A. Thought Crimes

In the book, I mention that in the earliest attempt cases, judges noted that it would be abhorrent for them to punish the defendant for his thoughts alone before explaining that they are punishing him instead for the conduct in which he engaged in furtherance of those thoughts. And I then go on to offer a rationale for the “means requirement,” the principle according to which guilt for an attempt requires proof that the defendant engaged in an action as a means to the satisfaction of his criminal intention. The means requirement, I assume, in line with the early judicial pronouncements about it, protects defendants charged with attempt from punishment for thought alone. Bratman observes, however, that the transfer principle—according to which an attempt is legitimately criminalized if the corresponding completed crime is—implies that an attempt can be legitimately criminalized even if it does not involve, as Bratman puts it, “an action in the public space,” much less one that is performed as a means to the satisfaction of the defendant's criminal intention. And so Bratman worries that these two principles are inconsistent with one another.

In response, I start by noting that the punishment of thought alone is different from the criminalization of thought alone. It could be that thoughts are justifiably criminalized but not justifiably punished. It is, in fact, common for justified criminalization and justified punishment to come apart. As I note in the book, it is unjustified to punish those who are immune, for instance, despite the fact that they have engaged in criminal conduct. A foreign diplomat who drives drunk in Washington, DC, for instance, has committed a crime for which she is not justifiably punished, given immunity agreements between the United States and other countries. Maybe we each have something like immunity for that which takes place only in our heads. The hacker, in Bratman's example, who is doing mental work in furtherance of her intention to commit theft, then, might have committed the crime of attempted theft, a crime for which we are unjustified in punishing her.

In fact, when it comes to thought crimes, something like this must be right. To see why, imagine that D tries and fails to steal a diamond ring, and imagine that D is seven years old. The criminal punishment of seven-year-olds is entirely unjustified, no matter what they have done. Let us raise this obvious fact to the status of a principle and confine it to attempts. According to the resulting toy principle, let us call it the “age principle,” a person is guilty of a criminal attempt only if he is (much) older than seven. Is the age principle incompatible with the transfer principle? One might follow Bratman's reasoning and conclude that it is. After all, the transfer principle allows the criminalization of all attempts of justifiably criminalized conduct and not just those that have a particular property, such as that they are committed by someone over seven. But the seven-year-old's attempted theft is not legitimately criminalized, despite the fact that theft is. The right response to this line of argument is that the seven-year-old's behavior is legitimately criminalized, it is just not legitimately punished. A similar point applies, we might say, to attempts that do not involve any action performed as a means to doing as one intends.

In fact, the point here just illustrates a fact about the different objects of criminalization and punishment. Types of action are criminalized; token acts are punished. To say that a person's act is a crime is to say that it belongs to a criminalized type; to say that it is justifiably punished is to imply that additional conditions are met or that potentially undermining conditions are absent beyond those that qualify it as a member of the criminalized type. So what justifies criminalization are facts about the type or, equivalently, facts about properties shared by all of the tokens of the type. But the tokens can differ from others of the type in ways that undermine the justifiable punishment of them, even when the type is justifiably criminalized.

If this is right, however, then the means requirement does not in fact protect defendants from being punished for thoughts alone. Whatever grounds the immunity-like status to what goes on in our heads, rather, is what provides that protection. The hacker, after all, satisfies the means requirement since she has indeed performed some acts—mental acts—as means to doing as she intends. It is also true, however, that frequently the defendant cannot appeal to the immunity-like status of his thoughts in his defense, for it is proven that he has engaged in an act in furtherance of his intention that is not entirely in his head but instead has entered the “public space.” That is, it is often the case that we can point to that in virtue of which the defendant satisfies the means requirement—namely, a “public” act in furtherance of his criminal intention—to demonstrate that we are not punishing him for thoughts alone. So while the means requirement does not place a constraint on the state that ensures it will not punish for thought alone, as I assumed when I wrote the book, it still helps.

In the book, I argue that the means requirement is justified for evidential reasons. In the absence of an act performed as a means to doing as intended, we lack adequate evidence, I claim, that the defendant's intention is guiding him, in the sense of “guidance” that is necessary for attempt under the guiding commitment view. So, as I see it, there is a very important difference between Bratman's hacker and the protagonist of another one of Bratman's examples, namely the paralyzed assassin. The paralyzed assassin has, by hypothesis, tried to squeeze the trigger but has failed because he is paralyzed. In contrast to the hacker, the paralyzed assassin has not performed any act as a means to doing as he intends, namely, killing somebody, because, we can assume, while his intention caused something, what it caused did not rise to the level of action. Or, to put it another way, the only act that he has performed is the act of trying to pull the trigger. But he has not performed any act as a means to pulling the trigger nor to killing anyone. The result is that he does not satisfy the means requirement, in contrast to the hacker, who does.

Now, in a philosophical example we can stipulate anything that we like, as, in this example, it is stipulated that the paralyzed assassin is trying to pull the trigger and trying to kill. With those things stipulated, it is also stipulated that the paralyzed assassin's intention is guiding him when it exerts a causal influence over him. But the world does not come with facts stipulated. And so we need to ask what the difference is between a paralyzed person who is guided by his intention to kill and another whose intention has the same effects but is not guiding him. Say, for instance, that an intention to pull a trigger, when it guides, causes various forms of preparatory neural activity before causing a muscle contraction and a subsequent finger movement. And let us assume, in line with some plausible theories of action individuation, that the first action here is the finger movement. If the causal process halts before that point, there has been guidance by the intention, but there has been no action. What would lead us to conclude that the causation of the relevant neural activity by the intention, in a given example and in the absence of anything else, is an instance of guidance by the intention?

To see such causal influence as guidance, we need to see it as explicable by virtue of reasons. It must be the case that in the absence of obstacles, those reasons will sustain the causal process all the way to the world coming to match the intention. There needs to be something that has happened that is rationally explicable and otherwise inexplicable, and rationally explicable by virtue of its contribution to a causal process through which the world will come to match the intention. But in the case of the paralyzed assassin, we are barred from describing the relevant neural activity as a means to anything, because it is not even action. It is certainly not believed by the agent to be a means; the agent may not even know it is taking place. We are barred from describing it in the very terms that it must be described in if the causal influence of the intention is that of guidance. For all we know, that is, the intention's effect is not one of guidance at all. For all we know, the removal of obstacles would result, for instance, in the expression of the intention—in, perhaps, a confession of it—or in an increase in blood pressure, or a feeling of embarrassment, rather than in the performance of conduct that is a means to the satisfaction of the intention. Again, we can stipulate that in causing the relevant neural activity, the intention is guiding. But to make such a stipulation is to insist on a fact that is not present in what has happened at that point, for there has been no act performed as a means, and so there has been nothing that can be explained only if the causal influence of the intention is of the rational, guiding variety.

So, it seems to me, as it did when I wrote the book, that acts performed as means play an essential and nonreplaceable role in evidencing a condition that is essential to trying, namely, guidance by intention, in contrast to other forms of causal influence that intentions can have on the agents possessing them. It is further true, as in the hacker example, that such evidence could be provided by thoughts, provided that those thoughts are also actions performed as means to the satisfaction of criminal intention.

One of the important implications of this discussion is that while it is abhorrent for the state to punish people merely for their thoughts, a fact that attempt law is consistent with despite failing to encode an explanation for it, attempt law may rest on the claim that it is not abhorrent but in fact could be fully justified to criminalize thoughts even when they have not given rise to actions “in the public space,” provided that at least some of those thoughts are acts. After all, attempt law, I hold, rests on the transfer principle, and Bratman is correct, I think, in his contention that the transfer principle warrants the criminalization of tryings that are all in the head. In fact, even the evidential restriction on the punishment of such tryings that is provided by the means requirement does not rule out such attempts. Is this problematic? Is it just as bad to criminalize thoughts as it is to punish them?

Notice that for all that has been said, it could be entirely unacceptable to criminalize thoughts for nonderivative reasons. Attempted murder should be criminal because murder should be criminal. Attempted murders that consist of nothing but thoughts should be also criminal because murder should be criminal. That is, the justification of the criminalization of anything that is criminalized under the transfer principle is derivative: it is justifiably criminalized because something else, namely, completion, is justifiably criminalized. We might think that the nonderivatively criminalized forms of conduct, such as murder, cannot consist of nothing but thought if they are to be justifiably criminalized. Nothing said here in response to Bratman requires the denial of this. All that the argument offered here requires is that the criminalization of attempts consisting of nothing but thoughts does not run afoul of whatever principles exclude the nonderivative criminalization of thoughts nor those principles that exclude the punishment of thought.

If there is a lingering sense that there is something wrong when the state criminalizes pure thoughts, even when some of those thoughts are acts and even when what justifies their criminalization is the fact that something that is not pure thought is justifiably criminalized, perhaps that uneasiness can be quelled by reminding ourselves of what a person like Bratman's hacker has done: she has tried to commit theft. While there is a meaningful sense in which that act is not “in the public space”—it is all in her head—there is another important sense in which it is, for theft is in the public space, and what the hacker has done is criminal, even if not punishable, precisely because it is an attempt of that; it is an attempt to invade the legally protected interests of others, and in this respect it is just like other attempts that do involve conduct in the public space and that are, because they, too, are attempted thefts, justifiably criminalized. The intuition that these remarks are meant to invoke is just the intuition that animates the transfer principle. The legitimate criminalization of theft transfers to attempts of theft, it seems, even when such attempts consist of nothing but thought.

B. Types of Act Criminalized under the Transfer Principle

As Bratman notes, the argument for the transfer principle that I offer in the book appeals to the following complex property possessed by some act types: all of the tokens are legitimately censured, and some are legitimately sanctioned.Footnote ¹ The argument for the transfer principle proceeds like so:

1. (1) A type of conduct is legitimately criminalized if and only if all of the tokens are legitimately censured, and some are legitimately sanctioned.Footnote ²

2. (2) If all of the tokens of C are legitimately censured, then all of the tokens of attempts to commit C are legitimately censured.

3. (3) If some of the tokens of C are legitimately sanctioned, then some of the tokens of attempts to commit C are legitimately sanctioned.

4. (4) So if C is legitimately criminalized, then attempts to commit C are legitimately criminalized.

Bratman raises a worry about premise (3). The worry is that it could be that some of the tokens of attempts to commit C are legitimately sanctioned not because they are tokens of that type but because they are tokens of a narrower type, such as, for instance, attempts that involve conduct “in the public space.” Although Bratman links his worry here to his remarks about thought crimes, it need not be so linked. I argue, for instance, for the claim that some tokens of attempts to C are appropriately sanctioned by citing the fact that some get arbitrarily close to completion. One might have the worry, parallel to Bratman's, that they are appropriately sanctioned not because they are attempts per se but because they are attempts that get arbitrarily close to completion. If that is right, then I do not provide an adequate argument for the transfer principle but only instead for a restricted principle to the effect that if C is legitimately criminalized, then attempts that get arbitrarily close to completion are legitimately criminalized. In general, if the attempts that are properly sanctioned are so thanks to the fact that they possess some property that is not possessed by attempts that are not properly sanctioned, then it appears that my argument supports not the transfer principle but instead a weaker principle justifying the criminalization of only the attempts that possess that property.

It seems to me, however, that I can respond to Bratman's concern not by drawing on ideas developed in my argument for the transfer principle but by drawing instead on an idea that plays a prominent role in a much later discussion in the book: the discussion of the fairness of giving lesser punishments to attempts than to corresponding completed crimes. The idea I have in mind is that an important function of sanction is the expression of censure. The state censures defendants by sanctioning them. We censure conduct under a description. We do not censure someone for moving his finger. We censure him for killing someone else (by moving his finger in pulling a trigger). So to say that a token act is worthy both of censure and of sanction is to imply that it is appropriate for it to be censured for belonging to a particular type by being sanctioned. That is, the implication is that the type in virtue of which it is appropriately censured is also the type in virtue of which it is appropriately sanctioned. This is implied provided that the sanction is appropriate as an expression of the appropriate form of censure, namely, censure for being of a particular type.

So we might put the idea of the argument for the transfer principle like this: if all of the tokens of C are legitimately censured, then so are the token attempts to C. And if this censure is sometimes appropriately expressed through the sanction of some tokens of C, then this very same censure is sometimes appropriately expressed through the sanction of some of the token attempts to C. The attempts to C are appropriately censured because they are attempts; but then it further follows that at least sometimes the censure of them for being attempts is appropriately expressed through their sanction by the state. Although it might be that it is appropriate to sanction only those attempts that possess a particular property not shared with those that are not appropriately sanctioned—such as being arbitrarily close to completion or as involving an act in “public space”—they are still appropriately sanctioned because they are attempts and not because they are attempts possessing this further property. The proof is that the sanction is appropriate in part because it expresses deserved censure; and the censure is deserved because the conduct is an attempt to commit a crime, and in particular a crime that is deserving of censure.

What is offered here is an amendment to premise (1) in the argument for the transfer principle just rehearsed. What is necessary and sufficient for the justifiable criminalization of a type of conduct, that is, is that all the tokens are appropriately censured thanks to being members of the type, and for some tokens this censure is appropriately expressed through state sanction. So understood, premise (3), when appropriately adjusted to fit with this adjusted version of premise (1), is somewhat more difficult to defend than I thought when I wrote the book. There is something intuitive about the idea that those who get arbitrarily close to completion of a justifiably sanctioned act are justifiably sanctioned. But are they justifiably sanctioned as an expression of the censure that they deserve, or are they instead justifiably sanctioned for other reasons, such as the fact that they are very close to completion?

I think that the former is true, but I have little more than plausibility arguments to offer in its support. Still, plausibility arguments are better than nothing. And it does seem plausible. After all, we are assuming that the corresponding completed crime, to which the attempt gets very close, is appropriately sanctioned as an expression of the censure it deserves in virtue of being of the prohibited type. But the attempt deserves the very same censure as the completed crime does, and so it is hard to see why sanction would not be appropriately issued in expression of the censure that it deserves, given how close it is to completion. Closeness to completion, that is, seems to defeat a set of defeaters, namely those features of an act that undermine the appropriateness of sanction, such as the fact that it takes place only in the defendant's head. But as I say, this is little more than a consideration in light of which my view is plausible rather than a full defense of it.

A. Degrees of Attempt and Abandonment

As an intuitive matter, trying admits of degree. In fact, trying seems to admit of degree along at least two separate dimensions. First, people try harder or less hard—that is, they can expend more or less effort or energy when trying to do something or can encounter greater or lesser internal resistance when trying. And second, people can do more or less in trying to act in a particular way or can fall short of doing what they are trying to do with less or more left to do. Brink is particularly concerned with the second sense in which trying admits of degree. He thinks that an account of the sense in which one can have done more or less in trying to act—an account of trying under which there can be greater or lesser tryings, that is—will have implications for the bearing of abandonment on moral and criminal responsibility for trying. As Brink suggests, it might end up following from the relevant account of trying-by-degree, in the second sense, that agents who abandon early are worthy of less blame or censure than late abandoners. And this in turn will bear on the question of how, exactly, the criminal law ought to respond to a defendant who has abandoned an effort to commit a crime.

As Brink notes, this is all a bit sketchy, but it is also interesting and well worth exploring further. For reasons that I try to explain here, I believe that what I say in the book about abandonment can account for the sense in which early abandonment seems to count more in the agent's favor, when it comes to blame and censure, than does late abandonment. This does not follow, however, from any account of the way in which trying admits of degree, and so I suspect that Brink will not be satisfied with it.

Brink notes that the account of trying that I offer in the book—the guiding commitment view—does not in itself imply that trying admits of degree in either of the senses just described. Still, the guiding commitment view is not incompatible with things that one might want to say about that. There could be more or less energy expended, or more or less internal resistance encountered, for instance, when one is guided by an intention. And so, even under the guiding commitment view, a person could be trying more or less hard; tryings could vary by degree in the first sense mentioned above. And similarly, given an appropriate metric, a person could have done more or less or have more or less left to do when being guided by an intention. And so the guiding commitment view could be coupled with an account of the sense in which trying admits of degree in the sense that Brink thinks matters to the normative import of abandonment. Under the guiding commitment view, that is, tryings differ from one another by degree in several different respects, and it could be that some of those respects correspond exactly or roughly to the intuitive senses in which trying admits of degree.

In connection with an account of the act element of attempt, in the book I discuss what I call the “geographical metaphor.” To think of things under the geographical metaphor is to think of the criminal intention and the intended completed crime as dots on different places on a map. The agent who is trying to perform an intended act is following (what she takes to be) a route from the one to the other. From this point of view, the question that is answered by an account of the act element of attempt is how far along the road from intention to completion the defendant needs to have gone for guilt for attempt, or how far she needs to travel and still be guilty. In the end, I suggest, the only meaningful sense that can be given to this metaphor is this: the agent has gone far enough if, in the absence of obstacles, she would go all the way to the end.

However, for reasons that I try to explain, we are not to imagine away all obstacles in thinking this through; we are to imagine away only those that undermine ability or opportunity or prompt change of mind or execution failure. Therefore, if the agent is her own obstacle, as it were, if she would fail because she is irrational or because she is ill equipped to recognize what she needs to do to act as planned, then she has not done “enough”—or, better, she has not done what she needs to do—to be guilty of an attempt. The talk of “distance” and “travel” in the geographical metaphor, that is, needs to be unpacked in the language of relevant and irrelevant obstacles to or defeaters of success. So unpacked, the metaphor is not really of much use.

Under the picture that I present, if the temporal location of abandonment—whether it is early or late—makes any difference to responsibility, blameworthiness, or worthiness to be censured by the state, it is for reasons that are independent from anything about whether the agent has done enough for guilt for attempt. In determining whether the defendant has gone “far enough,” we imagine away future change of mind or abandonment of the plan, among other things, and ask whether she would have “traveled” all the way to the end.

Since we are imagining actual or hypothetical abandonment away, the answer to this question does not vary depending on whether the abandonment is early or late. Roughly, the rationale for imagining away abandonment that happens or would have happened after the conduct that is proposed as sufficient for guilt for the act in attempt is this: say that two people are duplicates in conduct, effects of conduct, mental state, and circumstances. And say they are both arrested at time t for the attempt of crime C. Since they are duplicates in all of these various respects, they must have been trying to do the same things. How could what you are trying to do at t depend on anything other than the facts up to and including t? But say that the first person would have changed his mind had he not been arrested, and the second would not have. What follows is that unactualized, future change of mind makes no difference to what one is trying to do. And so it cannot make a difference to whether there has been commission of an attempted crime. Nor can it make any difference to whether there is evidence of commission of an attempt.

Similar lines of thought support disregarding actual change of mind in assessing whether a person was guided by a criminal intention and so was trying to commit a crime. Say one of two otherwise identical agents changed his mind at t1. Still, what he was doing prior to t1 cannot differ from what the otherwise identical agent was trying to do prior to that time. For our purposes here, the important result is that to explain the sense in which the temporal location of abandonment matters, we need to draw on additional resources beyond those that I employ in the book's account of the act element of attempt.

However, I believe that what I say in the book about the mitigating force of abandonment can explain the sense in which abandonment's temporal location is normatively important. Early abandonment is, ceteris paribus, a greater mitigating factor than is late, under that account. (As we are about to see, however, quite a lot is worked into that “ceteris paribus” clause.) In the book, I suggest that the reason that abandonment mitigates is that among the reasons for issuing a particular sanction S for an attempt is this: had the defendant prospectively recognized that she would suffer S in response to her completion of the crime, she would have recognized sufficient reason to refrain from the crime. Since, by default, if there are just as good reasons to inflict one sanction as another, we should inflict the smaller, this reason supports inflicting the lowest sanction such that, had the person anticipated it, she would have recognized sufficient reason to refrain. Since the person who abandons for good reason rather than, for instance, just to avoid detection has already recognized sufficient reason to refrain, without anticipation of any sanction for performance, much less sanction S, this reason supporting the infliction of S is canceled. A fact that cancels a reason to sanction mitigates. And so abandonment is a mitigating factor.

Given an empirical assumption, this reasoning will support the claim that abandonment is a greater mitigating factor when it comes early. To see this, first note the following sense in which a factor can be a greater mitigating factor than another: say that F1 cancels a reason for sanction S1, and F2 cancels a reason for sanction S2, and say that neither cancels a reason favoring any lower sanction than S1 or S2, respectively. If S1 is lower than S2, then, in a sense, although probably not the only sense, F1 is a greater mitigating factor than F2. The idea, in short, is that a factor is a greater mitigating factor than another if it forces us to look for a lower sanction that there is sufficient reason to issue. F1 gives us reason to search for a lower sanction than S1 that we have sufficient reason to issue, but F2 does not undermine whatever reasons there are to issue S1, since S1 is lower than S2, and so F2 does not give us reason to search for a lower sanction than S1. A defendant is better off, in the sense that he has an argument for a lower sanction, that is, if he can cite F1 in his favor than if he can only cite F2.

Now, let F1 be early abandonment and F2 be late. D1 abandons early, D2 abandons late, and both were guided by an intention to commit the same crime; they are otherwise the same. And let us assume that the later the defendant abandons, the more resources he has invested in performance. Then it would seem to follow that D2 failed to recognize sufficient reason to abandon until he had invested more than D1. And so it would seem to follow that he does not weigh the reasons against the crime as heavily as D1 does. It would seem to follow that a lower sanction would have prompted D1 to recognize sufficient reason to refrain at the outset than would have prompted such recognition by D2. Had either defendant anticipated being sanctioned in the precise amount that he in fact invested prior to abandonment, he would have abandoned. But since D1 invested less, a lesser sanction would have prompted abandonment at the outset. So D1's abandonment is a greater mitigating factor than D2’s.

There are many simplifying and potentially distorting assumptions here. Most notable is the assumption that the amount of resources one sinks into an enterprise indicates a lesser willingness to abandon it. While this is true in many cases, it is not true in all. Some people might be prompted to recognize reasons to abandon that have nothing whatsoever to do with how much they have invested in the enterprise. In fact, some of the most sympathetic abandoners are like this. Consider the would-be Palestinian suicide bomber who is so struck by the idea of killing innocent children when she sees them face to face that she gives up the plan. Her reasons for abandonment have nothing to do with the quantity of resources she has sunk into the enterprise. However, she is not less sympathetic having abandoned late than she would have been had she abandoned early. So this example shows not that there is something inadequate with the account just offered of the significance of early rather than late abandonment but rather that the account does not imply and should not imply that late abandonment is in general a lesser mitigating factor. We want, that is, to account for the intuition that early abandonment is better than late only in the conditions in which it is true. It is true when how much the defendant has done tells us something about what it would take to prompt him to recognize sufficient reason to refrain.

I believe that Brink will be dissatisfied with what I say here in support of the idea that the temporal location of abandonment can matter, that early can count more in a person's favor than late. The reason is that he thinks early abandonment is better because there has been a lesser trying, and he thinks the less one tries to engage in wrongful or criminal conduct, the better. But I am doubtful that blame or censure is appropriately modulated by the amount of trying all by itself. Of course, it is hard to say for certain in the absence of a particular account of the metric of “more” and “less” trying. So, consider what Brink suggests in this regard. He seems to have in mind a tree format for representing a person's action plan, with nodes and branches. At each of the nodes the agent has information, and the branches represent what the agent will do in light of the information he has. At the store, if the tomatoes look good, I will buy ingredients for a caprese salad; if not, I will buy the ingredients for a caesar. New node: I am buying the caesar ingredients, and the store has whole anchovies, then I buy them; if only paste, then I buy that. On a parallel node: I am buying caprese ingredients, and the store has burrata, then I buy that; if not, then mozzarella. Et cetera.

To see how much the person has tried to do something, Brink suggests, we need to measure both how many nodes and branches he has passed through and also the significance of those nodes and branches; some matter more than others. There is a lot of detail that would need to be filled in here to develop a full account. What constraints are there on the drawing of nodes and branches? What makes one branch or node more significant than another? But still it is possible to have a picture of the kind of thing Brink has in mind.

Now, notice that it is certainly the case that some people, while trying to commit a crime, engage in significant forms of wrongdoing along the way—they engage in forms of conduct that are wrongful quite independently of the fact that when the agent engages in them, she is trying to commit (another) crime. The person who pistol-whips the teller while trying to steal the bank's money has done something wrongful and criminal. To know that she was doing something wrong when she pistol-whipped the teller, we do not need to know that in doing so, she was trying to rob the bank. Of course, it would have been better if she had abandoned her plan to rob the bank before she pistol-whipped the teller rather than after, but what is significant about early abandonment in such a case is not that there is less trying but that nobody gets pistol-whipped. At least some of the intuition that if you are trying to do something wrongful, it is better to engage in less trying than more probably derives from the assumption that you will do some things wrongful in themselves on the way to trying to do something wrongful.

But this need not be the case. The hard cases for Brink are those in which the only wrongful act in which the agent has engaged prior to completion is the act of trying to commit the crime. In such a case, is less trying better than more? I cannot see any reason why we should think so. Consider two agents with the same action plan. The first has traversed four nodes and three branches; the second has traversed all of those plus another node and another branch. Unless the second has engaged in a form of wrongdoing that the first has not engaged in, I cannot see how, from this description alone, there is any reason to distinguish them morally. We might add to the description. We might explain that having sunk more into his plan, the second agent shows himself to be less willing to give it up in the face of reasons to do so. But then what is doing the work in morally distinguishing the agents is not the difference in the amount they have tried but a difference of a kind that my account of abandonment's significance accommodates. What matters is not how much they have tried. What matters is something else that, as a contingent matter, is perhaps conjoined with that. The result: the degree of trying does not seem to be of normative significance in itself.

I should confess that I am not certain that this reply to Brink is satisfactory. There is something right about the idea that, assuming that you are going to try to do something wrongful, it is better to try less than more. Whether I here provide an adequate account of what is right about that is not entirely clear to me.

B. Commission, Guilt, Censure, Sanction, and Luck

In his remarks about my discussion of the censure-sanction distinction and the related claim that while there is no censure-luck there is sanction-luck, Brink argues that I am committed to some false claims. Let me disentangle two lines of objection in Brink's remarks. First, he says that I am committed to the claim that while a legally insane person is properly censured, he is not properly sanctioned. And then Brink points out that this is problematic, since excuses such as insanity typically block the appropriateness of censure by the state, just as they typically block the appropriateness of moral blame. Second, Brink suggests that I provide an inadequate defense of the claim that there is sanction-luck because I do not explain how consequentialist considerations can legitimately affect what sanction is appropriate without also legitimately affecting what degree of censure is appropriate. The first of these concerns is, I believe, rooted in a misunderstanding. The second, however, is a just criticism, although I think more can be said in defense of my position than I say in the book. I take each of these points in turn.

Why does Brink think that I am committed to the claim that an excused defendant is still worthy of censure but not of sanction? In fact, I imply the opposite, writing, “[C]onsider an excused or justified act that is justifiably criminalized. . . . Such conduct might not be deservedly censured—justified conduct never is, and excused conduct rarely is—and it is certainly not legitimately sanctioned; but this is consistent with its being criminal” (32).

Of course, sometimes one is committed to things that one explicitly denies. Perhaps that is the case here. This remark is made in the midst of a (probably excessively) terse discussion of the relationship between criminalized conduct and punishable conduct. My idea was that the censure-sanction distinction could help us to understand the relationship between the two. The relationship I postulated, as discussed above in my reply to Bratman, was this: in the absence of justification or excuse, a type of conduct is justifiably criminalized if all the tokens are justifiably censured and some are justifiably sanctioned. In saying this, I did not reach the question of the conditions, if any, under which excused conduct is justifiably censured, although I took it as given that it is not justifiably sanctioned. In any event, nothing that I can see about the relationship that I postulated between criminality and punishability commits me to the claim that excused conduct is justifiably censured. And that is a good thing, because it rarely, if ever, is.

Perhaps Brink is making an assumption about the relationship between the censure-sanction distinction and a different distinction that I draw, namely, the distinction between guilt and commission. I claim that a person can commit a crime without being guilty of it, as when there is insufficient evidence establishing her guilt despite her commission; and I claim that a person can be guilty of a crime that she did not commit, as when she helps someone else to commit a crime and is thereby implicated in the other person's commission. I use the guilt-commission distinction for various purposes in the book. But I do not believe that any of the purposes for which I use it requires postulating any particular relationship between the guilt-commission distinction and the censure-sanction distinction. I do claim that excuses are frequently present where there has been commission of a crime, as when an insane person commits a crime. Excuses need not negate commission. But this does not imply that excused conduct is censurable, for commission of a crime is insufficient for the appropriateness of censure. In fact, the relationship between these properties of action is quite complicated and contestable.

For instance, consider the case of someone who aids another to commit a strict liability crime, as when D1 helps D2 to drive over the speed limit but does not drive himself, and neither knows that the limit has been breached. In such a case, it is plausible to say that D1 is guilty of the crime but did not commit it and that he is not justifiably censured since he has an excuse (namely that he did not know the limit had been breached) that does not cancel guilt (since the crime is strict liability). Other descriptions of this case are possible, as well. And I do not mean to suggest that this is necessarily the correct one. My only point is that postulating the distinction between censure and sanction and between guilt and commission does not commit me to any particular account of the relationships between these complex properties. And I am certainly not committed by my use of these distinctions to the claim that excuses do not block censure.

This brings us to Brink's worry to the effect that I do not supply sufficient room for the claim that differing sanctions might be appropriate for utilitarian reasons that do not support the claim that differing degrees of censure are appropriate. In the book, I argue that it can be fair to sanction a last-act attempter and a completer differently because the only relevant constraint on what sanction is deserved is this: the issuance of the sanction must not express a greater degree of censure than is deserved. Provided that two different sanctions could express the same degree of censure, it is possible for two people who differ only in facts that are under the control of neither to be worthy of different degrees of sanction even though they are not worthy of different degrees of censure. There needs to be a reason to sanction the two differently; each must be given a sanction for which there is good reason. And in the case of the last-act attempter and the completer, I suggest that one important reason to give a greater sanction to the completer is that the state thereby expresses acknowledgment of the value of that which the victim has lost, a loss that we do not typically find in failed last-act attempts. However, nothing that I say precludes the possibility of not expressive but instead utilitarian reasons for sanctioning one person more than another who deserves the same level of censure, provided that the two sanctions express the same level of censure.

If, for instance, there are reasons of deterrence for punishing the completer more heavily than the last-act attempter, and the deterrent effect can be produced without change in the expressed level of censure, then we could without unfairness punish the completer more heavily than the last-act attempter. The explanation, that is, that I offered for the absence of unfairness in issuing different sanctions, despite equivalence in deserved censure, extends in principle also to the case in which the reasons for the difference are utilitarian.

Now, Brink expresses doubt that there are either utilitarian or expressive reasons to sanction last-act attempts less heavily than completions. I agree about the case of utilitarian reasons, although I disagree about expressive ones. It seems very likely that there is no additional deterrent value in the increase in sanction for completion. One might think that the increase deters second attempts following failure. But that is not true. What deters a second attempt is a sanction for the second attempt, whether it succeeds or fails, that adds to the sanction for the first. Nobody thinks that there needs to be a punishment discount for a first murder in order to prevent people from committing a second; it is good enough to promise to punish the second. Ditto for attempt and completion. Or one might try to support the punishment discount for attempt on different utilitarian grounds. One might think that a punishment discount incentivizes abandonment, for instance. But that would favor a punishment discount only in those cases in which the defendant believes (truly?) that he can still prevent completion by abandoning, and few last-act attempts, if any, are of that sort. So there are at least no obvious utilitarian considerations that support harsher punishments for completers than last-act attempters.

If there are utilitarian reasons favoring an increase in punishment for completion, I suspect that they derive from the utilitarian value of the state expressing solidarity with the victim and expressing our shared belief that the victim has lost something at the defendant's hand that is, perhaps, of irreplaceable value. One can imagine that such expressive acts have valuable consequences—perhaps they even serve to deter by serving to inculcate a respect for the interests of others that in turn leads some to refrain from committing crimes they might otherwise have committed. That is an empirical speculation and might be false, but it is plausible enough. I myself think that the expressive reasons for an increase in punishment for completion are reasons enough, even if the expressive act is costly and does not buy its costs back in deterrence or some other consequentialist value. Or, put more carefully, the expressive reasons are reasons enough, provided that in expressing the value we place on what the victim has lost, we do not thereby express more censure of the defendant than he deserves.

Brink notes, and in the book I acknowledge, that I do not offer an argument for the claim that there is no censure-luck. I suspect that the best argument for that claim is a “wrong kind of reason” argument of the sort now popular. The reasons for censuring to a particular degree must be grounded in facts about the agent being censured or about the conduct in which she engaged of a sort that are or ground facts about the way in which she has misused her capacities for recognizing and responding to reasons for action. Other reasons for censure are of the wrong kind. Do I have an argument for that claim? No. But it seems to me to be true. And, importantly, it is not the case that a parallel argument supports the claim that there is no sanction-luck. Sanction is one of the state's most important tools both of expression, and of control. So the reasons for sanction must be of a sort that support the state's use of this tool for its legitimate purposes. But the expression of censure is only one of the purposes that sanction serves, and so there could be reasons for sanctioning one person to a particular degree that are not reasons for sanctioning another who is deserving of the same degree of censure. Or at least, so it still seems to me, despite Brink's doubts.

A. Attempts of Negligent and Strict Liability Crimes

Alexander suggests that my aim in the book is to identify the necessary and sufficient conditions under which a person's conduct is, in fact and independently of the law, an attempt of a particular crime. He then sees this project in the philosophy of action as conjoined with a further claim on my part to the effect that the law does or ought to assign criminal liability for an attempt of a crime only if the defendant did indeed attempt that crime. But this is not the way to see my view. Rather, my first claim is that there is a type of attempt that is implicitly criminalized whenever we criminalize a completed crime and that is legitimately so criminalized provided that the completed crime itself is legitimately criminalized. This immediately raises a question: What are the necessary and sufficient conditions for being an attempt in that sense? The question, that is, is: What are the necessary and sufficient conditions for being an attempt that we can justify criminalizing by pointing to the fact that the completed crime is justifiably criminalized? The guiding commitment view is an answer to that further question.

So I do not claim to have identified, nor do I believe that I have identified, the ordinary notion of attempt. There are multiple ordinary notions, and only one of them is implicit in our criminal prohibitions of completed crimes. I think there is a notion of attempt that is implicit in the very act of criminalizing, whether that is accomplished through some complex democratic process, as in our society, or whether it is accomplished instead through fiat in service of the whims of the sovereign, or through acts of judicial crime creation, or what have you. To criminalize C is also to criminalize the attempt of C, in a sense of attempt that I try to identify and analyze. And further, when C is itself legitimately criminalized, then so is the attempt of C, in the relevant sense of attempt. This implication is cancelable—a legislature could announce that although C is a crime, the attempt of C is not—but in the absence of cancellation, the implication goes through. And even if the inference is canceled, it would be legitimate to criminalize the attempt, provided that completion is legitimately criminalized.

We can see why the difference between the view I hold and the view Alexander attributes to me is important by looking at what Alexander has to say about attempts of negligent and strict-liability crimes. In one of his examples, it is a crime to kill the governor, and the crime is strict liability with respect to the fact that the person killed is the governor. Alexander notes that under (one reading of)Footnote ³ the Model Penal Code Section 5.01(1), a person who tries and fails to kill someone who is the governor but whom he (reasonably) believes not to be, has committed an attempt of the crime. And he notes correctly that under the guiding commitment view, such conduct does not constitute an attempt of the crime of killing the governor (although it may be an attempt of some lesser crime, like plain old murder). The reason is that someone who flat-out believes that his target is not the governor is not committed by his intention to that person being the governor. But under the guiding commitment view, only conditions to which you are committed by your intention can be appealed to in a description of what you tried to do. Alexander is happy with the result—he does not want to punish such a defendant for attempting to kill the governor—but he thinks that I misdiagnose the problem with doing so. He thinks that I reason like so: this person is not in reality trying to kill the governor, so he should not be held criminally liable for trying to kill the governor.

But that is not how I reason—or I should say, if something that I wrote in the book makes it appear that I reason that way, then that is a mistake of expression on my part. For one thing, there is a perfectly good sense of “try” in which this person is trying to kill the governor. The person he is trying to kill is the governor, and under what I call the “wide” sense of trying, any facts about the act that the person would perform were he to do as he intends can be appealed to in a description of what he tried to do. The wide sense of trying is a perfectly ordinary sense of trying, it is just not the sense identified by the guiding commitment view. We use the word “try” that way all the time, especially when describing what other people are trying to do and especially where we anticipate facts about their intended conduct that they are unaware of. (“There he goes, trying to screw up again!”)

The point is not that the man who is trying, in the wide sense, to kill the governor is not really trying to do so. The point is, rather, that he is not trying to do so in the sense that is implicitly prohibited when we prohibit killing the governor. So, if the Model Penal Code implies that such a person has committed an attempt to kill the governor, then the Model Penal Code is identifying a particular type of behavior as falling under a criminalized type—namely “an attempt to kill the governor”—that it does not fall under. The problem with the Model Penal Code is not that it uses the word “attempt” in some nonstandard way. The problem is that it identifies a behavior as a particular crime when it is not that crime.

Here is another way to put the point: say statute 1 defines the crime of killing the governor. And statute 2 is Model Penal Code Section 5.01(1). I would say that statute 1 implicitly criminalizes as an attempt all and only those behaviors that would count as attempt under the guiding commitment view. If statute 2 implies that behaviors that would not so count are attempts, then statute 2 has it wrong. The criminal code in such a case is inconsistent, and statute 1, because it gets it right, rules. To label the person who tries to kill someone whom he believes not to be the governor as having criminally attempted to kill the governor is to violate the principle of legality. Statute 1 does not criminalize his behavior.

Why does not statute 2 criminalize it? After all, I just said that the implication that the attempt is criminalized when completion is can be canceled. Why should we not think of statute 2 as the legislature's way of canceling that implication and replacing it with an alternative? The answer is that nothing prevents us from thinking of things that way. But if we do, then the answer to the question “Why is it a crime to attempt to kill the governor?” cannot be answered by saying, “Because it is a crime to kill the governor!” That is, we cannot at once help ourselves to the most commonsensical explanation for the criminalization of attempts—namely, that the corresponding completed behaviors are crimes—and at the same time stipulate a definition of attempt under which more or less counts as an attempt than counts as such under the guiding commitment view.

The defendant who does not know that the person he is trying to kill is the governor has not attempted to kill the governor in the sense that is implicitly criminalized when we criminalized killing the governor. Since the reason that an attempt of C is legitimately criminalized is because C is, the Model Penal Code (read in the way that Alexander suggests) offers a definition of attempt that is indefensible. The problem is not that there is mismatch between the Model Penal Code's definition and that which is really an attempt. The problem is that there is a mismatch between that definition and what attempts must be, given that they are implicitly criminalized when we criminalize completion.

B. The “Real” Impossibility Puzzle

Alexander says that despite all the many pages that I write and many forms of conceptual and philosophical mechanisms that I develop in the book for responding to the problems of so-called “impossible” attempts, I never directly engage with the worry that there is a merely formal difference, or a difference merely in statutory drafting, between cases of “pure legal impossibility” and typical cases of attempt. The worry that he has in mind goes something like this. Start with D1, who tries and fails to have an adulterous affair while falsely believing that adultery is illegal. Full of guilt, she runs to her local police station to turn herself in. The police laugh in her face and send her home, for attempted adultery is no crime at all. Why not? Because adultery is not a crime. This case seems to instantiate a principle:

The problem is that if this principle is true, then differences in statutory drafting that seem irrelevant make a difference to what forms of conduct are legitimately criminalized as attempt. Alexander illustrates this with a version of the case of Lady Eldon, who brings French lace into the United Kingdom without paying duty and falsely believes that French lace is a dutiable item. We can then compare what she has done under two statutory regimes. In the first, the relevant statute is:

Some other part of the law announces that the dutiable items are I1, I2, I3 . . . In, and the list does not include French lace. In the second, the relevant statute is:

Here again, French lace is not listed. Lady Eldon seems to have attempted the S1 crime, for she intends not to pay duty on a dutiable item, and that intention guides her conduct. But she appears not to have attempted the S2 crime for her intention does not commit her to paying no duty on an I1, or an I2, or an I3, and so on. This is puzzling, given that the two statutory schemes seem to criminalize exactly the same behaviors.

Alexander suggests that this is the real puzzle about impossibility in attempts. If you can (usually) escape criminal liability for attempt when that which you are intending to do is not criminal, then it seems that whether or not you can escape criminal liability for attempt will turn on normatively irrelevant differences in the wording of statutes. How can that be?

There is something puzzling here, to be sure. But Alexander is mistaken in thinking that the puzzle here has anything to do with attempt per se. To see that the issue here is not special to attempt, consider an example involving not attempt but mistake. Let us assume that French lace is a dutiable item. Lord Eldon, bearing a gift for his wife, succeeds in bring some French lace into the United Kingdom without paying any duty on it. Italian lace is also dutiable, although Lord Eldon has none. He reasonably but falsely believes that the lace he brought in was not French but Italian and reasonably and truly believes that the lace he brought in was dutiable. If the crime requires belief that the lace is French, then Lord Eldon has not committed the crime. If it requires belief that the lace is dutiable, then he has. So, whether Lord Eldon has committed the crime seems to turn on whether S1 or S2 is law. How can that be, given that the two statutes seem to criminalize exactly the same things?

Recognizing that Lord Eldon should not avoid criminal liability in this case, jurisdictions have various forms of “lesser crimes” doctrines, according to which a mistake cannot excuse if there would have been a crime under the believed facts. That is a formal doctrinal fix that circumvents rather than solves the same problem that springs up in the kinds of cases that concern Alexander. The problem is this: Statutes often refer to the elements by appeal to properties of those elements different from the properties that need to be used in the mental representations of the elements that are involved in the mens rea of the crime.

For instance, say that, confronted with Lord Eldon's case, a court decided that the mens rea of the crime is present, given either a belief that that which one is carrying is dutiable or a belief that that which one is carrying is French lace. To be guilty, the defendant need not believe the French lace he is carrying is French, although such a belief would suffice; he need only believe it to be dutiable. Notice that a court could decide this regardless of whether S1 or S2 is law, although whether it is legitimate for the court to do so would depend on the right theory of statutory interpretation. But notice that if a court were to do this, there would be no need for a “lesser crimes” doctrine in order to resolve the case in a way that makes sense. When the mens rea elements of the crime are construed in this way, Lord Eldon's mistake about the lace's provenance does not negate mens rea.

Most importantly, if the court chooses to construct the mens rea elements of the crime in the way just suggested, the problem about attempt would disappear. Had Lord Eldon not made it across the border, he would still have been guilty of the attempted crime. So the problem is not in the transition from the statute to the attempted crime. The problem is in the transition from the statute to the completed crime. If we can specify what forms of completion are criminalized by a given statute—including, particularly, a full specification of what mental states are necessary and sufficient for the mens rea of the crime—then we will also settle the question of what does and does not constitute an attempt. So the hard problem here—and it is hard—has next to nothing to do with attempt or impossibility. The problem is the central problem about statutory interpretation in criminal law: given the words of the statute, what are the mens rea elements of the crime? This is a real problem, but it is not the real problem about impossibility in attempts because it is not a problem about attempts, in particular, at all.

C. Practical Competence, Inherent Impossibility, and Last Acts

Although Alexander does not link his remarks about my resolution of the so-called “inherent impossibility” problem and his remarks in support of the “last act” test for the act element of attempt, the two issues are linked, and worthy of discussion together. Let us start with the last-act test. Alexander asserts that there should never be guilt for an attempted crime if the defendant has not performed an act in the belief that he need perform no further acts in order to complete the crime. It is well known that this is an extraordinarily lenient standard for attempt. It is so lenient, in fact, that no court or legislature has ever accepted it, and no theorist before or since Alexander (and his coauthor, Kim Ferzan) has advocated it for more than the sake of argument. But I do not let the normative absurdity of the last-act test detain us. For my purposes here, what is important is that by advocating the test, Alexander unwittingly commits himself to a claim that he denies in his argument against my view of inherent impossibility. In justifying the last-act test, Alexander writes, “until the defendant takes what he believes is the last act necessary to complete the crime, he can always change his mind, and he always has conclusive reason to change his mind.”

There is a question as to why a defendant should get credit for possible future change of mind when he would not have changed his mind in fact. Alexander, for instance, is committed (unwittingly and unknowingly, but still) to the claim that attempted rape is not possible. The reason is that there is no point prior to penetration at which a would-be rapist believes that he has done all that he needs to do in order to accomplish penetration. Or, to put it in the language of change of mind, there is no point prior to penetration itself at which the defendant could not change his mind. To me, this result wears its absurdity on its face. Part of what makes it absurd is that it is just obvious that attempted rape is possible. But set that aside, for another part of what makes the view absurd is that when the would-be rapist is close enough to penetrating his victim it is clear that he is not going to change his mind. And so it is just silly to give him any kind of credit for the possibility that he will, a possibility that is not going to be realized.

But let us grant for the sake of argument that the possibility of change of mind matters, even when it will not occur. Alexander needs this in order to support the last-act test. But he is therefore committed to the perfectly true claim that sometimes things that are not going to happen, could.

Alexander is fond of the claim that the probability of every event is either one or zero. And he further claims that an event that has a zero probability of occurrence is not possible. These are false claims about modality and probability.Footnote ⁵ But that is not the important point about them for my purposes. The important point about them is that if they are true, then Alexander cannot justify the last-act test on the grounds that it alone accommodates the putative fact that change of mind is always possible in the absence of the last act. If Alexander is right (counter to fact), then the probability of change of mind is either one or zero. If zero, then it is not possible. And if it is not possible, then it should not count in the defendant's favor. So any time the defendant was not actually going to change his mind is one in which he has done enough for attempt liability, by Alexander's standards, even if the reason he is not going to change his mind is that the police wrestle him to the ground before he can do very much in furtherance of his intention. So this reason for refusing to assign criminal liability prior to the last act is undercut. Charity, I believe, forces us to say that when Alexander asserts that the probability of every event is either zero or one and asserts that any event with zero probability is impossible, he is speaking metaphorically or figuratively.

Now let us move to inherent impossibility. Alexander notes that I justify acquittal for attempt in cases involving voodoo, or other cases in which wildly hopeless means are undertaken in order to further a criminal intention, through appeal to what I call “practical competence.” I get to the notion of practical competence through the notion of guidance by intention, which is essential to attempts under the guiding commitment view. I note that to be guided by an intention is for that intention to have a causal influence on the agent that will culminate in the world coming to match the intention, given the absence of a crucial set of obstacles, namely, those that defeat ability or opportunity or prompt change of mind or execution failure. Often, our confidence that the causal influence that the intention is having is an example of guidance depends in part on our confidence that the agent will take further steps toward the satisfaction of his intention in the future; sometimes additional action in furtherance of the intention is necessary to the world coming to be as intended, and so we cannot characterize the intention's influence as one of guidance if we do not think that the agent will engage in that further action. But a necessary condition of engaging in such further action is practical competence—the capacity to recognize what needs to be done to do as intended.

My claim was that the means undertaken by someone like Carole Hargis, who thought that baking a tarantula in a blackberry pie was a good means to killing someone, shakes our confidence in her practical competence. And that in turn shakes our confidence that she would indeed undertake the means necessary to do as she intends, even given ability, opportunity, and the absence of change of mind and execution failure. Practical competence is necessary for guidance when more needs to be done. And guidance is necessary for attempt. And so sometimes doubt about practical competence even when it is present provides doubt about attempt because it provides doubt about guidance even when it is present. The problem is not that there is no attempt—there often is—the problem is that we lack sufficient evidence that there is.

Alexander's first objection is inconsistent with his views about the act element of attempt. He suggests that nothing about what the defendant might or might not do in the future ought to matter to his criminal liability for attempt. But for the reasons just explained, Alexander himself relies on facts about what a person might or might not do in the future to justify assigning or refusing to assign criminal liability for attempt. And any general considerations about modality or probability that seem to make such appeals suspect are utterly mistaken. So Alexander cannot think that that aspect of my view all by itself is objectionable. He must think, instead, that there is something wrong with the particular thing to which I appeal about what the agent might or might not do in the future. And he does think that, for Alexander notes that my line of thought does not support the claim that there is no guilt for attempt when the defendant has engaged in the last act. And so he does not understand why I would claim that the voodoo practitioner, or Carole Hargis, is not guilty of attempt.

This objection equivocates on the notion of a last act; neither the voodoo practitioner nor Carole Hargis has engaged in the last act in the sense of relevance. If no further action on the agent's part is actually required in order to complete the crime, then we can indeed know whether there is guidance without inquiring about the agent's practical competence. But the mere fact that the agent believes that no more action is required is irrelevant to the question. To obviate the need to inquire about practical competence, the act must be the last in a different sense from the one Alexander has in mind in offering the last-act test.

Carole Hargis, for instance, could not kill her husband with the tarantula pie without engaging in additional conduct; she would need, for instance, to beat his head in with a curtain weight, as she in fact eventually did. But if she thought that the tarantula thing would do the trick, what supports the claim that she would seize the opportunity to beat him to death with a curtain weight? Maybe she would try to bake the curtain weight in a pie for him to eat, thinking that would get the job done. Another way to put it: we lack sufficient evidence that when her murderous intention caused her to bake the tarantula pie, it was having a causal influence that would culminate in her killing him. Eventually, her intention had such an influence. And it is quite possible that the influence it had was merely an extension of the very same causal influence it had on her when she baked the pie. But there is insufficient evidence supportive of that claim and so insufficient evidence that when she baked the pie, she was trying to kill her husband.

A. Intention, Content, Circumstances, and Foresight

Moore calls attention to the fact that under the guiding commitment view, an agent's intention can commit him to a condition in one of two ways. The condition can be “in” or “included in” the intention's content or else it can be a necessary concomitant of the intention's playing its proper causal role. Moore thinks that this second way through which an intention can constitute a commitment to a condition is part of the view in order to allow for an intention's constituting a commitment to a circumstantial element of a crime. In fact, however, I appeal to commitments constituted by an intention's proper causal role for a different reason. It is my way of responding to the observation that a person can intend a particular event without his intention's content by itself committing him to any fact about how that event is caused. After all, a belief that p and an intention that p have the same content,Footnote ⁶ but the belief that p obviously includes no representation of how it comes to be the case that p, and so neither does the intention.

However, it is often the case, although not always the case, that the intention will nonetheless commit the agent to himself being the cause of the event it represents. For instance: D intends [that V dies].Footnote ⁷ Why does this intention commit D to the following: D causes it to be the case that V dies? The answer is not that the content of D's intention represents her as the cause of V's death; it does not, as can be seen by noting that D's intention does not represent D at all, but only V. Rather, the answer is that if D's intention plays its proper causal role, then D causes V's death.

This point is important to me in part because it allows me to specify a different mens rea standard in attempts with respect to the act and result elements of crimes. An intention that a result comes to pass suffices for the mens rea of the attempt with respect to the result elements of the completed crime, but an intention to perform a particular act is required with respect to the act elements of the completed crime. (This in turn provides for a principled answer to the question of when a solicitation rises to the level of an attempt.) The reason for this difference is that a commitment to causing a particular event is insufficient for a commitment to being the agent of it or to doing it. Moore disagrees with this further claim, but that disagreement need not detain us here, for even if causing is sufficient for doing, it needs to be acknowledged that our intentions can at once represent future events, fail to represent the causes of those future events, and nonetheless constitute commitments with respect to their causes. My appeal to commitments constituted by an intention's proper causal role explains how this is possible.Footnote ⁸

In fact, I believe that the only commitments constituted by an intention thanks to its playing its proper causal role are commitments to one of two things: being the cause of an event, or being the cause in the special sense that is involved in being the agent of the event. Since when an event is a circumstantial element of a crime, the cause of the event is irrelevant to the crime's commission, commitment to its being caused by the agent or by something else entirely is irrelevant to attempt of the crime. And so commitments to circumstantial elements of crimes are not constituted by an intention thanks to anything about the intention's causal role but instead thanks to the intention's content. Someone who is trying to damage property not his own, for instance, has an intention the content of which commits him to property not his own suffering damage and has an intention that commits him, thanks to its proper causal role, to being the cause of the damage and perhaps to being the cause of that damage in the special sense that is required for him to be the agent of it. But since completion does not require that he be the cause of the fact that the property is not his own, attempt does not require that he be committed to causing that. Rather, his commitment to that condition is constituted by his intention thanks to its content rather than its proper causal role. It is because his intention represents the relevant object as not his own, as belonging to someone else, that he is thereby committed to that in the sense required for attempt.

The remarks just made, like the guiding commitment view itself, draw on a distinction between an intention's content and the set of commitments to conditions that are constituted by the intention. The latter is broader than the former. Moore describes the appeal to commitments constituted by an intention's causal role as a “supplement” to those commitments constituted by the intention's content. And that is the right way to think of them. But still, more should be said, and Moore invites me to say it, about what is and is not in an intention's content and, even more importantly, how an intention's content constitutes commitments to particular conditions.

Let me approach these questions first by saying a word about Moore's skepticism to the effect that what distinguishes an intention from, say, a belief are the norms of rationality that apply to the person in light of the mental state. Say mental state M1 represents future condition G (for “goal”). Is the agent under rational pressure to form a mental state M2 representing conditions that must first come to pass if G is to come to pass? If M1 is an intention, then the agent is under such rational pressure: we are under rational pressure to intend the means to those things that we intend to bring about. But if M1 is a belief, then he is not. One can rationally believe things about the future without having any mental states whatsoever representing what needs to happen first for the world to come to be that way. The claim is that intentions, and beliefs, too, are normatively constituted.

Moore prefers a functionalist view according to which intentions are constituted by their causal roles.Footnote ⁹ M1, on such a view, counts as an intention rather than a belief because in part, for instance, it plays the role of prompting a mental state like M2, representing actions that must be performed if G is to be achieved, which itself causes the agent to engage in such conduct. And not just this; there are many distinctive causes and effects of intention that are different from the causes and effects of belief, or desire, or other types of mental state. I do not deny that intentions play the distinctive causal roles that such a functionalist position takes to be constitutive of them. They do. And in fact, if they did not, they would be all but useless. It is worthwhile to have intentions because more often than not they play these causal roles, roles that allow us to achieve many goods that are difficult or impossible to achieve without them. Similarly, I am confident that Moore agrees that those who have intentions are governed by the kinds of norms of rationality that Michael Bratman has done more than anyone else to describe. Further, I suspect that the reason that the distinctive norms of rationality apply to intenders is because intentions typically play a certain causal role in agents’ psychologies. But I do not think that this weakens or invalidates the claim that what is constitutive of intention is the applicability of norms rather than causal role.

Consider an analogy. We can make two observations about the biological structures we call eyes: they play a causal role in generating visual experience, and they do not function as they ought if they fail to generate visual experience. If things similar in crucial respects to my eyes, for instance, the eyes of healthy people, did not play that causal role, then it would be ridiculous to suggest that my eyes ought to. If nobody can see with those things in their heads, then those things in their heads are not failing to function as they ought if they fail to help people see. But it does not follow that causal role is constitutive of what an eye is and the applicability of norms are not. It can be the case that an object has property P1 because it has property P2 and also that it is of type T thanks to the fact that it has P1. Here P1 is the property “ought to help you see,” and P2 is the property “things like it help people to see.” If someone points to my eye and asks, “Why do you think you ought to be able to see with that?” it is perfectly sensible to answer, “Because it is an eye.” And if someone points, as it were, to M1, a representation of a future world in which G obtains, and asks me why I think, thanks to that, the agent ought to form an intention with respect to what he believes to be a means to G coming to pass, it is perfectly sensible for me to answer, “Because M1 is an intention.”

To be sure, in offering that answer I am implying that those who have mental states similar to M1 in crucial respects have a mental state that plays the causal role of, among other things, prompting the formation of intentions with respect to the means. But still, what is crucial about M1 that makes it an intention is not that mental states like it play the causal role that intentions play but instead that thanks to that fact, M1 should.

Of course, there are hard questions as to whether those mental states that are “relevantly similar” can be identified noncircularly; maybe those are just the mental states that are like M1 in the following respect: they are also intentions. But this is no more of a problem for those who take intentions to be normatively constituted than it is for those who take them to be constituted by causal role. An explanation needs to be given for the possibility of a token intention misfiring and failing to play the causal role distinctive of intention while still being an intention. And providing such an explanation requires identifying a set of mental states that are relevantly similar to the token intention and play a causal role that the token does not. Still, I was too quick to say that what makes a mental state an intention is the norms that apply to those with intentions, period. The causal role played by similar mental states and the mental state itself also matter and may be essential. The point of real importance for my purposes is a point about which Moore and I agree: intentions play distinctive causal roles, and those who have them labor under distinctive rational pressures. I emphasized the norms governing those with intentions because it is a normative notion of commitment—and of the commitments constituted by an intention—that is crucial to trying and to the proper characterization of what a person is trying to do. But causal role matters too.

The side trip just taken about the causal roles and norms applicable to intention bears on the question of what I mean by an intention's “content” in contrast to the full set of conditions that one is committed to thanks to having the intention. One might try to explain what it is for a mental state to have a particular content by citing that mental state's causal role. To have a belief that p rather than q is to be disposed, for instance, to say things like, “I believe that p.” It is also to be disposed to form new thoughts with content q when one comes also to believe that if-p-then-q. However, the causal role associated with content p will itself depend on what kind of mental state it is that has content p. A belief that p and an intention that p play different causal roles in the psychology of a rational agent. To take an obvious example, the person who intends that p may not be disposed to say, “I believe that p,” for she might be in serious doubt as to whether she will succeed in achieving what she intends. And even if she is so disposed, it is not because she has the intention that p but because she has a separate mental state with the same content, namely a belief that p.

So it seems to me that efforts to reduce content to causal role must rest on a prior and independent characterization of what distinguishes intentions from other sorts of mental states, such as beliefs. Of course, the same point can be made about norms of rationality. Rationality requires something different from those who believe that p and from those who intend that p, and so it seems that we cannot hope to reduce content to applicable norms unless we have a prior and independent characterization of what distinguishes different sorts of mental states.

It seems to me that the right response to this line of thought is to embrace the contrapositive: since beliefs and intentions are distinguished on the basis of either causal role or applicable norms or both, content is not to be reduced to causal role or applicable norms. And this makes me lean toward a “language of thought” approach to content. Each contentful thought, whether an intention or belief or what have you, is associated with a sentence or a phrase in the language of thought. That sentence or phrase has a meaning. And that meaning is the thought's content. So, as I see it, some (but not all) of the “match” questions that Moore discusses and mentions in his commentary on my book are really questions in the philosophy of language. Why is it that someone who intends to take something of someone else's has an intention favoring “the asportation” of the relevant goods? Because “to take” means the same thing as “to asport.” The phrase in the language of thought has the same meaning as the phrase in the statute defining larceny, and so there is sufficient “match” between the two respective contents. Explaining why they have the same meaning is a project for philosophy of language. If we knew what principles determine the meaning of a phrase or a sentence, then we could apply such principles here.

However, we can move forward pretty effectively in theorizing about criminal law without an articulation of such principles, because ultimately such principles are those on which we rely when we make everyday meaning judgements. Those who understand English and understand the language of thought, can tell that “taking” and “asporting” mean the same thing. You do not need to be a philosopher of language or a linguist to see that there is something bogus about the defendant's argument when he says in his defense, “I did not intend to asport the necklace, I merely intended to take it!”

The lines of thought just sketched are thoroughly explored in the philosophy-of-mind literature, especially the literature responsive to functionalist proposals and the literature developing the language-of-thought hypothesis.Footnote ¹⁰ I know that I am not contributing in a new way to those debates. Rather, I am trying to sketch roughly where I stand on the map that has been drawn so well by others. But notice that the position I take has some important implications, in particular for understanding the views that I offer of the mens rea required for the attempt with respect to the circumstantial elements of the crime and of the relevance of the intent-foresight distinction to attempt.

First, let us consider mens rea with respect to circumstances. Moore seems to think that if I required that the commitment to the circumstantial elements derived from the intention's content rather than its proper causal role, I would not need any of the complex mechanism I develop for understanding the different commitments generated by de re and de dicto intentions. But he is mistaken about that. The reason is that the articulation of the middle way between an intent standard and a belief standard—the articulation of an account under which it is at once true that an attempter must intend the circumstance and that that is not too high a standard—requires an appreciation of the way in which a condition can be included in an intention's content without it thereby being the case that the agent is committed to causing the condition.

If a defendant has a de dicto intention [to kill a prison guard], to use Moore's example, then, when that intention plays its proper causal role, the defendant causes it to be the case that it is a prison guard that he kills. Of course, he does not cause the person to be a prison guard—assuming he does not work in the prison's personnel office. But he does cause it to be the case that it is a prison guard and not, say, another inmate whom he kills. He does not cause it to be the case that the victim possesses the property of being a guard, but he does cause it to be the case that two properties are conjoined: the property of being a prison guard, and the property of being killed. He is thus committed by his intention's content to the person he kills being a prison guard and committed by his intention's proper causal role to being the cause of that. To require such a de dicto intention for the attempt would be too high a standard precisely because completion of the crime does not require being the cause of the fact that the person killed is a prison guard. Since completion does not require causation of that condition—causation of the kind that is present when the target was selected because he is a prison guard, as in the guard-killing contest—attempt does not require an intention-based commitment to it.

The challenge, however, is then to explain why a de re intention [to kill a prison guard] constitutes any kind of commitment at all to the target's being a prison guard. The constraint that is placed on such an explanation is that it must identify some sense in which rationality places the agent under pressure to accept that condition, given his intention. If a person is committed by his intention to causing a condition, there is a whole series of well-known ways in which he is thereby under rational pressure to accept it. If he is rational, he does not deliberate under the assumption that it will fail to come to pass; he forms intentions to cause conditions that are necessary for its occurrence; in the absence of new information, he does not consider courses of conduct that are incompatible with it; and so on. What is lacking—or, I should say, with all due modesty, what was lacking prior to the writing of my book—was an account of the sense in which rationality requires acceptance of a condition included in the content of an intention that the agent is not thereby committed to causing. My description of commitments of nonreconsideration and noncomplaint is an effort to explain the sense in which there is a distinctive set of rational pressures with respect to a condition's occurrence that derive from an intention that includes the condition in its content but does not commit the agent to causing the condition.

Moore recognizes that I distinguish between and predicate differences of treatment under the law on a de re intention [to A the thing that is P], on the one hand, in contrast to, on the other hand, the conjunction of an intention [to A the thing] and a belief [that the thing is P]. Moore says, and I think he is right, that this “cuts the bologna very thin.” I say a word elsewhere in defense of the claim that the difference in the commitments of these two agents is of some normative significance.Footnote ¹¹ But even putting aside the question of whether there is a normatively significant difference between these two people, I think there is a difference in what they are trying to do (in the sense of trying that matters, I believe, to the law—the sense identified by the guiding commitment view). The first, and not the second, for instance, is trying to kill a prison guard. The second, by contrast, is merely trying to kill someone (who happens to be a prison guard). Whether this difference in what these two agents are trying to do ought to make a difference to the way they are treated under the law is a more complicated question. What I do think, however, is that if they should be treated the same, it is not because they both have committed the same attempt; it cannot be because of that, because that is not true.

In a related vein, Moore suggests that I should embrace a “closeness” approach to the fraught distinction between intention and foresight in contrast to the approach I do adopt. What I suggest in the book is that there is a difference, to use another of Moore's examples, between the person who is guided by an intention to chop off John the Baptist's head while foreseeing his death, and a person who is guided by an intention to kill him. The former is not trying to kill him, while the latter is. Thus the grounds for the criminalization and punishment of these two agents’ behaviors must be different. The latter is committing a crime because it is a crime to kill another person; the former is committing a crime, and maybe even the same crime, for a different reason. My sketch of a suggestion of that alternative reason is that when she tries to cut off John the Baptist's head, she is committing a crime that is just as bad as, and bad in the same way as, the crime of the person who is trying to kill John the Baptist.

This is to appeal to an additional substantive principle over and above the general principle allowing for the criminalization of attempts of that which is legitimately criminalized, namely the transfer principle. It is not a crime to try to chop off John the Baptist's head because and only because it is a crime to kill him. It is a crime to try to chop off his head because it is a crime to kill him and because trying to chop off his head is just as bad as trying to kill him. To use a closeness approach, by contrast, is to insist that the person who is trying to chop off his head is trying to kill John the Baptist even though her intention does not commit her to killing him. That requires a conception of trying distinct from the guiding commitment view. But if it is tryings in the sense of the guiding commitment view that are criminalized under the transfer principle, as I propose, then the closeness approach requires that more be said about the grounds for the criminalization of attempts generally, merely to accommodate the criminalization and punishment of one bizarre form of attempt, namely closeness attempts.

I and someone who advocates a closeness approach agree that additional principles need to be appealed to in order to explain what is happening in cases of that kind; we need to say something to explain why closeness attempts should be treated under the law as something they are not, namely, attempts to commit the relevant completed crime. But I put such additional principles in a different place from what Moore urges. I think the guiding commitment view gets it right about what attempts are and are not criminalized implicitly when completion is criminalized. That set does not include attempts such as that involved when one tries to chop off the head and merely foresees the killing. But the accommodation of such cases does not warrant a wholesale revision of the position I adopt. Such cases warrant treatment as exceptions to the rule rather than warranting revision of the rule itself. We should accept the guiding commitment view and admit that closeness attempts are to be treated like the attempts they are close to being for reasons that go beyond the transfer principle, rather than giving up on the guiding commitment view in order to insist that a close miss is as good as a hit.

Incidentally, it is striking that Moore is drawn to a closeness approach when it comes to foresight of close, unintended consequences—he thinks that there is probably some adequate analysis of closeness under which the foreseen consequences are indeed intended—but at the same time does not seem drawn to the parallel view when it comes to circumstances. One could say, for instance, that the person who intends to kill someone whom he knows to be a prison guard but does not represent as such in his intention has, thanks to a rule according to which a close miss is a hit, an intention to kill a prison guard. I reject this kind of analysis in the case of circumstances and also reject it when it comes to results. A miss is a miss, even if close. But it seems to me that if Moore thinks that when it comes to results, a close miss is a hit, he should say the same thing about circumstances. At the least, he—or anyone drawn to a position of the sort he sketches—owes us an explanation for why the two cases are to be treated differently.

All of this bring us to the heart of the issue of why attempts are crimes in the first place. If they are criminalized because they are tryings and if they are criminalized because of what they are tryings to do, then wherever there is a difference in what two agents are trying to do, there is a difference in the criminality of their behavior or in the grounds for the criminality of their behavior. This second disjunct is important. Both the person who is trying to kill a prison guard and the person who is trying to kill someone (who happens to be a prison guard) are engaging in a criminal attempt. In that sense, their tryings do not differ in their criminality. But there is still a difference in the grounds for conceiving of their distinct tryings as crimes. The conduct of the first is a crime because it is a crime to kill a prison guard. The conduct of the second is a crime because it is a crime to kill a human being. And the same can be said about true attempts and closeness attempts. The further question, then, is how we need to think about the criminalization of attempts in order to make good on this picture.

B. The Criminalization of Attempt and Neutrality among Theories of Criminalization

The points made in the final few paragraphs of the last section foreshadow what I want to say in reply to the first part of Moore's commentary, in which he suggests that in my discussion of the grounds for the criminalization of attempt, I cannot and in fact do not maintain my avowed neutrality about the right theory of punishment and the right theory of the desert basis of crimes. My idea was that it is a desideratum against which a theory of criminalization and an intertwined theory of punishment are to be tested that it supports the transfer principle. If your theory of criminalization implies that it is sometimes not justified to criminalize the attempt, even when completion is justifiably criminalized and in the absence of a specific legislative decision against the criminalization of the attempt, then there is something wrong with your theory of criminalization. Seeing this just requires a little common sense about the nature of the act of proscription.

To use one of my favorite examples, the father who prohibits his daughter from jumping on the couch prohibits, also, her trying to jump on the couch. If the precocious child asks for a rationale for this latter prohibition, given that the father said nothing at all about trying when he prohibited jumping on the couch, all the father needs to do to justify it is to point to the fact that it is legitimate to prohibit jumping on the couch. He can says things like, “Jumping on the couch makes it dirty” in justification of the prohibition of the attempt, even though the failed attempt might leave the couch clean.

Moore thinks that an adequate justification of the transfer principle will make itself superfluous. Once we see why it is justified to criminalize the attempt if completion is justifiably criminalized, we also see why it is justified to criminalize the attempt period, and without appeal to anything about the justifiable criminalization of completion. For a retributivist like Moore, for instance, what is required to explain why it is justified to criminalize attempts is nothing more nor less than an account of why attempts deserve punishment. That account might appeal to features of attempts that they share with completions and which also serve to justify the criminalization of completion, but the fact that the features are shared plays no role in the justification of the criminalization of the attempt. Or so Moore thinks.

Now, I do not deny that this is possible. There may be more than one good rationale for the criminalization of attempt. And in fact, there may be a rationale that appeals to features of attempt that are shared with completion but does not appeal to the fact that they are shared in rationalizing the criminalization of attempt. This is perfectly possible. But it is not the form of rationale for the criminalization of attempt that interests me, in part because I do not think it is the form of rationale that animates or should animate attempt-law doctrine. Rather, I think that attempt law is and should be animated by the idea that attempts are implicitly criminalized by default when completions are. I think the law is and should be animated by the idea that to justify the criminalization of attempt, all we need to do is to note that completion is justifiably criminalized. This is probably true because completion and attempt have crucial features in common. But it is not those features that we appeal to independently in order to justify the criminalization of attempt. It is instead the fact that attempts are attempts of justifiably criminalized behavior.

It is worth noting one important implication of the approach that I am trying to take that bears on the question of the degree to which advocacy of the transfer principle is consistent with neutrality about the answers to the biggest questions in criminal-law theory concerned with the appropriate rationale for punishment. Moore thinks that my argument for the transfer principle gets dangerously close to committing me to the claim that the results of conduct are not included in the desert basis of the conduct. He thinks that I get close to committing myself to the claim that what you deserve is not a function of what you cause.Footnote ¹² After all, reasons Moore, recognizing that attempts and completions often have different results, I cannot appeal to commonality in result in order to justify the criminalization of attempt. And on the assumption that the features of attempt that I do cite serve to show how attempts are deserving of punishment and so are justifiably criminalized because they are so deserving, it seems to follow that the results are of much less importance to desert than one might have thought.

If Moore is right in this line of reasoning, it is a fault with my argument for the transfer principle. In fact, things are complicated here, for, as demonstrated by the jumping-on-the-couch example, the results included in completion can serve to justify the criminalization of attempt, including attempts that do not cause those results. Part of the reason that it is justified to criminalize attempted murder is that murder involves the causation of death. This is a puzzle. How could the features of a different act (namely completion) serve to justify another's (namely attempt's) criminalization when the latter does not have that feature?

To answer this, we need to say more than I say in the book about the commonalities between attempt and completion that support criminalizing attempt when completion is criminalized. The really crucial commonality between an attempt to C and C is that the attempt is intentional under a description that implicates all of the features of C in virtue of which it is justifiably criminalized; in particular, it is an attempt to C and thus implicates all of the features of token C-ings in virtue of which they are of that type.

In the argument for the transfer principle offered in the book, I do not emphasize this in the way that I should, pointing instead to two important commonalities between attempt and completion that seem to support the criminalization of the attempt independently of its relationship to completion.Footnote ¹³ When we criminalize C, we are criminalizing anything that is a C-ing and anything that is intentional under a description that implicates all and only the features of C-ings in virtue of which they are criminalized. The latter category includes all and only the attempts of C, as defined by the guiding commitment view. It is only the guiding commitment view, of all the various ordinary notions of attempt, that allows us to appeal to a condition in the characterization of what the person has attempted if and only if that condition is one to which he is committed by his intention. And so it is only the guiding commitment view of attempt that identifies the set of attempts that are implicitly criminalized when completions are and are justifiably so criminalized when completions are.

There are the makings here also of a response to another aspect of Moore's commentary—his challenge to explain why guidance by intention is necessary for criminalization, given that, for reasons he explains, a person who intends to commit a crime but is not guided by that intention employs the very same modes of recognition and weighing of reasons as the person who is guided by the same intention. Moore is right that I tend to cover over the need to answer the challenge by speaking of modes of recognition and response to reasons. The response part brings in guidance. Moore suggests that what is really going on here is that I am smuggling in a particular account of the desert basis of censure—guidance is included in the desert basis. I suppose that that is what I think, to be honest. And it is probably true that because I think it, I failed to really see the need to answer the challenge Moore raises.

But I also think there is something to say in response to the challenge that does not draw on that particular (true, in my view) claim about the desert basis. In having an intention to C, the agent has not engaged in any act that is intentional under a description that implicates all of those features of C-ing in virtue of which it is criminalized; intending to C is not an intentional act. The result, I think, is that the intention to C, where there is no guidance by that intention, is not implicitly criminalized when we criminalize C. We cannot justify the criminalization of the nonguiding intention by simply saying, “C is legitimately criminalized.” The intention to C, that is, is not criminalized under the transfer principle. If the father punishes the child for intending to jump on the couch after he prohibited jumping there, she has a legitimate gripe, for although trying to jump on the couch was among the things he prohibited, intending to, in the absence of guidance by that intention, was not. (Perhaps things would be different if after prohibiting jumping on the couch, he had added, “Don't even think about it!”)

I hope, although I must admit to some uncertainty about it, that this does not amount to an appeal to the principle of legality in a way to which I am unentitled. I do believe that even if someone with authority did prohibit the intention to do something criminal, it would still be illegitimate to punish for such an intention. And I do think that the explanation for that is that punishment, and particularly the censure expressed by punishment, would be undeserved in a way that undermines punishment's legitimacy. The claim that I want, however, is not that intention cannot be criminalized, which is surely false, or even the claim that it cannot be legitimately criminalized, which is surely true. The claim I want is, rather, that the argument that supports the criminalization of attempt does not support the criminalization of intention in the absence of guidance. And the argument for that claim is not just that we have not, as a matter of positive law, criminalized the intention when we criminalized the completed crime. Rather, the argument is that the transfer principle supports criminalizing only those things that fall short of completion in a way that is compatible with their being implicitly criminalized when we criminalized completion. Unguiding intentions, because they involve no action intentional under a description that implicates all of those features in virtue of which C is justifiably criminalized—they are not tryings to C—are unlike completions in a crucial respect that blocks them from being criminalized under the transfer principle.