, not Yaffe's “active with respect to.” But with respect to Yaffe's concept, I do now deny that suffering abuse has as an essential property not only a human agent abuser (i.e., an actor) but also an intention of that abuser to abuse (i.e., an actor intentional with respect to abuse). (I deny, in other words, that one can suffer abuse only at the hands of one who is “active with respect” to that abuse.) “You are badly abusing him,” “he is suffering much abuse at your hands,” are both appropriate things to say to someone whose acts, unwittingly, are causing another to suffer. Intent of an abuser, like action of abuse by some human agent, is inessential to there being a suffering of abuse. And this is not because of some necessity claim, such as Yaffe's suggestion that only intrinsic properties can be essential properties. Rather, it is because the states of affairs (like suffering injury or abuse) referred to on the right-hand side of (1a) and (2a) do not happen to include any intention to inflict suffering, injury, or abuse among the properties making them the kind of states of affairs they are, anymore than the event type picked out by “big fires” happens to have as one of its properties: “intended by an arsonist.”

D. Is Causing Sufficient for Action (and/or for Activity)?

Yaffe very nicely describes my limited acceptance of (1b) (444–445). While I do go out of my way to defuse many well-known counterexamples to (1b) in my book,Footnote ⁸⁰ I myself restrict (1b) to cases of substantial causal contribution, as Yaffe notes. If x's acts (willed bodily motions) substantially cause y's death, then (but only then) does x kill y. But again, Yaffe would transform what I say—the amended version of (1b)—into something else, an equivalently amended version of (2b). Indeed, Yaffe describes me as being “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” (445). In my book I was not being cagey—I was being silent! Nonetheless, Yaffe eventually attributes to me the amended version of (2b), according to which x is active with respect to e only if x is a substantial cause of e.

Now, it is child's play to show that (2b)—either (2b) classic or (2b) amended—is false. For Yaffe thinks (as do I) that “there is more to being active with respect to another's suffering of abuse than [substantially] causing his injury; additional conditions may need to be satisfied” (448–449). Given the way that Yaffe conceives of being “active with respect to,” that is surely right. Needed is some mental state marking the actor's culpability, and causing e in no way requires intending e, foreseeing e, or unreasonably risking e. So Yaffe pounds into dirt his mythical opponent. Rightly so—that guy is an idiot.

Despite this unseemly torching of a straw man he himself has created, Yaffe actually arrives at a significant issue in the second half of his article. He does this by focusing on one very narrow class of examples: not only those act types that have means restrictions built into them as essential properties but also that subclass of these that Sandy Kadish dubs “nonproxyable” act types.Footnote ⁸¹ The question Yaffe raises is whether (1b) (or 2(b)) can be sustained in this narrow class of cases. It is to that issue that I turn in the next section.

E. Is Substantially Causing Some Event Sufficient to Have Performed a Nonproxyable, Means-Restricted Act Type?

In the cases considered here, we are again focusing on the second conjunct of the biconditionals (1) and (2). In my book I examine (1b); Yaffe examines (2b). We will thus again be talking past each other. But unlike last time, perhaps we can talk enough about the same subject so that we can meaningfully disagree about it.

Yaffe accurately recapitulates my discussion of means restrictions generally and nonproxyable means restrictions particularly. An example of the first is Donald Davidson's: Is the substantial causal contribution of a surgeon's (s) acts to removing (intransitive) of a victim's (v) appendix sufficient for it to be true that s performed the action of removing v's appendix?Footnote ⁸² Or that s was active with respect to removing (intransitive) of v's appendix? If there are means restrictions to acts/activities of removal, the answer is in the negative, as Davidson thought; and (1b) and (2b) would be false.

An example of the second, the special cases of means-restricted nonproxyable act types, is provided by bigamy. Arguably one can perform an act of bigamous marriage only if the person who does the act is the selfsame person as the person who is caused by that act to be married to two different people at the same time. If that is so, then a substantial causal contribution by x to the state whereby one person is married to two others will not be sufficient for it to be true either that x performed the action of bigamy (entering a bigamous marriage); or that x was active with respect to x's being married to two people at the same time. And (1b) and (2b) would be false.

Yaffe very nicely points out my reply to the second sort of case. I urge that careful attention to what one must cause for (1b) to be true shows that (1b) is indeed true in these cases. Of course, it is not sufficient to cause someone to be married to one person while already married to another; (1b) properly formulated asserts rather that it is sufficient (for doing bigamy) to cause oneself to be married to one person while already married to another. As Yaffe accurately puts it, on my view “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” (452). So interpreted, I assert that (1b) is true even of this class of cases.

Yaffe does not respond directly to my defense of (1b) just given. Rather, he seeks to commit me to an analogous defense of (2b). He does this by attributing the following belief to me: I supposedly deny the proposition (call it (ii)): [if x is active with respect to some event e (such as becoming married to one person while one is already married to another), then one has performed the action of bigamy (entering a bigamous marriage)]. Yet I of course said no such thing. Nor do I now think any such thing. Being active with respect to is Yaffe's phrase; in the meaning he earlier stipulated so as to achieve “normative relevance,” being active with respect to includes activity but more besides. Above I construe the “more besides” to be intending, foreseeing, or risking, the mental states of culpability. But whatever the “more besides” is, on Yaffe's conception of being “active with respect to e,” that notion includes doing the action of e-ing. So I do not deny what Yaffe says I deny, not in what I said, not in what I meant, and not in what I now think. Thus when Yaffe complains that “no reason has been given to reject (ii)” by me (452), I am at a loss. I do not have a reason to reject (ii) because I do not in fact reject (ii).

Yaffe is on much more solid ground when he leaves all of this and gets to something I actually do think. Take the example of x kissing y. About such nonproxyable cases, (1b) asserts that if x's acts substantially cause contact between z's lips and y's body, and if x = z, then x kissed y. Yaffe accurately attributes to me the following belief:

With the caveat that it is usually easier to make a substantial causal contribution to lips touching the body of another when the lips are one's own rather than those of a third party, I agree with this attribution of belief to me.

With what does Yaffe here disagree? What he says is this: “The problem . . . is that there is no reason to believe that we are holding fixed the facts about activity when we alter the facts about identity” (453); whereas, Yaffe thinks, there is reason to believe the opposite: “our willingness to apply the relevant verb of action in the one case [where it is one's own lips doing the touching] and not the other [where it is someone else's lips one causes to do the touching] . . . indicates that we sense a difference in activity in the two cases” (453). So here we have Yaffe's troublesome concept again, “active with respect to” (or “activity”).

Let me give two readings of “active with respect to,” one of which engages Yaffe in a genuine disagreement with me whereas the other just has him talking past me again. The first reading of “active with respect to” makes the degree of “activity” proportional to the degree of causal contribution made by some act to some result (like lip contact with the body of another). Yaffe's claim here, then, is that “the facts about identity make a difference to the facts about activity,” that is, to size of causal contribution.

I literally cannot imagine how this last claim of Yaffe's could be true. It is easy enough to see that, generally speaking, one's causal contributions will be larger for actions where one uses only one's own body as the means rather than acting through the body of another. But this is not always the case: imagine the germ-conscious person who uses another to do his kissing for him; he invents and assembles a kissing machine, fastens a paraplegic in it, and uses the machine-cum-paraplegic to touch the lips of the latter to the body of others. I take it that this abstemious kisser is as substantial a causer of lip contact as is the standard kisser.

More importantly, even if, in general, there is a correlation between degrees of causal contribution, on the one hand, and whose body gets used as the medium through which the causal contribution is made, on the other, this will not support Yaffe's claim. For Yaffe's claim is that “the facts about identity” [i.e., whose body gets used] “make a difference” to the facts about degree of causal contribution. These facts about identity are supposed to bleed into the facts about causation; identity (of agent and body used) is supposed to act like some kind of causal aphrodisiac, a causal enhancer or causal magnifier. Given what causation is, this dangerously borders on the unintelligible.

Despite being stuck defending a claim that is only marginally intelligible at best, Yaffe manfully adduces three sorts of evidence in support of it. The first is the linguistic fact to which I also advert in my book: that fact is that we do not typically use the causative verbs of action such as “kill,” “laugh,” “hit,” or “kiss,” to describe agents’ causings of deaths, sounds, or contact with the body of another when those causings are achieved through the behavior of a third party. And Yaffe is right: in my book I dismiss this “evidence” as not evidencing anything very interesting or very deep. Rather, whether a causative verb of action is idiomatically applied to some agential causing depends mostly on the accidents of familiarity and repetition. We are unfamiliar with agents causing avalanches that destroy villages; so no matter how greatly x's acts causally contribute to an avalanche beginning, when that avalanche destroys a village, we do not say, “x avalanched the village.” Yet when we are familiar with a kind of causing agents do with some frequency, we apply the causative verbs of action utterly indifferent to the fact that the causing was done through the body and/or agency of another. “x swindled her out of her money,” applies perfectly well no matter whether x acted alone or acted via a con requiring the participation (innocent or otherwise) of others. Likewise with “x deceived y,” which seems equally applicable when x says something false to y and when x induces z to do or say something that x knows will induce false belief in y.

Yaffe's exposition about his second piece of evidence harks back to the Davidsonian slogan with which Yaffe begins: “All actions are intentional under some description of them.” Recall that the modern construal of this is metaphysical and not (merely) linguistic: all actions have their source in some intention of the agents whose actions they are. Now, Yaffe urges (correctly, I think), all such intentions are self-referential in their content. That is, at the source of all actions must be an intention of the form:

And as Yaffe says, such action-originating intentions must cause act-tokens that instantiate the representations (“A”) that are the content of those intentions.

All well and good, but here comes Yaffe's hogchoker, stated using the example of kissing:

For the vast majority of action verbs in English (including both proxyable and nonproxyable verbs), (4) is obviously false. X can kill y, hit y, burn down y's house, offend y, defame y, and so on, without ever representing himself doing such things. That also seems to be true of kissing: x might think she is licking stamps, pressing her freshly lipsticked mouth against a tissue, leaving a kiss mark on a mirror, and so on, and instead she can actually be causing contact between her lips and y's body, in which case x is kissing y yet with no representation of herself “making such contact with,” as (4) requires.

The only verbs for which (4) is plausible are those that Elizabeth Anscombe dubs “intention-drenched” verbsFootnote ⁸³ and that I call “intentionally complex action descriptions.”Footnote ⁸⁴ Arguably at least, I cannot fish without intending to catch something, hide without intending to conceal myself, rape someone without intending penetration, and so on. Some such intention-drenched verbs may be nonproxyable, but most are not; again, the distinction is not marked by this (intention-drenched) feature.

So where does Yaffe go wrong in thinking (4) to be true? He confuses a true claim—all actions are sourced in self-referential intentions—with the false claim that such self-referential intentionality must extend not only to the most basic things one knows how to do (so-called “basic acts”) but also to the more extended results of those basic acts, results such as lip contact on the body of another. The only way (4) could be true would be if it were amended into (4b):

Yet (4b) shifts the meaning of “active with respect to” away from “degree of causal contribution” and back to the culpability reading of “intended, foreseen, or consciously risked.”

This culpability reading is the second interpretation of Yaffe's objection to me mentioned above. If active with respect to is again to mean “intended, foreseen, or consciously risked,” then of course (4b) is true (because activity is defined in such a way that 4(b) has to be true). But this is also irrelevant to any meaningful disagreement with me. So Yaffe needs not to go this route.

Yaffe's third piece of evidence stems, he thinks, from John Perry's solution to the apparent self-reference paradox in Locke's memory-based theory of personal identity (454–455). Yaffe has to stretch for this analogy, but he need not have reached so far, because he could make his point perfectly well without it. His point is the familiar (and correct) one that intentions are the source of actions only when those intentions cause the bodily movements partly constitutive of such actions, in the right way. Alvin Goldman introduces this notion in 1970 to deal with so-called deviant causal claims and the apparent counterexamples that such cases present to causal theories of action.Footnote ⁸⁵

Yaffe wants both to extend and precisify Goldman's requirement. The extension is from the causal relation existing between intention and bodily movement executing the intention (Goldman's concern), to the causal relation existing between that bodily movement and some further event in the world, such as lip contact, death, or injury. Yaffe urges that there are deviant causal chains here, too, and that such chains must be nondeviant if they can constitute being active with respect to such results (where activity is used in its causal sense, not its culpable-mental-state sense). In a general way this extension is surely warranted. Indeed, the lawyers’ notions of proximate causation exist precisely to rule out such deviant causal chains, as does my own “substantial causal contribution” requirement.

The precisification Yaffe desires is what should be viewed as a per se rule of deviancy (nonproximateness, insubstantiality) of this second causal link: the “right way” for bodily movements to cause lip contacts, and so on, excludes using third-party bodies or agencies as causal intermediaries. Yaffe in effect recreates, as a supposed implication of the philosophy of action, a kind of doctrine of intervening causation that lawyers have applied for centuries and that I seek to demolish in chapters 11 to 13 of my book.

Yaffe makes no argument in favor of thinking that an agent substantially causing some event e cannot be an agent causing e through use of the body or agency of another. Perhaps Yaffe thinks this to be so obvious that it needs no argument. The only way he could think this is if he confuses the second causal link in actions with the first. For with regard to the causal link between intending to do the most basic thing we know how to do and the bodily movement events that partly constitute such basic acts, it is plausible to deny that there can be (as a causal intermediary) an act or body of another person. If there were such an intermediary, there must have been something even more basic we were trying to do, and that more basic act would cause the other agent or body to behave as we want.Footnote ⁸⁶

Yet the plausibility of Yaffe's intervening cause thesis evaporates entirely when he moves from link one to link two. We cause remote events through use of others’ bodies and agency all the time. I argue this at length in chapters 12 to 13 of my book.

F. Yaffe on Accomplice Liability

Yaffe's conclusions about accomplice liability depend on his metaphysical conclusions in the last section. If his metaphysical conclusions are as shaky as I think they are, then his legal conclusions are also in trouble.

One conclusion Yaffe reaches is that those who culpably aid others in the latter's entering of bigamous marriages are worthy of criminal liability (457). I agree. The question is whether the law needs Yaffe's non-cause-based, accomplice liability to reach that result. Yaffe says yes, I say no. Secondly, Yaffe's says my “no” commits me to a form of strict liability (458). I deny it, so issue is joined.

On the first issue between us, Yaffe could be correct only if his reinvention of intervening causation could be sustained, that is, only if an agent causing some event e through the use of the body or agency of another was impossible. But it is not, as shown above. We can substantially and nondeviantly cause deaths, deceptions, injuries, lip contacts, and so on through all kinds of instrumentalities, other people included. So the law has no need of some noncausal accomplice theory of criminal liability, just as I argue in my book.

On the second issue between us, Yaffe can accuse me of imposing strict liability only if he changes yet again the meaning of his invented phrase, “active with respect to.” Yaffe says that “in addition to causation, activity with respect to the relevant events requires identity with the person who says ‘I do’ and is doubly married” (457). Liability imposed where there is no such identity is liability imposed in the absence of activity, and that, Yaffe concludes, is strict liability for what happens to us rather than for exercise of our agency.

Now parse again Yaffe's “active with respect to.” Does he mean: (1) the causal reading of activity, that is, there can be no substantial or nondeviant causing of some bad result by an agent if he does such causing through the body or agency of another? If I am right in my argument above, Yaffe's requiring identity for activity (in this sense) is false. Or does Yaffe mean: (2) the culpable-mental-state reading of activity, that is, there can be no culpable mental state causative of some bad result e if the mental state in question includes in its content a representation of someone else's body or agency being the immediate cause of e? This, too, I argue above is surely false. Or does Yaffe mean: (3) a new, identity reading of activity, that is, in addition to substantially and nondeviantly causing e, and (perhaps) in addition to intending to cause e through one's own body only and not through use of the body of another person, being active with respect to e requires an identity between the agent doing the intending and the causing, on the one hand, and the person whose body it is through which the causing is done?

I think Yaffe thinks he has established (3), yet the only arguments he presents are those that go through causation (1) or intention (2). And neither of those work. Of course, Yaffe can stipulate that what he means by activity is now (3); but that is not victory in any but the most empty sense.