Gideon Yaffe's Attempts: In the Philosophy of Action and the Criminal Law Footnote ¹ brings a sophisticated range of ideas from the philosophy of action to bear on important issues about the criminal law of attempts. It helps focus our attention in a philosophically disciplined, informed, and insightful way on many fascinating and perplexing issues that arise when we reflect seriously on the very idea of criminal liability for an unsuccessful attempt.

I focus here on one complex thread in the overall argument of the book. This leads to some thoughts about the relations between metaphysical and epistemological issues in the philosophy of action and normative issues about the legitimate criminalization of attempts.

Let us begin with Yaffe's thought that there is an extremely plausible principle connecting criminal liability for a successful action with criminal liability for the mere attempt to perform that action. Yaffe calls this the “transfer principle,” and its canonical formulation is: “If a particular form of conduct is legitimately criminalized, then the attempt to engage in that form of conduct is also legitimately criminalized” (21, italics in original). Yaffe avers that this principle “is reflective of a deeply entrenched element of our common-sense moral thought” (23). And, it is important to note, this transfer principle supports the criminalization of attempts tout court, not only attempts that have some further property—for example, attempts that raise the probability that there will in fact be the completed crime.

An initial worry about this idea, one that Yaffe highlights, is that in criminalizing mere attempts, we do not want to be criminalizing mere thoughts, even if these are thoughts about how to pursue a certain criminal activity. We do not want to be creating “pure thought crimes” (215). But the transfer principle supports the criminalization of attempts tout court: it does not include an explicit limitation to attempts that involve action on the part of the agent that is in fact, as I will say, in the public space. So does the transfer principle lead us to “pure thought crimes”?

Well, what is an attempt? Can there be an attempt in the absence of action in the public space? One of the significant accomplishments of this book is a clear and plausible theory of the nature of attempts. This theory involves two basic ideas: the idea of intention-based commitment, and the idea of being guided by such a commitment. An intention in favor of C involves a characteristic commitment in favor C. For an agent to attempt to perform C is (roughly) for that agent to have such an intention-based commitment to C and for that intention-based commitment to guide that agent. Jones attempts to kill Smith, for example, just in case Jones has an intention that involves a commitment to killing Smith and this intention-based commitment guides Jones. And this might be true even if Jones does not in fact kill Smith: perhaps Jones shoots but Smith is wearing a bulletproof vest. This is the “guiding commitment view” of attempts (73).

What is it to be guided by an intention-based commitment? Here Yaffe has a complex and subtle story, one that involves what he calls “the completion counterfactual” (94). For present purposes, we can work with the basic idea that you are guided by such a commitment just when the underlying intention motivates, and “the distinctive feature of motivation by intention is that, if all goes well, it leads to doing as one intends”(94). And the conditions for all going “well” are that there be relevant ability, opportunity, absence of change of mind, and absence of “execution failure.” (An example of what Yaffe has in mind in talking of execution failure—a failure simply to execute—is the failure of the golfer, in J.L. Austin's famous exampleFootnote ², to sink the putt he had the ability and opportunity to sink (94).) So, for example, if Jones is indeed guided by his intention to kill Smith, then, if all goes “well”—where this includes conditions of ability, opportunity, absence of change of mind, and absence of execution failure—he does kill him.

What is the relation between attempts of the sort described by the guiding commitment view and acts in the public space? You can intend to kill without acting in the pubic space. But if that intention guides you, must there be some act in the public space that is the issue of this guidance? Well, such guidance will normally involve acts in the public space—for example, successfully moving one's body. But, according to Yaffe, such guidance does not necessarily involve such acts. Yaffe's main reason for this is that the intention-based guidance can occur even if there is unanticipated, relevant paralysis. Perhaps one is guided by one's intention to assassinate the king—and so one attempts to assassinate him, and one would assassinate him if all went, in the relevant sense, “well”—but one is suddenly stricken with paralysis and as a result does not even successfully move one's body (91–92). Though what is attempted would, if performed, be an act in the public space, this attempt is not.

Though this is to go beyond what Yaffe says, there also seem to be other kinds of examples of guidance by intention without action in the public space. What I have in mind are cases of attempting to achieve some end by trying to think something through, where one's thinking and planning, while goal-directed, are entirely, as we say, in one's head. Perhaps my early steps in my attempt to defraud you involve solely my complex thinking about how I am going to do this. My thinking is guided by my intention to defraud you; and if all were to go, in the relevant sense, “well,” I would indeed defraud you. Granted, the world being what it is, my “last act” attempt will go beyond such thinking and include, say, my communication with you. But in my early thinking it seems that I am nevertheless involved in what is, according to the guiding commitment view, an attempt. This attempt is, so far, only a non-last-act attempt. But Yaffe explicitly includes among the attempts covered by the transfer principle non-last-act attempts. And here, again, it seems that though what is attempted (namely, fraud) would, if performed, be an act in the public space, my early-stage attempt (as Yaffe understands what an attempt is), consisting solely of my thinking, need not itself be an act in the public space.

So we have a problem: given that there are attempts that do not involve acts in the public space, and given that the transfer principle supports the criminalization of attempts tout court, not only attempts that involve acts in the public space, we seem to be in danger of endorsing “thought crimes.” This danger arises as a result of the combination of the transfer principle and Yaffe's theory of the nature of attempts, a theory that requires both intention and intention-based guidance but nevertheless allows that such guidance need not involve action in the public space. And each of these two elements of Yaffe's theory is initially quite plausible. So, what to say?

Yaffe's response is not to modify either of these two basic features of his theory but to provide a different explanation of why his account does not endorse criminalizing mere thinking. Yaffe's explanation appeals to a principle he calls the “means requirement.” This principle states: “a defendant has committed a criminal attempt only if he has performed an act in the class of means” (219, italics in original). We need to understand what this requirement says and what its role in the overall argument is supposed to be.

The first point to note concerns the idea of an “act” as it appears in this principle. As I read him, Yaffe understands this idea of an act as the idea of an act in the public space. It does not include entirely mental activities—for example, solving a math problem in your head. That is why his means requirement can be seen as a response to the problem about criminalizing mere thoughts.Footnote ³ What is required by the means requirement, then, is an act in the public space that is performed by the agent as a means to the end intended in the attempt.

The next point concerns the precise way in which the means requirement is supposed to help Yaffe retain both the transfer principle and the guiding commitment view without endorsing the criminalization of mere thoughts. Suppose that we see this means requirement as an independent, normative constraint on which attempts are legitimately criminalized. Given this independent, normative constraint, no attempts will be legitimately criminalized if they do not involve an act in the class of means. So we need not worry that in legitimately criminalizing attempts we are criminalizing mere thoughts.

So interpreted, the means requirement is a distinct normative constraint on the criminalization of attempts. Call this the normative interpretation of that requirement. So understood, we would need an explanation of why this is an appropriate independent, normative constraint. But the crucial point for now is that when it is interpreted in this way, the means requirement is incompatible with the transfer principle. After all, according to this normative interpretation of the means requirement, even if C is legitimately criminalized, the mere attempt to C is not legitimately criminalized if it does not involve an act in the public space that is in the class of means, though (for reasons cited) it is indeed possible for an attempt not to involve such an act. And that violates the transfer principle, since the consequent of that principle endorses the criminalization of attempts tout court, whereas we are now limiting that legitimate criminalization to attempts that satisfy the means requirement. But Yaffe seeks a solution to the problem about criminalizing mere thoughts that is consistent with—and indeed is supportive of—the transfer principle. So he cannot appeal to the normative interpretation of the means requirement.

Yaffe's proposed solution is neither to see the means requirement as an independent normative constraint on legitimate criminalization (since that is incompatible with the transfer principle), nor to see a performance of a public act in the class of means to C as a necessary element of an attempt to C (since that is precluded by his theory of the nature of attempts). His proposal, instead, is to see a performance of a public act in the class of means to C as a necessary and special kind of evidence of the attempt to C. This involves three ideas. The first is that performance of an act in the class of means to C really is evidence that the agent is attempting C—in the sense of attempt explained by Yaffe's guiding commitment view. As Yaffe says: “acts taken as means to doing as specified by one's criminal intention provide evidence that the defendant was, in fact, guided by his intention to commit the crime” (234). The second—more difficult—idea is that performance of an act in the class of means to C is necessary evidence for the attempt of C. It is not necessary evidence for the intention to C; we might have, according to Yaffe, other evidence for that. But it is necessary evidence for the claim that the intention to C is guiding the agent in the way essential to attempting C: “nothing other than action in furtherance of intention could serve as” such evidence, according to Yaffe (234). So even though it is possible for there to be an attempt of C in the absence of an act in the public space, a necessary condition for our having adequate evidence that there has been such an attempt is that there be an act that is, in particular, an act in the class of means.

Now, even if we grant these two claims about evidence, we might wonder if they establish what we want. We would be granting that we cannot have evidence of an attempt to C without evidence of an act in the public space that is performed as a means to C. But this does not yet show that what we would be criminalizing, in criminalizing the attempt, itself involves that act. It still seems that what we would be criminalizing is the attempt itself, and all that we have been told is that to be confident that there has been this attempt, we need this evidence. But if the attempt itself is solely a mode of thought, then we can worry that we are still legitimizing criminalizing mere thought, though we are granting that to establish that there has been such thought, we need evidence that involves action in the public space.

Yaffe's answer to this worry is his third idea. The idea is that when there is action in furtherance of the intention in favor of C, that action is not merely necessary evidence that is “correlated” (236) with the guidance that is involved in the attempt. That action is itself part of that in which the crime of attempt consists. As Yaffe says, “the evidence is so closely related to the fact that it evidences, that it makes sense for the law to treat it as an element of the crime” (236).

These three ideas, then, constitute what Yaffe calls his “guidance-based evidentialist approach” (228). The thought is that this approach is compatible with both the transfer principle and the guiding commitment view of the nature of attempts but nevertheless allows us to appeal to the means requirement to block the inference from those two doctrines to the legitimate criminalization of mere thinking.

A weak link is the second idea, the idea that an act in the class of means to a crime is necessary evidence of the attempt to commit that crime. Consider Hacker. Hacker intends to break into a computer system and steal data. If she did that, she would be acting in a way that, I take it, it is legitimate to criminalize. So, according to the transfer principle, it is legitimate to criminalize her attempting to break in and steal the data. But here is how Hacker is attempting to do this: she is thinking hard in her effort to understand certain aspects of relevant computer code, where this thinking is her initial means to her end of breaking in and stealing data. Hacker is the kind of person who does all this thinking in her head. This thinking does not essentially involve an act in the public space, though it is nevertheless thinking that she undertakes as a means to her criminal end: her thinking is guided by her intention to steal the data. So Hacker's thinking is not the sort of act in the public space that is required by the means requirement. Yet it seems that Hacker, in her thinking, is indeed attempting to commit the crime—attempting in the sense of the guiding commitment view—though this is not yet a last-act attempt. She intends to commit the crime, and this intention is guiding her thinking—her intention is what explains why she is thinking through various esoteric issues about the computer code. She is acting in furtherance of this intention to steal the data, and if all were to go “well,” she would steal the data. So we have a case in which there is an attempt to commit the intended crime but there is not, so far, a public act taken as a means to that crime. And Yaffe's view here will be that although there is indeed the attempt, in the absence of a public act in accordance with the means requirement we will not have adequate evidence that there is the attempt.

But suppose Hacker is also a blogger, one who chronicles almost her every thought. Her blog makes it clear that she is indeed thinking hard about the computer code as her way of stealing the data. In writing her blog, she is engaged in an act in the public space. But she does not engage in her blogging as a means to stealing the data, so her blogging act does not satisfy the means requirement. And her goal-directed thinking does not satisfy the means requirement because it is not an act in the public space. Nevertheless, it seems that she is attempting to steal the data by trying to figure out certain features of the computer code and that her blog might serve as good evidence for this.

Suppose now that, alerted by Hacker's blog, the police descend. They stop her before she takes any act in the public space as a means to her criminal end, though not before she is involved in attempting to steal the data by figuring out—in her head—the computer code. If we accept the transfer principle, we should say that it is legitimate to criminalize such an attempt even if does not involve an act in the public space. And it seems that Hacker's blogs might well provide evidence that Hacker has been attempting to steal the data even in the absence of an act that satisfies the means requirement.

Granted, the cited evidence—the writings in the blogs—involved public acts. Our evidence for what we are criminalizing is itself not mere thinking. But—in contrast with the kind of evidence purportedly provided by acting in furtherance of the criminal intention—this evidence is only merely correlated with what we are criminalizing; it is not itself part of what is being criminalized. So what we are criminalizing seems itself to be mere thinking, albeit goal-directed thinking.

Indeed, it seems that we will be led to an analogous conclusion about a paralyzed would-be assassin who is also a blogger. We might have good evidence from his blogs that he attempted to move his finger to pull the trigger, even though—because of the paralysis—there was in fact no act in the class of means. Though this is not to criminalize mere intention, it still seems a problematic conclusion.

So let us reconsider the conjunction of the transfer principle and the guiding commitment view of attempt. One idea would be to adjust the account of the nature of attempts. Here we might try to say that attempts must involve something like substantial steps, where such steps involve acts in the public space. In response I take it that Yaffe would say that this is to build into the metaphysics of attempts a condition that is, at bottom, a purported condition on the legitimate criminalization of attempts. In bringing together the philosophy of action with the concerns of the criminal law, we need to be clear about what is a matter of the metaphysics of various modes of agency, and what are the normatively relevant conditions on legitimate criminalization or other responses of the law to these modes of agency. Examples such as those of a paralyzed attempter or—as I conjecture—the case of Hacker suggest that this appeal to substantial public steps is motivated not by a plausible view of necessary elements of attempt but by a plausible view of necessary conditions of legitimate criminalization of attempts. So we should try to address these issues about legitimate criminalization without undue distortion of the metaphysics of attempt.

I am sympathetic with such a reaction. Attempts are what they are, not another thing. We risk confusion by adding to a theory of the nature of attempts conditions whose intuitive support comes not from a view about that but rather from a view about when it is legitimate to criminalize attempts. We do better to develop an independently plausible account of the nature of attempts and then take due account of important thoughts about legitimate criminalization elsewhere in our overall metaphysical, epistemological, and normative theory.

Granted, there are important action-theoretic concepts that are infused with normative considerations from morality or law. Negligence is an example. And there are hard questions about which action-theoretic concepts fall into this category. But we do well to seek foundational concepts that carve up our agency in ways that do not themselves build in complex and controversial moral or legal evaluations (though we can reasonably suppose that they will involve certain thin rationality norms). We can then proceed explicitly to consider which such substantive evaluations to accept and how to accommodate them within our overall theory. This is a kind of division of philosophical labor between metaphysics, on the one hand, and normative assessment, on the other hand. And as I see it, Yaffe's procedure is a version of this division-of-labor strategy: say what attempts are, and then try to explain the nonlegitimacy of criminalizing mere thoughts elsewhere in our overall theory, including our theory about the epistemology of attempts. On the assumption that intention and guidance are phenomena whose theoretical specification need not involve substantive moral or legal evaluations, we can see the guiding commitment view as being at the heart of Yaffe's version of this division-of-labor strategy.

The problem is that though this division-of-labor strategy is appealing, Yaffe's version of this strategy seems to run into a problem. Yaffe seeks to retain not only his theory of the nature of attempts but also the transfer principle. But this leads to difficulties concerning thought crimes. So perhaps we should reassess the transfer principle.

A natural thought here is that what is true is only a qualified principle along the lines, roughly,Footnote ⁴ of:

Qualified transfer qualifies Yaffe's transfer principle by explicitly adding the condition cited in the means requirement to the consequent of the principle. This avoids the need to defend the means requirement along Yaffe's evidentialist lines, a defense about which I have expressed some concerns earlier in this essay. (See above, pp. 106–107.) And it eschews an effort to defend the means requirement by building its condition into the metaphysics of attempts—a defense that is in tension with the division-of-labor strategy. Instead, qualified transfer builds a normative constraint on the legitimate criminalization of attempts directly into its understanding of the connection between legitimately criminalizing C and legitimately criminalizing certain forms of attempting C.

Of course, Yaffe would accept qualified transfer, since it follows from his transfer principle. But he would also say that there are strong intuitions behind his stronger transfer principle; and that is why he seeks a solution to the problem of criminalizing mere thoughts that does not reject that transfer principle. So if we want to opt instead for something like qualified transfer while rejecting Yaffe's transfer principle, we need to defuse that purported support for the transfer principle.

Now, it is tempting to understand the transfer principle in terms of the metaphor that the legitimacy of criminalization literally transfers from completed C to attempting C. But this is not, at bottom, how Yaffe understands this principle. His idea, instead, is that the transfer principle is true because when it is legitimate to criminalize C, there are underlying commonalities between completed C and attempting C, commonalities that are the ground both for the legitimate criminalization of completed C and for the legitimate criminalization of attempting C. In this sense his principle might better be called a commonality principle.

Yaffe's argument about these commonalities is framed by an important background principle concerning the relation between legitimate criminalization, deserved censure, and legitimate sanction. Here is his principle:

One condition for legitimate criminalization involves deserved censure; a second condition involves legitimacy of sanction. But not every “unexcused and unjustified token of that type of conduct” that deserves censure (as in (1)) needs to be legitimately sanctioned (as in (2)); instead, what is needed is only that some are legitimately sanctioned. Why? Yaffe's main example here is legal immunity (31). A member of a legislature may have performed an “unexcused and unjustified token of” a legitimately criminalized type of conduct, and yet, given her immunity to prosecution, it would not be legitimate for the state to sanction her. This is so even though her conduct deserves censure, as in (1). So legitimate criminalization of a type of conduct requires only that some “unexcused and unjustified token[s] of” that type are legitimately sanctioned.

I want now to describe the general structure of Yaffe's argument for the transfer principle, given (*). To do this, though, I need to adjust (*). (*) explicitly offers only a sufficient condition for legitimate criminalization. However, the use to which (*) is put by Yaffe in his argument for the transfer principle seems to me to require that (*) also provides necessary conditions for legitimate criminalization. So that is how I will understand (*).

That said, here is Yaffe's argument: We suppose, arguendo, that it is legitimate to criminalize C. We infer that versions of (1) and (2) in (*) are true about C.Footnote ⁵ We then argue that given the relation between completed C and attempting C, it will also be true that versions of (1) and (2) are true about attempting C. (To show this, we need to appeal not, of course, to the transfer principle itself—since that is what we are trying to prove—but to relevant normative views about what grounds desert for censure and legitimacy of sanction.) So we infer, drawing again on (*), that it is legitimate to criminalize attempting C. Since this form of reasoning is completely general in its application, we infer that if C is legitimately criminalized, then so is attempting C. And that is the transfer principle.

So there are two crucial steps. First, we need to show that if every unexcused and unjustified token of completed C deserves state censure, then every unexcused and unjustified token of the type attempting C deserves state censure. Second, we need to show that if some unexcused and unjustified tokens of completed C are legitimately sanctioned, then some unexcused and unjustified tokens of the type attempting C are legitimately sanctioned.

Yaffe's argument for the first step draws on the plausible normative view that unexcused and unjustified conduct deserves censure just in case it is a product of a faulty mode of recognition of or response to relevant reasons (38). With this normative view in hand, he then appeals to his theory of the nature of attempts—attempts involve intention and guidance—to argue that if completed C would involve such a fault, then so would attempting C. So if unexcused and unjustified tokens of completed C deserve state censure, so do unexcused and unjustified tokens of attempting C.

Yaffe's argument for the second step draws on the thought that “it is possible to get arbitrarily close to completion of a crime without actually getting there” (35). He reasons that if some unexcused and unjustified tokens of completed C are legitimately sanctioned, then at least some unexcused and unjustified tokens of the type attempting C—in particular, those that get arbitrarily close to a completion that would be legitimately sanctioned—are legitimately sanctioned. This draws on the plausible normative view that the legitimacy of sanction cannot depend on the tiny difference between a completed C and an attempt that gets arbitrarily close to that.

Given both these steps we infer that if versions of (1) and (2) in (*) are true about C, then versions of (1) and (2) are true about attempting C. So we conclude that the transfer principle is true.

This is a fascinating line of argument, but I think I see a problem. Let us partition the type attempting C into two disjoint subtypes: attempts that involve an act in the class of means—for short, means-act attempts to C—and attempts that do not involve such an act—for short, non-means-act attempts to C. Suppose that all of the tokens of attempting C that are legitimately sanctioned are tokens of the subtype means-act attempt to C. Granted, in making this supposition we are drawing on normative views about the legitimacy of sanctions; but, as we have seen, such normative theorizing is already at work in Yaffe's use of (*). Given this supposition, we can infer that while the type means-act attempt to C will satisfy condition (2) of (*), the type non-means-act attempt to C will not satisfy that condition. This is because we are supposing that no tokens of the type non-means-act attempt to C are legitimately sanctioned. But then there seems pressure to deny that it is legitimate to criminalize the more general type, attempting C. After all, that more general act-type includes the subtype non-means-act attempt to C. But since no tokens of that subtype are legitimately sanctioned, it is not legitimate to criminalize that subtype, according to (*).Footnote ⁶ But in criminalizing the more general type attempting C we are criminalizing both the subtypes in our partition. So there is pressure just to criminalize the subtype means-act attempt to C. But to narrow in this way the type of attempt that it is legitimate to criminalize is to give up the transfer principle.

A response to this challenge that is implicit in Yaffe's discussion draws on a footnote qualification to (*). As noted, given a type of conduct, condition (2) of principle (*) requires only that “some of the unexcused and unjustified tokens of that type of conduct are legitimately sanctioned by the state.” In his footnote qualification, however, Yaffe adds that “this principle should be understood to be saying that . . . the legitimacy of sanction of some [tokens], derives from the fact that the tokens are tokens of the type” (32, n.11). So, for a given act-type, what condition (2) requires is that “some of the unexcused and unjustified tokens of that type . . . are legitimately sanctioned” because they are tokens of that type. This means that to apply condition (2), we need to know not only what tokens are legitimately sanctioned but also from what type of act (of which this is a token) this legitimacy of sanction derives.

So let us return to the types: attempting C, means-act attempt to C, and non-means-act attempt to C. We are supposing that all tokens of attempting C that are legitimately sanctioned are tokens more specifically of the type means-act attempt to C, and none is a token of the type non-means-act attempt to C. So when we legitimately sanction these tokens, we need to ask whether the legitimacy of the sanction “derives from the fact that” it is a token of attempting C or from the fact that it is a token of the more specific type means-act attempt to C?

This question is not about why the token act deserves censure. We can agree with Yaffe that it deserves censure because it is an attempt to C (given that completed C would deserve censure). But the issue now is why this token act is legitimately sanctioned. Is it because it is an attempt to C, or is it rather, and more specifically, because it is a means-act attempt to C? And here it seems to me that our answer will depend on complex normative reasoning, not primarily on views about the metaphysics of attempts or about evidence for attempts. And there is a normative case for thinking that it is the narrower act-type—means-act attempt to C—that is the basis for the legitimacy of the sanctions. So there is room for insisting that it is because the tokens are tokens of the narrower type, means-act attempt to C, that they are legitimately sanctioned. But if we say that, then we are rejecting the transfer principle (as well as the inference from (*) to the transfer principle). Instead of the purported commonality between completed C and attempting C that is at the bottom of the transfer principle, we are now—in part on the basis of normative reflection—highlighting a different commonality, one between completed C and a means-act attempt to C.

Of course, every legitimately sanctioned token of the type means-act attempt to C is a token of the type attempt to C. But it does not follow that the legitimacy of the sanction of that token “derives from the fact” that it is a token of the type attempt to C. And we have seen reason to suppose that there will be cases in which this legitimacy of sanction does not derive from that fact.

This is compatible with accepting Yaffe's idea that if unexcused and unjustified tokens of completed C deserve state censure, then so do unexcused and unjustified tokens of attempting C. What leads to this rejection of the transfer principle are normatively plausible views not about desert of censure but about legitimacy of sanction. So we are respecting Yaffe's effort not to “cleave legal censure away from moral desert” (222).

Return to Hacker. In her attempt—her goal-directed thinking—to steal the data, she fails to be responsive to relevant reasons, just as does the successful data thief. So we can infer that she is as subject to censure as is the successful data thief. And her attempt is a token of the type attempted data theft, some tokens of which (those that satisfy the means requirement) are legitimately sanctioned. So we are in danger of arriving at the problematic conclusion that such attempts are legitimately criminalized. We avoid this conclusion by holding on to the normative view that it is not legitimate to sanction attempts that do not satisfy the means requirement (even though such attempts may deserve censure). We then see this as supporting an interpretation of condition (2) of (*), as it applies to attempts, that supports only the legitimacy of sanctioning (and so criminalizing) the narrower type, means-act attempt to steal. And that will both lead us to reject the transfer principle and make room for something like qualified transfer.

When we do this, we are returning to the normative interpretation of the means requirement. We are seeing the requirement of a means-act as a substantial normative condition on legitimate criminalization of attempts, because it is a condition on legitimacy of sanctions of an attempt. We are doing this without distorting our account of the nature of attempts and so we are respecting the division-of-labor strategy. And we are not trying to make an evidential argument do work that it is unclear it can do. The price we pay, however, is that we cannot retain the unqualified transfer principle.

What, then, to say about Yaffe's thought that the transfer principle “is reflective of a deeply entrenched element of our common-sense moral thought”? It seems to me that the element to which Yaffe alludes is best seen as a view about desert of censure, roughly: if you would deserve censure for completed C then you deserve censure for attempting C. This is the idea (supported by his theory of the nature of attempts) behind Yaffe's use of condition (1) in (*) as part of his argument for the transfer principle. And I do not challenge that. What I do challenge, on normative grounds, is a thought that lies behind Yaffe's application to attempts of condition (2) of (*). This is the thought, roughly, that if it would be legitimate to sanction a completed C, and if a particular attempt to C is legitimately sanctioned, then the legitimacy of the sanction of that particular attempt to C “derives from the fact” that it is a token of the type attempt to C. And if we reject this latter idea—since sometimes the legitimacy of sanction (if not the desert of censure) “derives from the fact” that it is a token of the narrower type means-act attempt to C—we will be led to reject the transfer principle.

But what, after all, is the normative argument for the view that the state should not sanction mere thoughts, even if they are goal-directed, even if they really do constitute attempts, and even if we really do have good evidence of such attempts? This is a hard question. But it is the question we need to ask in order to understand the ground for our rejection of “thought crimes” and the impact of this on the law of attempts. And I am skeptical that we can answer this normative question solely by appeal to the metaphysics and epistemology of attempts—which is not to deny that we also need accounts of this metaphysics and epistemology.

This is in tension with Yaffe's efforts to get more normative mileage out of the metaphysics and epistemology of attempts. But even if I am right to be wary of this as overreaching, we are much in Yaffe's debt for providing a deep and systematic framework within which we can examine this and other related issues about attempts in a clear way, one that promises progress in this difficult but important terrain.