R. A. Duff
Many criminal offenses are so defined that their commission requires the occurrence of the primary mischief or harm with which the law is concerned. The mischief with which offenses of homicide are primarily concerned is the death, or unlawful killing, of a human being; murder and other species of criminal homicide are so defined that they are committed only when someone is killed. The primary mischief with which the offense of criminal damage is concerned is the nonconsensual destruction of or damage to property; the offense is so defined that it is committed only if such destruction or damage actually occurs. We can call such offenses “consummate” (see Husak 1995): their commission involves the consummation of the mischief at which the law is aimed.
Systems of criminal law typically also criminalize attempts to commit consummate offenses: alongside the range of specific consummate offenses, there is a general offense of attempting to commit any such offense (above a minimal level of seriousness). If I try to kill someone, without justification or excuse, I am guilty of murder if I succeed, and of attempted murder if I fail; if I try unlawfully to destroy another’s property, I am guilty of criminal damage if I succeed, and of attempted criminal damage if I fail. Criminal attempts form one category of “nonconsummate” offenses. An offense is nonconsummate if the conduct that constitutes it is criminalized because of its relationship to some primary harm or mischief, but the commission of the offense need not involve the occurrence of that primary harm or mischief: attempts are criminalized because the conduct that constitutes a criminal attempt is related to a primary harm or mischief (as an attempt to cause it); but the commission of an attempt does not require the occurrence of that primary harm or mischief. Other general nonconsummate offenses are incitement and conspiracy: it is an offense to incite another to commit, or to agree with another to commit, a consummate offense, even if that offense is not actually committed (see Simester and Sullivan 2010: 285–328). But we will focus here on criminal attempts.
Three initial questions arise about criminal attempts. Why should we criminalize attempts to commit crimes? How should a criminal attempt be defined? How (how severely) should criminal attempts be punished? Each of these questions has occupied theorists of criminal law, and leads into larger issues in philosophy of action and moral philosophy; each generates further questions, which will emerge more clearly if we look initially at the question of definition.
We cannot assume that in defining offenses the criminal law will or should always use terms in their precise extralegal meaning: the law’s proper purposes might be better served by creating a legal definition of attempt which does not precisely match our extralegal concept. But we can assume that there must be (or have been) at least some quite close relationship between legal and extralegal conceptions, and a brief examination of the extralegal concept will bring out some of the further questions raised by the law of attempts.
Two aspects of the extralegal concept are particularly important for present purposes. First, attempts require intention: I attempt or try to Φ only if I intend to Φ. Furthermore, the intention must be a “direct” intention: it is not enough that I act with what Bentham (unhelpfully) called “oblique” intention (Bentham 1781: ch. VIII.vi; 2000: 70), consisting in the certain belief that my action would bring about the relevant result; it must be at least part of my purpose to bring that result about, even if I intend to bring it about only as a means to a further end.
(We should note two complications here. One is that the distinction between intended results and foreseen side effects comes under pressure when the connection between what is intended and what is foreseen is so close that they seem inseparable: if I intend to blow up an aircraft, I must surely also be taken to intend the deaths of those whom I know to be on board, even if my aim is only to destroy the aircraft; if I fail in my endeavor, I have surely attempted to kill them. What makes such cases plausible as cases of intention, however, is not the mere fact of certain foresight, but the way in which the foreseen effect is part, rather than a further consequence, of what is intended: if I intend to explode the aircraft, I intend to destroy the whole thing—which includes its (human) contents. Whatever my further purpose, my intended means encompass that destruction (see, generally, Yaffe 2010: 119–24; Simester 1996). The other complication concerns what must be intended when the consummate offense includes circumstantial elements that are essential to its wrongful character. A person commits rape when he sexually penetrates another person without that person’s consent, and is reckless as to whether the other person consents; he need not intend the penetration to be nonconsensual, or know that it is nonconsensual. But what if he tries but fails to sexually penetrate the other, non-consenting, person? For this to constitute attempted rape he must at least intend that penetration, but must he also intend that penetration to be nonconsensual, or at least know that it is not? Or is it enough that he is reckless as to whether it is consensual? It seems plausible to hold that the attempt should not require intention as to that aspect of the consummate offense: he attempts rape if the other person is actually non-consenting and he is at least reckless as to that possibility. We could then say that whilst an attempt requires intention as to the central consequential aspects of the consummate offense, it does not require intention as to every aspect. However, we then face the challenge of distinguishing those aspects of the consummate offense as to which intention is required from those as to which it is not—a challenge that we cannot take up here (but see Duff 1996, ch. 1; Yaffe 2010, chs. 4–5).)
Second, attempts require conduct that comes quite close (at least as the agent envisages it) to the completion of the intended enterprise. I might form the intention to rob a bank, make detailed plans for the robbery, recruit colleagues to help me, buy the necessary equipment and watch the bank to see when and how I can best carry out my plans: I am preparing to rob the bank, but not yet attempting to rob it. Even when I steal the car we will use, and turn on the engine, I am still preparing rather than attempting the robbery: I am attempting to rob the bank only once I move beyond preparation to the actual execution of the intended robbery, though the boundary between preparation and execution is far from sharp or clear. On the other hand, the extralegal concept of attempt reaches further than “complete” or “last-act” attempts in which the agent has done the last thing there is for him to do to complete the action: as I enter the bank with a drawn gun, shouting “This is a holdup,” I am attempting to rob the bank, although there is still more that I must do to complete the robbery. If a would-be assassin is captured as he aims his gun but before he actually pulls the trigger (which would be the “last act”), it is still true that he attempted to kill his target.
(I can be attempting to Φ even if what I am doing is in fact remote from actually Φing: if I put sugar in your cocoa, believing it to be arsenic and intending to poison you, I am attempting to kill you, even though putting sugar in someone’s cocoa is not close to the actual commission of murder. This counts as an attempt because, if the facts had been as I believe them to be, I would have been executing my plan to kill you. It is not clear, however, whether the agent’s beliefs are always conclusive. If I fire at what is in fact a tree stump, believing it to be the person I want to kill (who is actually miles away), is it true that I am attempting to kill him? Such examples raise the question of whether attempts must have some actual connection to the real world, rather than merely what the agent believes to be a connection (see, further, Duff 1996: 219–32).)
Attempting to Φ therefore involves action that constitutes the execution of an intention to Φ. Some philosophers have portrayed trying or attempting as a kind of identifiable (usually mental) occurrence: trying to Φ is a distinct activity by means of which I hope to bring it about that I Φ (see, e.g., O’Shaughnessy 1973; Armstrong 1973). All action can then be analyzed into a “trying” and such further effects as that trying produces. But this is not what we mean when we talk about “trying” or “attempting”: rather than picking out some distinctive element of an action, such talk constitutes an adverbial, contextualizing qualification of the action verb that it modifies. To say that A is trying to Φ, that is, is to set A’s conduct in a certain context. It might be to connect what A is doing to the Φ-ing that forms its aim: we explain why A is running so fast down the road by saying that she is trying to catch a train; the trying does not involve anything more or other than the running for that purpose. But talk of trying also sets the action in a context of failure, doubt or difficulty: if A has been unsuccessful in executing her intention to Φ, or we are not sure that she will successfully execute it, or its execution is arduous or difficult, we say that she is trying to Φ rather than that she is Φ-ing (see Heath and Winch 1971; Duff 1996, ch. 10).
Once we recognize these aspects of trying in the ordinary extralegal meaning of the concept, our initial three questions about criminal attempts multiply into a larger set of questions.
If we start with the category of consummate offenses, which actually bring about the primary harm or mischief at which the law is aimed, our first question must be why we should extend the criminal law any more widely than that. To avoid begging too many questions, we can put this question in the language not of attempts, but of nonconsummate crimes, whose commission does not require the occurrence of the primary mischief at which the law is aimed. Attempts constitute just one kind of nonconsummate crime; others will be noted below. We criminalize consummate crimes such as homicide and criminal damage: why should we also criminalize nonconsummate versions of those crimes, such as failed attempts to kill or to damage another’s property? Such attempts can of course cause other kinds of mischief that properly concern the criminal law: they can, if known, cause fear or anxiety, or disrupt social order (see Becker 1974). But that gives us no reason to criminalize attempts that are not noticed at the time, and does not seem to capture the wrongfulness of attempts.When we convict and punish someone for attempted theft, we are surely not simply convicting and punishing her for the fear, anxiety or disruption that her failed attempt might have caused.
Our further questions arise when we ask why, if we are going to criminalize nonconsummate crimes, we should do so through a law of attempts, and whether such a law should use “attempt” in its ordinary extralegal meaning.
One question is why we should extend the law as widely as the concept of attempt suggests. The most modest extension of the law’s scope beyond consummate offenses would capture only complete or “last-act” attempts, when the agent has done all that there is for her to do to complete the crime. As we saw, the extralegal concept of attempt is broader than that, but why should the criminal law go so far? Why should it not cover only last-act attempts (apart from the difficulty, in some cases, of identifying a “last act” that does not actually complete the crime)? One important consideration, to which we will return later, is that until she has completed her attempt the agent still has a chance to abandon her criminal enterprise voluntarily—a “locus poenitentiae.” So why should she be criminally liable before she has definitively and irrevocably committed herself to the commission of the crime (see Alexander and Ferzan 2009, ch. VI)?
If we can justify extending the reach of the criminal law beyond consummate offenses, and even beyond “last-act” attempts, further questions arise about why we should not extend it further than is suggested by the extralegal concept of attempt (and further than our existing laws extend it). Why should we not also criminalize other kinds of nonconsummate offense that are related to a primary mischief but that do not amount to what extralegal usage would count as an attempt? Our existing laws include a range of other types of nonconsummate offense, three of which are worth noting here.
First, our attempt laws typically follow extralegal usage in requiring an intention to commit the complete crime (although intention is sometimes interpreted to include “oblique” intention). They also criminalize some kinds of reckless endangerment: conduct that is not intended to cause harm, but that creates an unjustified risk of harm and whose agent acts recklessly as to that risk. Some such offenses are relatively specific as to the type of risk whose creation is criminalized: they criminalize risk-creation in the course of particularly dangerous activities, or by agents with particular responsibilities, or that threaten particularly vulnerable kinds of potential victim (see Smith 1983; Duff 2007: 161–64). Others are relatively general: the American Model Penal Code, for instance, criminalizes conduct that “recklessly endanger[s] another person” by creating a risk of death or serious injury, or that creates a risk of a “catastrophe” that would affect “the safety or property” of many (§§ 211.2, 220.2(2)). But none are as general as the law of attempts: whilst it is an offense to attempt to commit (almost) any consummate crime, it is not similarly an offense to take a reckless risk of causing any result whose actual causation would be criminal. I commit criminal damage if I damage another’s property either intentionally or recklessly—if I throw a stone that breaks my neighbor’s window, whether I intended to break it or recklessly took the risk that I would do so. If my brick misses the window, I am guilty of criminal attempt if I threw it with the intention of breaking the window, but have committed no offense if my throwing was merely reckless as to the risk that it would break the window. This is puzzling. We convict both the deliberate and the reckless stone-thrower of the same offense if they actually cause damage, and convict the deliberate thrower if his stone misses: but whatever our grounds for convicting him, whether they are that he is no less culpable than one whose stone hits its target, or that he shows himself to be as dangerous, or that we want to deter people from such dangerous actions, similar reasons seem to warrant criminalizing the reckless agent whose stone misses. If we are to have a general nonconsummate crime, why should it be a crime of “attempt,” requiring intention, rather than a broader crime consisting in conduct that comes sufficiently close to the consummate offense, and a “fault element” identical to that required for the consummate offense? If criminal damage requires conduct that damages another’s property, and either an intention to cause such damage or recklessness as to the risk of causing it, why should we not define “nonconsummate criminal damage” as consisting in conduct that comes sufficiently close to damaging another’s property, done with the intention or the recklessness that would make the agent guilty of criminal damage if he actually caused damage?
Second, although attempts do not require a “last act,” they do (in extralegal usage) require conduct that constitutes execution of the intended action, rather than merely preparation for it. Some attempt laws preserve this feature of extralegal usage: English law requires conduct that is “more than merely preparatory to the commission” of the intended offense (Criminal Attempts Act 1981, s. 1), which courts have interpreted to require that the agent have “embarked on,” or be “in the process of committing” the offense ([Gullefer