The argument to this point is that double jeopardy “offense” can be understood only in terms of the “life or limb” blameworthiness the legislature intends to create. We must next understand when the legislature intends to create distinct blameworthiness and thus different offenses. Ultimately, we need a test that determines whether the blameworthiness in two statutes, or two violations of a single statute, is singular or distinct. If it is distinct, no double jeopardy bar exists to imposing a punishment for each distinct blameworthiness, and no bar exists to seeking that punishment in two proceedings.
“Distinct” in this context means something other than “different.” The blameworthiness of felony murder is different from that of premeditated murder, for example, because the former includes commission of the underlying felony. But that difference in blameworthiness does not necessarily suggest that the legislature meant the blameworthiness of first-degree felony murder to be distinct from that of first-degree premeditated murder. If this were the best reading of legislative intent, a killer could be tried and convicted twice for first-degree murder for a single killing, hardly a satisfying outcome.
Many nineteenth-century courts used a blameworthiness test, albeit under the guise of comparing the “gravamen” of offenses. As Susanah Mead has observed, this test “had the potential for striking a balance between the act and offense, because it focused on the conduct of the defendant as it related to the social interests sought to be protected.”1 Examining the proscribed conduct in light of the social interests is just another way of asking what blameworthy conduct the legislature sought to proscribe.
I argue that singularity of blameworthiness is best understood as the convergence of harm singularity and act singularity. This, of course, requires an account of “act” and an account of “harm.” It is my basic thesis that legislatures write statutes with some form of this act/harm singularity as a background assumption for the dividing line that separates distinct blameworthiness from singular blameworthiness.
The standard account of criminal liability requires (1) an act, (2) mens rea, (3) attendant circumstances, and (4) harm.2 Because it is the goal of the criminal law to prevent harm—consider the traditional common-law crimes of murder, robbery, and rape—we can usefully begin by equating blameworthiness with the causing of harm. But what does it mean to cause harm? Joel Feinberg devoted four excellent volumes to the problem of defining the moral limits of the criminal law,3 and his definition of harm is useful: a wrongful setback to the interests of a person.
Harm is not a sufficient condition for blameworthiness. Feinberg explicitly states his account of harming to include the mens rea and act requirements—one must act to cause the harm and be sufficiently responsible for the act and harm.4 Focusing on the act requirement for the moment, some misfortunes in life do not permit us to ascribe blameworthiness to a human actor because no one acted to bring them about. Random chance and Mother Nature do not give rise to wrongful setbacks.
Moreover, some acts are harmful only because of the circumstances that exist when they are committed—for example, consensual sexual intercourse with a person under a certain age or who is unconscious and thus incapable of consenting. When the presence of the circumstance is what gives rise to the harm, the two cannot be separated conceptually. The “victim” has suffered no setback to her interests if she is of age and gave her consent. Of course, some circumstance elements are not connected to harm—for example, the requirement in some federal crimes that a state line be crossed.
It seems that blameworthiness is related to act and harm (with harm understood to include the circumstances that contribute to the harm). This is hardly surprising, but still incomplete. Some acts that cause harm are not blameworthy because the actor did not have a culpable mental state. If D bumps V accidentally, D is not blameworthy for V’s fall. But if D intended to kill V while making it look like an accident, we would say that D had a culpable mental state that proves the felony of assault with intent to kill.
As one would expect, the standard account of criminal liability thus states the conditions of blameworthiness: act, mens rea, harm (with harm understood as including the circumstances that give rise to harm). Stating an account of criminal blameworthiness is not enough. Because any double jeopardy project is ultimately about singularity, we need an account of how these elements combine to create singular or distinct blameworthiness. One place to begin that project is to focus on the singularity of acts, in effect returning to the nineteenth-century gravamen-of-offense test. As Susanah Mead described the test: “If the gravamen, or principal act, necessary to violate each statute was the same, then the offenses were the same for double jeopardy purposes.”5 The focus was on the “principal act” and thus on whether the two offenses manifested what I call “act singularity.”
In considering the role that “act” plays in criminal blameworthiness, we can safely ignore the controversial possibility that criminal liability can arise in the absence of an act, with act defined as a volitional movement.6These questions are important in substantive criminal law but not in formulating a basic account of when legislatively proscribed criminal blameworthiness should be considered distinct.
This chapter draws heavily from action theory, a metaphysical enterprise that seeks to state the conditions under which one act is distinct from another. This is a surprisingly difficult task, one that has given rise to a skeptical view of its utility. Recall Larry Simon’s critique involving the number of “acts” contained in the “act” of shaving.7 But, as Michael Moore has recognized, we should not dismiss action theory because it fails to provide self-evident answers.
Moreover, Simon’s use of action theory confuses acts that exist in the world with blameworthy acts proscribed by the legislature. A double jeopardy theory does not need an account of when acts in the universe are distinct; it merely needs an account that determines how many legislatively proscribed acts have occurred. While this is not an easy task, it is easier than the metaphysical enterprise.
The necessary relationship between “act” and “offense” has for centuries been one of the irreducible principles in substantive criminal law. The common law has long held that criminal liability requires proof of a voluntary act that is proscribed by law.8 The last sentence states two different kinds of act requirements. Proof of a voluntary act requires that an actor have committed a particular act at a certain time and in a certain place. The requirement that the act be proscribed by law is satisfied by showing that the particular act fits the act description contained in the criminal prohibition.
Action theory draws the same distinction between descriptions of acts, called universals or act-types, and particular acts in the world, called event-particulars or act-tokens. Act-types are found in statutory proscriptions. The offense of larceny consists, in part, of the act-type of taking and the act-type of carrying away. Each particular instance of an act-type is known as an act-token. The larceny at the Watergate Hotel was an act-token.
Some form of action theory is deeply embedded in existing judicial approaches to same offense. The form it takes may constitute bad metaphysics, as Moore occasionally charges,9 but action theory is part of our intuition about blame and responsibility. For example, no one doubts that the statutory offense of larceny from a person is the same offense as the statutory offense of larceny from a person, as long as the same larceny (the same taking) underlies both charges. Right away, we see an action theory problem. What if the actor removed the victim’s wallet and returned five minutes later to take his watch? Are there two takings, two offenses, or only one? Courts have struggled with this issue, as we see later, but at least they ask the right set of questions. Did the actor remove the wallet and watch by one “act” of larceny?
Courts have failed to see that action theory can also provide the answer to the equally difficult question of whether different statutory prohibitions define the same offense. Early courts struggled with the question of whether burglary and larceny are the same offense when the larceny is used to prove the requisite intent in burglary. This issue, superficially at least, looks different from two charges of larceny. In the burglary/larceny context, two statutory formulations exist and can be laid side by side. The difficulty, however, is the lack of a way of comparing the offense descriptions for “sameness.” Courts, using a variety of analyses, eventually decided that burglary and larceny were not the same offense. But the real reason these offenses are not the same is that the burglary “ends” when the threshold is crossed; the larceny is therefore a subsequent and different act.10
Now compare the statutory offense of larceny from the person with that of larceny at night. There are two questions. First, are the statutory act-types the same? If so, are the different charges based on different act-tokens? Obviously, committing larceny from the person on Tuesday is a different offense from committing larceny at night on Friday. The reason is that different act-tokens occurred. Thus, whatever the relationship between statutory act-types, different act-tokens always render offenses different. This must be true because two act-tokens of the very same act-type are two offenses—for example, two robberies, two larcenies, two rapes.
Similarly, if the two statutory act-types are different, they proscribe different offenses, and we need not worry about whether separate act-tokens occurred. Rape is a different act-type from murder, and it should not matter (I later argue) that the act-token of rape also caused the victim’s death (arguably the same act-token).
If, however, the statutory act-types are the same and only a single act-token of that act-type occurred, there is singular criminal blameworthiness and thus only one double jeopardy offense. Let us assume, for the moment, that larceny at night is the same act-type as larceny from a person. On that assumption, if a thief takes property from the victim’s pocket at night, the account developed here presumes that it is one offense even though proscribed by different statutes. Blockburger presumes otherwise, because the distinct element in each larceny offense makes them different whether or not those distinct elements manifest blameworthiness.
Blockburger permits judges to answer the same-offense question without thinking about substantive sameness. It is a mechanical solution. The act-type that we call larceny could, by virtue of additional trivial descriptive elements, be held to be several different offenses. This seems wrong. If “offense” is a substantive concept, which it is in criminal law, it is a mystery why courts and commentators rely on mechanical tests to measure the double jeopardy sameness of different statutory offenses. Why not rely on a test of substantive sameness? Moreover, why not use the same analysis that helps sort out how many violations of a single larceny statute have occurred?
There is more work to be done. I must offer a way to determine what parts of a statutory crime are act-types; statutes often include circumstances and results, which must be kept separate from act-types when constructing the likely statutory blameworthiness. If all parts of a statute are blameworthy act-types, then my account would be indistinguishable from that in Blockburger. The next section offers a method of distinguishing act-types from circumstances and results.
Michael Moore develops a comprehensive theory of action in his excellent book, Act and Crime. Part of his insight is that act singularity is the best account of same offense. In my view, Moore’s approach takes insufficient note of harm singularity, a point that I develop later in this chapter, but Moore is otherwise very helpful.
The difficulties in applying act singularity, while real, are manageable. Consider, for example, the aggravated first-degree murder offense of killing a police officer, which entails (1) an act (2) done with murder mens rea, (3) which results in the death (4) of someone who is a police officer. What is the proscribed act-type in this offense? The death can be described as a result, and the victim’s status can be described as a circumstance. On the other hand, it is plausible to say that the act is killing a human being (murder is defined that way) and, from there, not incoherent to say that the proscribed statutory act is killing a human being who is a police officer.
The Model Penal Code distinguishes between conduct and what it calls “attendant circumstances”11 but offers no definition of either concept. Paul Robinson and Jane Grall criticize the Code for this omission and propose a narrow reading of “conduct” and thus a correspondingly broader definition of “result” and “circumstance.”12 Under the Robinson-Grall approach, conduct is limited to “actual physical movement.”13 The effect of this narrow formulation, as they recognize, is to make conduct “a relatively unspecific and unimportant aspect of an offense.”14 Indeed, many acts are non-blameworthy, or much less blameworthy, when taken out of context of their circumstances or results.
For my purpose in measuring legislatively proscribed blameworthiness, I need a broader measure than “physical movement.” Fortunately, I do not need to draw a precise line between acts and the context that makes the acts blameworthy. Rather, I need simply to identify which elements are arguably acts and then determine whether those elements create or add to blameworthiness. My goal is to distinguish between blameworthy act-types proscribed by statutes and other elements that (1) are clearly not act-types or (2) have little or nothing to do with blameworthiness.
We can apply this methodology to the aggravated murder statute as follows.15 First, ask whether the status element is even arguably an act-type. If the answer is yes, the next question is whether it is a blameworthy act-type. Mens rea elements do not constitute act-types and can be disregarded at the first stage. To say that X does act A intentionally is to describe not A but X. It seems unlikely that the status of a victim is an act-type; the standard account of act excludes all properties that cannot be brought about by the actor. When X kills Y, X can do nothing by an act to bring about the property that Y is a police officer.
Assuming, arguendo, a nonstandard account of act that includes the victim’s status, the next inquiry is whether status adds measurably to the blameworthiness that would be present without the element.16 Any element that does not add to the blameworthiness is unlikely to have been considered blameworthy by the legislature. Removing the victim’s status would leave killing a human being with murder mens rea. A useful way to approach the question of whether an element adds measurably to the proscribed blameworthiness is to apply Feinberg’s harm principle. In a homicide, the wrongful setback to a person’s interests occurs by being killed, not by being a police officer who is killed.
To be sure, the legislature created aggravated murder by naming a particular victim status, but the legislative purpose is presumably to deter this form of murder, or enhance the incapacitative effect, by increasing the penalty. It requires an unusual morality, or an unusual status, to find significantly more harm on the basis of the job classification of the victim.17 Thus, the blameworthy act-type in the aggravated murder statute seems likely to be the killing of a human being, without regard to the status of the victim.
This produces the satisfying result that aggravated murder and nonaggravated murder proscribe the same blameworthy act-type (and thus the same offense); the victim’s status as a police officer is not an act-type and, if it were, is not necessary to the blameworthiness. Tracking Blackstone, all homicide statutes proscribe the same blameworthy act-type.18 This result is satisfying because, otherwise, we would be forced to say that killing a police officer entails two blameworthy act-types: killing and killing a police officer.
Michael Moore agrees that courts should disregard act-types that are not, in his words, “morally salient” when comparing act-types for double jeopardy identity. To use Moore’s example, drunk driving, driving without a license, and driving an overweight vehicle all share the act-type of driving. As there is nothing about driving qua driving that is morally salient, there is no sameness of morally salient act-types in drunk driving, driving without a license, and so on.19
For a more difficult example, consider forcible rape and statutory rape. Forcible rape is typically defined as intercourse “against the will and without the consent” of the victim. Statutory rape is sex with a person younger than a certain age. A useful way of isolating the issues when comparing act-types is to subtract the blameworthy act-type that the two offenses have in common and then to look at each left-over element to see if it (1) is an act-type that (2) defines a Feinberg harm distinct from the common act-type.
The threshold problem in this example is deciding whether the offense descriptions have a blameworthy act-type in common. They obviously have sex in common, but sex qua sex is not blameworthy. Considered in the abstract, it is not a wrongful setback to anyone’s interest. These offenses share the act-type of wrongful sex, which is presumably blameworthy but too general to be useful. In constructing the blameworthy act-type from the statutory language, the act-type must be made sufficiently precise to manifest the particular blameworthiness against which the statute is directed. In rape, the relevant blameworthiness is sex against the will of the victim; in statutory rape, the relevant blameworthiness is sex with an underage person. Using these descriptions as the blameworthy act-type, the two offenses have no blameworthy act-type in common and must be different offenses.
It can be argued that the age element in statutory rape is a proxy for lack of consent.20 On this view, the blameworthy act-type is sex without consent in both statutory rape and forcible rape. While the argument is not without merit, it forfeits the benefits of using the legislative language to construct the act-type. Just as we should reject broadly defining the act-type as wrongful sex, we should also reject rewriting the statutory language in constructing the act-type. The legislative justification for statutory rape might be lack of capacity to consent, but what the legislature chose to proscribe is sex with an underage victim.
Treating rape and statutory rape as different act-types and thus different offenses seems right, viewed though the lens of probable legislative intent of the relevant harms. The actor who has forcible sex against the will and without the consent of an underage person has committed two harms simultaneously. The act-types should be construed to reflect these different harms.
A useful contrast is the Blockburger pair of offenses: selling narcotics not in or from the original package and selling narcotics not pursuant to a written order of the purchaser. If we could say that selling narcotics is an innocent activity, then the Blockburger offenses would be just like Moore’s various driving offenses and my example about sex offenses. We would have to add to the innocent act-type the element that creates blameworthiness, which would in this case be the two circumstance elements (selling not from the original package and selling without a written order of the purchaser). Once we do that, of course, neither act-type is included in the other, which would endorse the Court’s view that the offenses are different.
To support this interpretation, we might note that selling narcotics is not a mala in se offense and is not always prohibited by statute. Narcotics are sold every day in hospitals and pharmacies. But selling narcotics is qualitatively different from driving or having sex. The last two activities are lawful unless proscribed because of a dangerous or harmful way of doing the act-type. Selling narcotics, whatever its deontological moral status, has been unlawful in this country for decades unless expressly permitted. The difference is between lawful unless proscribed because of a circumstance and unlawful unless permitted because of a circumstance. This difference locates the blameworthiness in the act-type of selling narcotics, rather than in the circumstances attending the act-type, as was the case with the driving and sex offenses. On this reading, the blameworthy act-type is selling narcotics without statutory permission.
Moreover, the Blockburger narcotics offenses are unlike the driving or sex offenses because the additional elements in the narcotics offenses are not themselves blameworthy, as they are in Moore’s driving example and my rape example. If selling narcotics is an innocent activity, it is difficult to conceive why it would be criminally blameworthy to sell narcotics not in the original package. It is not typically criminal to buy a package of items and sell the individual items from the package or to sell the package without a written order of the purchaser. In the driving offenses and the rape offenses, the additional elements are what create the blameworthiness. Not so in the Blockburger offenses.
Subtracting the blameworthy act-type of selling narcotics from each Blockburger offense leaves “not in or from the original stamped package” and “not pursuant to a written order.” Neither of these residues is an act-type, nor are they blameworthy. There was, then, a single blameworthy act-type in Blockburger, which makes it less likely that Congress intended cumulative punishments.
On the account developed here, blameworthiness is singular when the criminal statutes define a single blameworthy act-type. One obvious and uncontroversial way that statutes can define a single blameworthy act-type is for a series of offenses to be arranged in a hierarchical manner depending on variations in mental state, result, or the scope of the act-type. This can be called “alternative” forms of blameworthiness—giving the prosecutor, judge, and jury a range of choices for a single conviction—rather than distinct blameworthiness that can be punished or prosecuted separately. The next section begins with these hierarchies and then expands the analysis to other, less obvious examples of when blameworthiness is alternative rather than distinct.
First-degree murder, second-degree murder, and manslaughter are usually defined so that the salient difference is the actor’s mens rea. Adapting Model Penal Code terminology to the common-law homicide offenses and oversimplifying for ease of explanation, first-degree murder could require purpose, second-degree could require a knowing state of mind, and manslaughter only recklessness as to the risk. This sort of hierarchy naturally produces Blockburger included offenses—at least if one assumes, as the Model Penal Code does, that proof of a higher culpable state always proves the lesser.21
As Scalia has noted, this kind of inclusion naturally gives rise to the same offense.22 Though Scalia seems to attribute this to the plain meaning of “same offense,” my claim is different: in a mens rea hierarchy, the legislature is creating a set of alternatives for the prosecutor and the jury. The blameworthiness of first-degree murder is greater than that of manslaughter, creating a choice between these offenses without also creating distinct and cumulatively punishable offenses.
A hierarchical pattern can also be created by changing the result element. A legislature could proscribe assault, assault causing injury, assault causing severe injury, and assault causing death. Again assuming that proof of the more severe injury always proves all the less severe injuries, this is a hierarchical pattern.
Another inquiry awaits us. To say that murder is the same blameworthy act-type as manslaughter or that assault with intent to kill is the same blameworthy act-type as assault does not tell us whether robbery is the same blameworthy act-type as larceny. The question here is how to analyze act-type inclusion. This is a form of hierarchical blameworthiness in which what varies is the act-type itself, rather than mens rea or result elements.
The answer, again, is found in the best inference about legislative intent. Though Moore seeks to state a metaphysical account, rather than one based on legislative intent, he argues that “partial identity” of morally salient act-types makes offenses the same.23 Partial identity of act-types presupposes conjunctive act-types. For example, the conjunctive act-type of robbery includes the act-type of larceny and the act-type of assault.24 The principle can thus be stated as one of act-type inclusion rather than partial identity.
The philosophical difficulties with the notion of partial identity can be put to one side25 because my project is not about metaphysical act identity. Thus, I do not need to defend the proposition that acts that are partially identical are nonetheless the same. But Moore’s act-type inclusion states a good presumption about legislative intent. It is unlikely that the legislature meant to create distinct blameworthiness when a conjunctive act-type always includes two or more act-types. Robbery is a form of aggravated larceny, aggravated by the conjoined act-type of assault, a relationship that suggests that the legislature did not create distinct blameworthiness when it prohibited both robbery and larceny.26
As I discussed in chapter 2, the Court held in Grady v. Corbin27 that the inclusion of act-types should be decided not as the statutes were written (the Blockburger approach) but as they are prosecuted (the Grady approach). Which approach is to be preferred on my legislative intent account? As Justice Holmes once said about a double jeopardy question, it is a short point.28 The Grady approach probably achieves a fairer result in some cases. A friend who is a federal public defender steadfastly maintains it is unfair to allow the government to convict of both conspiracy and the planned substantive offense.29 Surely, he argues, the liability for a serious substantive offense should include that for the planning stages.
Perhaps, but this is a criticism of conspiracy law,30 rather than a prescription for same offense. The reason it seems unfair to punish both the conspiracy and the substantive offense, perhaps, is that it is unfair to punish the mere making of plans to commit crime in the future. We gloss over that unfairness when the conspiracy is the only harm the defendant commits, but it assumes more glaring import when the defendant proceeds to commit the planned harm.
More important, it is legal realism of the first magnitude to use unfairness as a synonym for double jeopardy. While a concern for fairness may partially explain the long history of double jeopardy prohibitions, the only way fairness manifests itself in a hard-edged view of double jeopardy is to prevent more than one jeopardy for the same offense.
In deciding the Grady inclusion question, we once again seek the best presumption about legislative intent. A legislature that criminalizes conspiracy almost certainly wants it punished separately from crimes committed pursuant to the conspiracy. The whole purpose of a conspiracy offense is to punish earlier stages of criminal activity. Why assume that this desire to punish somehow disappears when the criminal goes on to commit other crimes?
More generally, a legislature that proscribes distinct blameworthy acts in different offenses would intend these to be separately punishable. That the facts establishing an element in one offense could be used in some cases to prove the other offense is merely a coincidence. On the facts of Grady, the use of drunk driving to prove reckless manslaughter is a coincidence that does not affect the legislative intent to make these offenses cumulatively punishable. The actor has, after all, caused two harms: the risk creation associated with drunk driving and the realization of that risk in manslaughter. That they can be proved on the same evidence does not make them the same offense.
If the legislature wishes offenses to merge when they are proved by the same evidence, it can, of course, provide for that outcome. A vehicular homicide statute, for example, could provide that driving offenses merge into the homicide if used to prove it. A conspiracy statute could provide the same about planned substantive offenses. Lacking that evidence of intent, there is no merger, and the Court was thus right to reject Grady’s fine-grained act-type inclusion.
So far, I have identified three types of hierarchical relationships among offenses—they may vary by gravity of the mens rea, the severity of the harm caused, or the scope of the act-type proscribed. Blockburger treats hierarchical overlaps as creating a single offense. It is not a bad starting place for a definition of same offense, because it captures an important truth about distinct blameworthiness: if proving one offense always proves another, it is unlikely that the legislature meant to authorize both penalties. It is more likely, in the hierarchical context, that the legislative intent was to create a choice of offenses to allow a single conviction commensurate with the gravity of the blameworthiness shown in the particular case.
But hierarchical patterns are not the only way to infer that the legislature intended to create alternative, rather than distinct, blameworthiness. For example, the legislature might use different but nonincluded circumstance elements to create alternative ways of aggravating a basic offense. Consider larceny, an offense of ancient origin. If the legislature wished to create aggravated larceny offenses, it could do that by adding different aggravating circumstances—for example, larceny at night, larceny from the person, and grand larceny. It is unlikely that the legislature means each of these forms of aggravated larceny to be different from the others so that three convictions could result from a single larceny of valuable property at night from a person. (There is some question about grand larceny, which I take up later.) No natural hierarchy exists here, but the offenses share the larceny core and seem to be alternative ways of aggravating the basic offense of larceny.
In effect, this kind of alternative blameworthiness varies by the particular mode of how the offense is committed, rather than by the hierarchical gravity of offense (by mens rea or result). That blameworthiness can attain alternative status by mode of commission as well as by varying gravity explains why Blockburger produces counterintuitive results in some overlapping nonhierarchical offenses. Judge Rutledge saw this problem and rejected Blockburger as the only means of inquiry:
[I]n some instances legislative refinement has defined generically identical offenses with narrow differences in intent or in the means or methods of perpetration, e.g., assault with various specific intents and with variously specified weapons. Some of these differences are substantial, others too slender for, in effect, nullifying the constitutional protection against double jeopardy. When they are so or the question is doubtful, the [same-offense analytical] step should be taken consciously and deliberately, not ignored or taken automatically as is done when the process stops with applying the [Blockburger] test.31
For one example of “slender” differences, the offense of diverting electric current is a different Blockburger offense from larceny because the former does not require intent permanently to deprive the owner of possession (and the latter obviously does not require that electricity be diverted).32 But the best proof of Blockburger’s inadequacy is that sufficiently different ways of describing a single homicide are different offenses. The proposition is flawless if Blockburger tells us all we need to know about what makes offenses different. Felony murder, for example, requires proof of an underlying felony; premeditated murder requires proof of premeditation. Several state courts have rejected the Blockburger outcome in this context.33 Others have mindlessly followed Blockburger and upheld two convictions for one killing. And why not? When the Supreme Court tells us confidently that Blockburger is the sole test of same offense, prosecutors can be excused for believing that the Court means it.
The courts that reject the Blockburger outcome are without an analytical tool and thus resort to fairness or common sense. The Colorado Supreme Court justified vacating one of the convictions because “[i]t would be a strange system of justice that would permit the defendant to be sentenced to two concurrent life sentences for the killing of one person.”34 The court cited an Alaska Supreme Court opinion in which the Alaska court said, “It would indeed be a strange system of justice that would allow … two life sentences for the killing of one person.”35 Two state courts solved the analytical problem by not offering any justification, instead flatly stating that a single death could not support two homicide convictions.36 The Iowa Supreme Court reached the same conclusion, justified only by a citation to two legal encyclopedias.37 One court seemed to rely on an implicit “same-act” kind of analysis, stating: “It is manifestly impossible to kill or slay one person twice.”38 Another court explicitly found a “same-act” limitation on murder convictions: “A defendant cannot be convicted of more than one murder arising out of the same physical act.”39 Several courts used a “surplusage” argument to avoid the issue—as long as the sentences are concurrent, the defendant has not been harmed, and no double punishment issue is presented.40 The Supreme Court rejected this “concurrent-sentence” theory in 1985, and at least two state courts have also rejected the surplusage theory to hold that only one conviction can attend one killing.41
One court resorted to a due process analysis to avoid the Blockburger result: “conviction of both charges, arising from the slaying of the same person amounts to piling punishment upon punishment. Fundamental fairness precludes such a practice.”42 Later, the court limply said, “We think that point so obvious as not to need further comment.”43 The jurisprudential theory from these cases seems to be a rule that strange Blockburger results, those that are obviously wrong, are not acceptable. Needless to say, this is not a very ringing endorsement of Blockburger. How much easier, and more satisfying, to include this protection within double jeopardy by defining “offense” so that a single killing can lead to but one conviction.44
In the larceny hypothetical, it is highly doubtful that the legislature would intend larceny from a person to be punished in addition to larceny at night when a single act-token of larceny occurred, but Blockburger produces this result. A blameworthy act-type account, with its embedded harm singularity principle, gets it right here: the owner’s interests are not wrongfully set back any differently because the property was taken at night than because it was taken from her person. Grand larceny is a harder question because the owner’s interest suffers a greater setback when $500 is stolen that when five cents is stolen.
Of course, to say that the owner who loses $500 suffers a greater setback than one who loses five cents is not to say that the harm is distinct, only that it is greater. Understanding the difference between distinct blameworthiness and alternative blameworthiness is the point of this chapter. A sensible presumption would be that elements differing in quantity (of goods stolen, of drugs sold) are intended to create alternative blameworthiness, rather than distinct blameworthiness. Elements that differ in quantity seem to do the work of hierarchical grading, as in the laws of Aethelberht (ca. A.D. 600), when a single blow that pierced both cheeks was punished at twice the rate of the blow that pierced one cheek.45
To the extent there is doubt about whether the amount of money stolen creates alternative blameworthiness, two other analytical moves are available. First, an element must do more than manifest blameworthiness to count as part of the blameworthy act-type under my double jeopardy principle. It must also be an act-type. The amount of money taken by a single act-token of larceny is not likely to be included as part of the act-type. The act-type is taking property, with the value of that property a circumstance existing in the universe. If the actor takes a watch or a ring or a stack of bills, the dollar value is not part of what the actor is seeking to bring about, except in the most general sense that thieves want to steal property with greater, rather than lesser, value.
This, too, is not completely free from doubt. It is, after all, sometimes possible for the actor to bring about the property of how much is stolen (by taking a larger stack of bills, for example). Thus, a second analytical move resorts to a global presumption that the Court uses sporadically, called the “rule of lenity.”46 As the Court’s terminology suggests, if doubt exists about whether the legislature meant to create distinct blameworthiness, courts should presume singular blameworthiness. If error is made, it should be made in the direction of finding fewer “units” of blameworthiness. At a minimum, the grand larceny value element raises doubt about whether it is a blameworthy act-type, and that doubt should be resolved in favor of singular blameworthiness.
If grand larceny and larceny at night define a single blameworthy act-type, and thus singular blameworthiness, then the narcotics offenses in Blockburger also define only a singular blameworthiness. This is not, of course, to say that the legislature somehow lacks the power to punish both narcotics offenses cumulatively (or the larceny offenses). Chapters 1 and 2 made clear my view on that question. Rather, it is to argue that when the legislature creates the same blameworthy act-type in more than one offense description, a single act-token of that blameworthy act-type should be presumed to manifest singular blameworthiness.
Having offered a method of isolating statutory blameworthy act-types and comparing them for sameness, I must now offer an account of when act-tokens (discrete acts in the world) are separate. A little-noted part of Blockburger addressed the act-token issue with more success than the Court had with the issue of whether prohibiting one sale of narcotics with different attendant circumstances was two offenses of selling narcotics. Blockburger was convicted of two different narcotics offenses and also of two counts of one of those offenses. The two counts arose from the sale of two quantities of narcotics; when one transaction was done, the purchaser paid for another quantity, to be delivered the next day. The Court used Wharton’s “impulse” test to decide that there were two offenses of selling: “when the impulse is single, but one indictment lies, no matter how long the action may continue. If successive impulses are separately given, even though all unite in swelling a common stream of action, separate indictments lie.”47 While this basic notion is consistent with legislative intent, separating “impulses” is a difficult inquiry.
How many theft offenses did the legislature authorize when a thief steals two horses standing side by side? Does it matter if they have different owners? Does it matter if the horses are roped together so that the thief can steal them with one motion? If stealing two horses is two thefts, what about stealing a purse that contains one hundred different items? The act-token question turns in part on the question of how many harms have occurred. This requires an account of harm singularity, which the next section introduces.
Usually, when A acts to harm B, a single act-token causes a single harm. A murders B or robs B or steals B’s wallet. It would be an easily stated rule if we could say that each act-token is a distinct double jeopardy harm and thus a separate offense. Michael Moore develops and defends this proposition. The defense is not without difficulties, however.
There are two basic sets of problems. First, a single act-token can cause multiple harms. If D strikes fifty persons with a single shot from a shotgun loaded with peas, Bishop found only one offense of battery, because occasioned by one volition (drawing a distinction between “one volition” and “one transaction”).48 But Bishop acknowledged contrary authority, and a crude harm analysis would contend that because fifty persons are harmed, fifty batteries have occurred. In the Oklahoma City bombing, a single act-token of detonating a bomb killed 167 people. To insist that this is one murder offense because it is one “act” of murder seems wrong. Perhaps, then, courts should consider the distinctiveness of harms in a yet unspecified way when counting the number of offenses.
The other set of problems for action theory comes when multiple act-tokens occur. When act-tokens are distinct, the harms are often distinct as well. For an easy example, the robbery of six people sitting around a table is six offenses of robbery.49 Robbery is, of course, the same offense as robbery, but six act-tokens of robbery are six different offenses. Killing four victims with four separate shots is four offenses of homicide.50 Another example is larceny and burglary. Hale noted that they are “several offenses, tho committed at the same time. And burglary may be where there is no larciny, and larciny may be where there is no burglary.”51 The act-token of larceny is distinct from that of burglary, and, by the same reasoning, the harm of the larceny is different from that of the burglary.
But multiple act-tokens sometimes seem to cause a singular harm as measured by Feinberg’s harm principle. Taking two items from the top of a dresser with separate motions may be two act-tokens of taking but only one Feinberg harm. The owner’s interests are wrongfully set back when the thief takes one item from the dresser; the second taking (or the tenth) seems cumulative rather than distinctive. Using a roughly similar theory, the Supreme Court held that multiple violations of the Fair Labor Standards Act are only a single offense (harm) for each type of violation.52
Feinberg defines harm to include mens rea—A does not wrongfully harm B unless A is at least negligent with respect to the consequences threatened by his act. Notice, though, that mens rea has nothing to do with distinctiveness of the harm. The harm (as a noun) is the same whether A has one or several mental states concerning his act as he wrongs B. A may watch B eat poisoned food intending B’s death, then expecting it, then hoping for it, then hoping against it (but knowing that it is likely), then hoping against it (and believing it unlikely). Multiple mental states do not multiply the harm. B’s death or sickness from the poison is one death, one sickness, one harm.
In what follows, the “harm singularity” principle draws from Feinberg’s definition of harm and thus focuses on the real-world effect of the act, circumstance, and mens rea. The harm singularity principle is that, unless the legislature speaks clearly to the contrary, the scope of the act-token that proves one (and only one) blameworthy act-type is coextensive with the harm proscribed by the act-type. The next section begins with the Court’s most sensitive and sensible analysis of the harm/act-token issue.
Larry Simon is right that counting act-tokens in the world is unlikely to provide much guidance on the same-offense question. Shaving is one act-token or several act-tokens or many act-tokens, depending on how coarsely one defines the relevant act-type. But a double jeopardy account merely needs to be able to tell how many tokens of the statutory act-type have occurred. This can be done by careful reference to the statutory language, applying my harm singularity principle.
How many joyriding offenses is it to keep a car for nine days and drive it many times, as in Brown v. Ohio? Martin Friedland states that “[m]ost courts” uphold multiple convictions when each violation “is brought about by a different physical act.”53 But his authority for that statement is a case holding that a separate sentence may be imposed “where each offence charged in each count is separate and distinct.”54 Of course. What we need is a test for deciding when each offense is separate and distinct. The answer to that question in Brown is the number of times Brown committed the statutory act-type of joyriding. To answer that question requires understanding as precisely as possible the scope of the blameworthy act-type and the way act-tokens instantiate act-types.
Brown v. Ohio not only raises and resolves in an appealing way many facets of the same offense problem but also provides hints of a profound conceptual double jeopardy universe. To begin, it seems right to say that a single taking and operation of a single car is a single offense. The same rule holds in England.55 Yet Justice Blackmun dissented on the ground that Brown must have committed more than one driving episode during the nine days he kept the car. In effect, Justice Blackmun argued that regardless of the relationship between the act-types of auto theft and joyriding, Brown’s liability for each act-type could be made to fasten on a different act-token.
Justice Blackmun’s view that Brown’s driving episodes were different act-tokens seems, at first blush, to make sense. The first and last episodes were probably separated by nine days. If Brown had stolen property on two separate days, we would have two act-tokens of larceny. Why would two driving episodes separated by nine days not be two act-tokens of joyriding?
The answer is that any distinction between act-tokens cannot be calculated until the scope of the statutory act-type is understood. The criminal taking of property from X is a different act-token (of whatever act-type) than a taking from Z. But taking Y’s property and keeping it nine days would be only one act-token of larceny. Indirectly making this point, the Court in Brown noted that a different case would be presented if the statute made each day of joyriding a separate offense.56 In the Court’s hypothetical case, each day of possession is a separate statutory act-type, and each day of Brown’s nine-day possession is a token of one of the act-types.
The act-type question is necessarily antecedent to the act-token question.57 Justice Blackmun’s common-sense view about driving episodes is commonsensical only if a single taking of a car instantiates more than one statutory act-type. Brown’s taking is, on standard action theory, more than one act-type in the universe: starting the car, putting it in gear, driving it away, getting out, getting back in, driving, stopping, getting out, getting back in, driving, and so on. But, as I have been at pains to note, the double jeopardy same-offense question has nothing to do with metaphysical action theory and everything to do with the scope of the proscribed act-type. Thus, the crucial question is whether the Brown majority was right to find that both joyriding and auto theft proscribed a single act-type of taking a car. It is a more difficult question than the majority acknowledged, which is probably what drew Blackmun’s dissent.
The joyriding statute in Brown used the verbs “take, operate, or keep” while the auto theft statute used “steal.”58 “Steal” implies that a single act-token of that act-type occurred, regardless of how long Brown had the car. Thus, for purposes of theft liability, only one act-token occurred.
“Take” and “keep” in the joyriding statute have the same scope as “steal.” Once an actor has taken, kept, or stolen, there is no future act-token in continued possession or use of the property. “Operate” is murkier, perhaps lending itself to Blackmun’s interpretation that each driving episode could be a separately punishable act-token. Since the joyriding act-type is stated in the alternative, “operate” could justify finding more than one act-token on the basis of separate operations. But the Court’s rule of lenity might presume that when the legislature uses more than one act-type word to describe what is forbidden, the words should be understood, for purposes of counting act-tokens, as creating a conjunctive act-type.
The Court has long applied a rule of lenity in double jeopardy cases.59 If a statute is somehow unclear how to count violations (the Court has never specified how this threshold lack of clarity is established), then the Court places a burden of clarity on Congress.60 In the Court’s words, there is a “policy of not attributing to Congress, in the enactment of criminal statutes, an intention to punish more severely than the language of its laws clearly imports in the light of pertinent legislative history.”61 When the statute proscribes alternative act-types, we can be sure we are not punishing in excess of legislative intent if we count only act-tokens of the conjunctive act-type, rather than act-tokens of any of the alternative act-types.
It is an old idea. In an early same-offense case in this country, the Court relied on a 1777 English case in which the court refused to count four sales of bread on Sunday as four violations of a statute that provided that “no tradesman or other person shall do or exercise any worldly labor, business, or work of their ordinary calling upon the Lord’s day.”62 Each sale of bread could be an “exercise” of business, but the King’s Bench indulged a more lenient construction, finding that the intent of Parliament was to penalize “exercising his ordinary trade on the Lord’s day.” On this construction of the scope of the act-type, the English court held that selling baked goods on Sunday could be committed only once per Sunday, regardless of how many sales occurred that one day. Otherwise, the court observed, “if a tailor sews on the Lord’s day, every stitch he takes is a separate offence.” Judges in 1777 appreciated the extraordinary number of act-tokens that can exist if the act-type is defined narrowly enough.
A skeptic might ask why the legislature bothered to define the act-type in the alternative if courts and scholars are going to collapse it into a conjunctive act-type. The answer, of course, requires paying attention to the context in which the question arises. Presumably the legislature created the act type of “take, operate, or keep” to make certain that actors who did any one of them would be guilty of one count of joyriding. Obviously, Brown could not be convicted of three counts of joyriding on a charge that he took a car, or he operated it, or he kept it. Only one conviction can be entered here because of the nature of alternative act-types. More important, there is no reason to believe that the legislature wanted an actor convicted of three counts of joyriding if he took a car and he operated it and he kept it.
If that is the right way to read legislative intent, then it seems equally likely that the legislature did not intend to authorize multiple convictions if Brown operated the car more than once during the period he kept it. This demonstrates an important point for my account: counting act-tokens is different from asking if a single act-token has been committed. Whether an offense has been committed is different from when that offense ends and another begins. Though the relationship has never been stated this way, I think it uncontroversial to say that a single act-token can be proved on any of the alternative criminal act-types, but counting act-tokens requires that statutory alternative act-types be considered a conjunctive act-type.
On a conjunctive reading of the joyriding statute for purposes of counting act-tokens, the proscribed act-type is “take, operate, and keep.” Brown committed only a single act-token of that act-type, supporting the majority’s conclusion that the prosecution had attempted to “divid[e] a single crime into a series of temporal or spatial units.”63 The legislature can make that division, the Court stated, but not the prosecutor.64 That, of course, is precisely my theory from chapters 1 and 2.
The value of a lenient approach to act-type construction can be seen in Brown. Blackmun’s view of driving episodes as multiple joyriding offenses means that a defendant could be punished more severely for the misdemeanor of joyriding than for the felony of auto theft. A single taking could never be more than one token of theft as it is normally defined, but a defendant could receive many consecutive sentences for multiple driving episodes based on a single taking. This could not be what the Ohio legislature intended when it relegated joyriding to misdemeanor status.
Crimes that separate the act from the result, such as murder, are a special case of act-type/token relationship. The best example is Diaz v. United States,65 where Diaz was convicted of assault, the victim died, and the government then prosecuted for homicide. Normally, the scope of the act-type of homicide includes the act-type of the assault that caused the killing, regardless of how long afterwards the victim dies. The question in Diaz is whether the verdict for assault somehow legally ended the part of the homicide act-type that included assault, in effect making the homicide a different act-token from assault.
Philosophers are split on whether killing is identical to the act that causes the death. Moore defends the coarse-grained view that the act of killing includes the result that the victim died.66 This leads to problems: how can we say that V’s death occurred several hours (or days) after V was killed? In the legal context, however, we can avoid philosophical difficulties. The legal liability for any act-token ends when it is manifested in a verdict, and future occurrences causally related to the earlier act must be a new legal act-token. As Diaz put it: “At the time of the trial for the [assault] the death had not ensued, and not until it did ensue was the homicide committed.”67 The Court thus affirmed the homicide conviction.
The last, and in many ways the most difficult, issue is how to count statutory violations when we unquestionably have multiple act-tokens of the statutory act-type but when one act-token may have “consumed” the blameworthiness of the act-type. One solution is to consider multiple act-tokens of the statutory act-type always to be multiple offenses, but this can produce counterintuitive results. For example, a prosecutor once filed 1,800 different informations for sales of beer in a single day; the statutory violation was selling beer without a certificate of inspection.68 Though the beer seller has clearly committed 1,800 act-tokens of the act-type of “selling beer,” it is not inevitably true that he has committed 1,800 violations of the criminal offense of selling beer without a certificate. That, after all, is the issue; when to presume that the legislature intended multiple, duplicative act-tokens to be counted as more than one violation.
Thus, the question is whether the legislature meant 1,800 act-tokens to count as 1,800 violations of the blameworthy act-type proscribed by statute.
In 1937 Frank Horack wrote, “Carefully analyzed, there appears no obstacle, as a matter of substantive law, to the prosecution of each consequence of a criminal act.”69 If he meant that courts can, by looking hard enough, find a theory to support a conviction for each consequence, he is clearly right. But if he meant that the substantive law always contemplates a separate conviction for each consequence, then Horack is wrong. It seems beyond question that the Missouri legislature did not mean to authorize 1,800 convictions for sales of beer without a license if those sales took place in one day.
There are two questions about the relationship between act-type and act-token, and courts make a mistake when they do not see the difference in the questions. The first question, which is more familiar to courts, is whether at least one act-token of a particular act-type is made out on the state’s evidence. If not, of course, no conviction can be entered or sustained. But the question of whether one act-token is made out on the facts proved is not the same as the question of how many act-tokens should be counted as violations of the statutory act-type. The first question is a minimum, sufficiency-of-the-evidence question. The second question is a counting question: yes, we have more than one act-token on these facts, but should we count all the act-tokens as statutory violations?
Courts have struggled with the counting question. In Ebeling v. Morgan,70 the statute made it a crime to “tear, cut, or otherwise injure any mail bag … with intent to rob or steal any such mail.”71 Ebeling had cut into six mailbags in a single transaction. Unlike Brown’s ongoing posession of the car, Ebeling’s first act-token was complete when his knife reached the second mailbag. That the statutory act-token occurred six times does not, however, necessarily mean the legislature authorized counting all six act-tokens.
Insight into Ebeling can be gained by asking the question first posed by a student note in the 1907 Harvard Law Review: how many separate injuries has the state suffered?72 This is a particularly insightful way to ask how many act-tokens the legislature meant to count. The student note applied the “injury-to-state” test to argue that only one theft occurs when someone steals goods belong to several different owners. “[I]t is clear that the state has been injured but once; and where there is only one transaction and one injury to the state, the offenses are identical.”73
Suppose J plays seventy-five hands of poker in a state where it is a crime to “play a game in which money is bet.” Convicted on the basis of playing one hand, he raises a double jeopardy objection to a new trial based on a different hand. These are different act-tokens of the act-type of playing a game in which money is bet, but it strains credulity to believe that the state is injured seventy-five times when consecutive games are played. Surely this is, as one commentator noted, a single “continuing injury.”74
Stealing, from the same room, a purse and a watch that belong to different owners is a single larceny.75 Bishop concluded that the better rule was that stealing any number of objects in one transaction was only one offense, regardless of the number of owners.76 Now suppose B breaks and enters a mansion and then makes six trips into and out of the burgled structure as she steals goods from inside. B is obviously guilty of both larceny and burglary, but is she guilty of six burglaries? Any entry would constitute an act-token of burglary, but the additional entries do not add significantly to the blameworthiness manifested in the act-type of burglary and then consumed by the first entry. Stating it in Feinberg’s terms, the second trip across the threshold does not constitute a separate wrongful setback to the victim’s interests. The six trips do not appear to be separate “injuries” to the state. Stealing many goods from a mansion is grand larceny, but six convictions for burglary overshoot the blameworthiness mark.
The Court made precisely that mistake in Ebeling, assuming that the question of what constituted the first act-token is no different from the question of how best to count statutory violations. On that superficial understanding, Ebeling committed six statutory violations. The Supreme Court accordingly affirmed the six convictions and five consecutive maximum sentences imposed by the trial court,77 thus permitting a sentence five times longer than could have resulted from a single conviction.
Ebeling noted that congressional intent was “to protect each and every mail bag from felonious injury and mutilation,”78 but this, too, misses the point of whether Congress intended multiple injuries in one transaction to be multiple offenses. Once the Court’s category mistake is put to one side, a question still remains regarding Congress’ intent. Phrasing the issue suggests the obvious: Congress surely had no intent at all about how many convictions should attend cutting six mailbags during a single transaction. So the real question is, What would Congress have answered had the question been put to it?
The rule of lenity is not particularly helpful in these cases because, even when recharacterized as a burden of clarity, it offers no useful principle of construction. In Ebeling, for example, can it be said that the “language” of the mailbag law “clearly imports”79 six penalties for six cuttings? Well, yes, on one reading. It is, as the Court noted, a crime to injure “any mail bag.” But, on the other hand, Congress said nothing about how many penalties should attend multiple cuttings of mailbags in one location. So, on another reading, the language does not clearly import six penalties.
Without some principle to guide the rule of lenity, it is thus nothing more than an ad hoc rationalization for a decision reached on other grounds. The Court upheld the six penalties in Ebeling. The chance of upholding one hundred penalties under the same statute is very small. In this case, the Court would trot out the rule of lenity as the basis for its decision, but the real basis would be some rough intuition that one hundred penalties is just too harsh. What is needed to avoid ad hoc analysis is a principle that better gives voice to the intuition about the relationship between blameworthiness and the number of violations.
The best answer will reflect an understanding of how tokens manifest the blameworthiness created by the act-type. It will require an inventory of interests that are subject to wrongful setback. That task is beyond the scope of my present project—it would take a volume itself—but some tentative thoughts can advance the double jeopardy inquiry.
Consider a stack of one hundred one-dollar bills. How many larcenies occur when those bills are stolen by one hundred takings in rapid succession? Is the blameworthiness in this case different from that in stealing one hundred bills in a single taking? In a metaphysical sense, the multiplicity of takings (or blows given in a fight) multiplies the number of act-tokens and moral wrongs committed.
To conclude that each morally wrong act-token should always be counted as a separate criminal violation is to conclude that act-token individuation tells us all we need to know about same offense. But it does not, as Moore seems to concede.80 The number of bills taken may make no difference. The legislature may think in terms of cohesion of result rather than fragmentation into separate act-tokens when the very same act-token is repeated, with the same goal in mind, within a short time span. Taking one hundred one-dollar bills in a single motion is indistinguishable from taking one one-hundred-dollar bill. Why should the blameworthiness differ if the actor takes the one hundred one-dollar bills in two motions? In ten? In one hundred?81
If stealing one hundred bills can be counted as a single larceny regardless of the number of takings, the number of mail bags Ebeling cut may not tell us how many offenses he committed. His intent may be more important than the number of act-tokens. The actor’s objective is one way of gauging the singularity of harm. The actor who intends to take all one hundred bills has done as much harm with one taking as with one hundred.
The thefts of X’s horse from one pasture and of Z’s horse from a pasture ten miles away are two larcenies, at least in part because we infer an intent to deprive two owners of their property. But stealing a saddled horse is one larceny, not one larceny of a horse and one of a saddle, at least in part because the intent was to steal a saddled horse. Similarly, stealing a purse is one larceny, not a different larceny for each object in the purse.82 Moore rejects individuating offenses by intent, but, in these cases, his action theory reaches the same result. “This is because defendants often do just the wrongs they intend to do, in which case the same-intent test individuates identically with wrong-relative individuation.”83
But consider stealing two of X’s horses standing a few feet apart. Though there are two act-tokens of larceny here, it could be counted a single larceny as long as the thief had the “single intent” to steal X’s horses.84 This is similar to the Court’s “single impulse” analysis from Blockburger.85 Action theory is an unsatisfying way to approach these unusual cases. Stealing a gym bag with the intent to deprive two people of their property would be two larcenies, in my view, despite the single taking.86 Stealing one hundred one-dollar bills by one hundred takings would, on the other hand, be a single larceny if the thief had formed the intent to steal all one hundred prior to taking the first one.
The offense gravamen of the mail bag statute in Ebeling was not merely cutting mailbags, but cutting mailbags “with intent to rob or steal any such mail.” One may thus infer that Congress’ principal concern was to protect the contents of the mailbags, rather than the canvas bags themselves. If this is right, then why does it make a difference in blameworthiness that six mailbags were cut? Stealing the contents of six mailbags would be one larceny. If it is, then why should it matter that the statute forbade cutting as well as stealing? Ebeling would be guilty of only one offense if he cut into a giant mailbag that contained as much mail as six regular bags. Again, it is unclear why the result should be different just because he made six cutting motions.87
Justice Holmes came to see the issue in this light. In an earlier case, he had followed Ebeling in holding that each act of putting a letter in the mail was a separate fraud violation, even though only a single scheme was involved.88 In that case, Holmes commented, “[T]here is no doubt that the law may make each putting of a letter into the postoffice a separate offense.”89 Of course, but that merely restates the issue: did Congress intend that each letter put into the mail, as part of a single scheme, be counted as a separate violation? Holmes, following Ebeling, did not seem to think that these were different questions.
More than a decade later, in United States v. Adams,90 Holmes saw the difference. Adams had made two false entries in bank books, both relating to a single deposit and credit. The government prosecuted the false entries successively, as separate violations of the offense of making a false entry with intent to injure or defraud the bank. “It is a short point,” Holmes commented.
The Government contends for the most literal reading of the words, and that every such entry is a separate offense to be separately punishable. But we think that it cannot have been contemplated that the mere multiplication of entries, all to the same point and with a single intent, should multiply the punishment in proportion to the complexity of the bookkeeping.91
To ask whether the literal reading was “contemplated” is to seek congressional intent. Holmes’s rejection of “multiply[ing] the punishment in proportion to the complexity of the bookkeeping” shows that his benchmark for inferring legislative intent was the blameworthiness of the sequence of act-tokens. Holmes thus stated a version of my principle of consuming blameworthiness. The first false entry, in effect, consumed the blameworthiness of the second entry done with the same intent and with respect to the same deposit.
Using intent to help individuate double jeopardy offense explains why it is a single offense to take two women across state lines in a single vehicle in violation of the Mann Act.92 On Moore’s account, this would be two act-types of violating the Mann Act, one for each victim. Individuation by intent explains why the Court held thirty-two counts under the Fair Labor Standards Act to state but three offenses; the Court treated “as one offense all violations that arise from that singleness of thought, purpose or action, which may be deemed a single ‘impulse.’”93 At a minimum, Moore’s action theory suggests eleven violations because eleven employee-victims were named; perhaps he would find thirty-two violations because of the thirty-two act-tokens of the various act-types involving minimum wage, overtime, and record-keeping.
To this point, the harm singularity principle has supplemented action theory as a way of gleaning legislative intent (1) on what constitutes one act-token of a statutory act-type and (2) in cases in which the legislature may intend to authorize only a single criminal violation even though multiple act-tokens of the statutory act-type have occurred.
The last act-token issue arises when there is only one act-token but multiple harms. The harm singularity principle is useful here as well.
Moore claims that the results of the state cases involving single act-tokens and multiple harms are a “checkerboard” that reveal “conceptual confusion.”94 As an example, Moore identifies the following set of results that are the majority rule. First, two murders result when Smith kills Jones and Long with one shot from a powerful gun; Moore posits that Smith intends to kill Jones and is silent about Smith’s mens rea with respect to Long. Second, only one homicide offense occurs when Smith kills both Jones and Long while possessing a negligent mens rea as to Jones and no criminal mens rea as to Long. Third, only one larceny is committed when Smith steals a bag containing property belonging to both Jones and Long.
Moore is right, given his premises. These results are conceptually confusing if one seeks to find the answers in act metaphysics, as Moore does. If one individuates act-tokens by the number of victims, each of these examples is two offenses; if, on the other hand, one believes that the act of killing includes the result that the victim dies—as Moore does—then each of these examples is but one act-token and can be only one offense using traditional action theory.95 What act metaphysics does not permit is the result that courts reach: sometimes there are two offenses and sometimes one.
Moore’s coarse-grained individuation defines act-token by what the actor does and not by the consequences, “no matter how numerous the bad consequences.”96 Without another principle, he would be compelled to say that only one offense occurs in each of his sample cases because there is only one act-token. If murder is a single act-type, this would lead Moore to conclude that only a single homicide occurred in each case where there was a single token, and the same with the larceny hypothetical.97 But, Moore argues that morality here is victim-relative as well as agent-relative. Thus, the act-type of killing Jones must be understood as different from the act-type of killing Long; indeed, an act-type of killing exists for every human who lives in the jurisdiction. On this view, the killing of two people by one act would be not two separate act-tokens but two separate act-types.98
While this seems right, it creates problems of its own because it commits Moore to say that eighty-four homicides occur whenever eighty-four victims are killed as a result of a single negligent act-token.99 Would this be the right answer if the actor was reasonably unaware of the presence of eighty-three of those eighty-four victims? Rejecting act metaphysics in the context of multiple-consequence cases makes it possible to consider what Moore expressly puts aside: “culpability measurement in double jeopardy contexts.”100
If we reject act metaphysics as the sole guide to individuating offenses, we are free to consider the actor’s mens rea as to the unintended victims. So, for example, where Smith intends to kill Jones and also kills Long, we would need to know Smith’s mens rea as to Long. In all likelihood, Smith would be acting either knowingly or recklessly with depraved indifference concerning Long. Firing a powerful gun with other people in the vicinity seems at least reckless, and depraved indifference is implicit in Smith’s intent to kill Jones. Thus, those courts that find two counts of murder here are right, but not because there are two act-types of murder (the act-type of murdering Jones and the act-type of murdering Long), which is Moore’s solution. Nor is it true that the presence of two dead bodies (two consequences) justifies this outcome. There are two bad consequences, to be sure, but Smith is criminally liable only for those outcomes that he causes with the requisite mens rea.
The real reason two murders occur here is that the violations of the act-type of murder should be individuated by Smith’s mens rea. If Smith has murderous intent toward both Jones and Long, there are two murders. This seems right. Why should bringing about two deaths with murderous mens rea be treated differently depending on whether it took two shots or only one? On the other hand, if Smith is reasonably unaware of the presence of other humans when he shoots Jones, there is only one murder because only one killing with criminal mens rea.
In the second homicide case, Moore expressly posits that Smith is negligent with respect to Jones but not to Long.101 On my account, courts are right to consider this only one homicide offense because Smith violated the homicide statute but once. Though Smith is not negligent with respect to Long, by definition, Moore nonetheless claims that Smith “negligently or recklessly kills” Long.102 How can this be? Moore would convict Smith of the act-type of Long’s manslaughter when he did not have the mens rea required for manslaughter. This result, by itself, is good reason to eschew action theory in these unusual cases.
A mens rea solution seems implicit in the notion of legality and statutory construction. The question is how many times Smith has violated the murder statute. To answer this question, we simply apply the homicide statute to Jones’ death and then to Long’s death. If Smith had no mens rea as to the death of Long, he would have violated the terms of the homicide statute only once. Indeed, Moore’s assertion that Smith has done more wrong here is puzzling. More harm has occurred, but how is Smith criminally liable for it? If Smith is liable for Long’s death because the bullet passed through Jones, presumably Smith would be liable when his bullet passed through a target. But as long as Smith lacks mens rea in the target case, he is not criminally liable. Why should this result change when the bullet passes through Jones rather than a target? Moore offers no reason to treat these cases differently.
The only problem with the way courts analyze these cases is the assumption that each dead body is a separate harm. Moore is right that this analytic manuever, in effect, introduces a third measure of wrongdoing—the number of harmful consequences—in addition to counting the number of wrongful act-types and the number of times each act type occurred (the act-token question). Moore argues that this manuever is incoherent because the degree of wrongfulness of Smith’s action is “fixed once we admit he did one wrong of a certain significance, once.”103 Courts are sometimes guilty of this incoherence.104
Despite our agreements on some points, Moore will likely think me guilty of bad metaphysics for mixing harm, action theory, and mens rea to individuate blameworthiness. But I remain convinced that this approach is the only one that treats the statutory language with the respect that it must have—harms are multiple only when the actor had the requisite statutory mens rea as to each consequence. Once we clarify the harm category, by adopting Feinberg’s terminology and individuating by intent, the harm manuever is both coherent and compelled by my legislative prerogative account of double jeopardy.
The answer to the larceny case follows from the homicide cases. If Smith does not know that the bag contains the property of two people—the typical case in which courts apply the “single larceny” rule—he has committed one larceny. But if Smith knows who owns the property in the bag and intends to take the property of two people, he has committed two larcenies. The legislature made the taking of the property of another a crime when the actor intends to deprive the owner permanently of possession; two offenses occur when Smith takes property with that intent as to two owners.
Moore’s victim-relative morality and coarse-grained individuation of act-tokens lead him to say that two thefts always occur when a thief steals a gym bag containing the property of two persons.105 This seems counterintuitive; Bishop argued that the “better legal reason” supports finding only one offense when property belonging to two owners is stolen at one time,106 and I think Bishop is right when the actor lacks intent to steal from two owners. The history of larceny, defined as a crime rather than a tort, has focused on punishing the infringement of ownership and not on restitution of a particular owner.107 If infringement on individual ownership were crucial to the legislative scheme for defining larceny, as Moore’s victim-relative morality requires, restitution would be a crucial remedy; it is not. Moore’s view commits him to find one million larcenies in the theft of a truck carrying one million checks, each made payable to a different person.108
But courts who treat all “single larceny” cases the same way, irrespective of Smith’s knowledge of the multiple ownership, are also missing the point. Why treat a single taking where Smith knows of the multiple ownership differently from one in which Smith opens the locker of Jones and then the locker of Long to steal their property? Why should the singularity of act make a difference when the intent is so clearly multiple? Act metaphysics are not helpful in these cases.
To be sure, individuating act-types by moral wrongs usually mirrors legislative thought on the singularity of offenses. Moore deserves credit for this powerful insight. But on a legislative-intent account of same offense, it is a mistake to ignore the mens rea of the individual offender. When a single act-token of one proscribed act-type causes multiple consequences, the best solution is to individuate by the defendant’s culpability. As long as this is understood to mean comparing the defendant’s mens rea to that of the statute,109 it is exactly right. Courts go off the track when they look for answers in the metaphysics of culpability—for example, trying to determine when more than one statutory violation is incident to a “single objective.”110
Moore properly rejects the “single intent” test when applied as a substitute for act-individuation.111 A series of shots is not a single act because motivated by a single intent.112 Three sales of narcotics to three different buyers on three different days is not a single act because made pursuant to a single intent.113 On the account offered here, mens rea is relevant only after act-individuation has produced the equation “single blameworthy act caused multiple consequences.” At that stage, carefully applying the statutory mens rea sometimes produces the answer that the actor has violated the statute once and sometimes that he has violated it more than once.
On this understanding, the distinctiveness of harm provides the answer both to the single-act-token/multiple-harm question and to the multiple-act-token/consuming-blameworthiness question. This seems right. When writing criminal statutes, legislatures consider the nature and distinctiveness of the harm done. That is why joyriding and cohabitation are defined to include a broad time period, while crimes of violence focus on the narrow act causing harm—rape requires penetration, murder requires killing, and battery requires touching.
The distinctiveness of harm caused by the defendant’s act-token is, therefore, the best principle to use when seeking legislative intent about how many act-tokens to count as criminal violations or how many criminal violations to ascribe to a single act-token.
Determining whether one blameworthy act-type is included in another provides the best presumption about legislative intent to create singular blameworthiness in the difficult category of cases when there is no natural hierarchy of mens rea or results. I cannot catalog every conceivable pair of act-types to demonstrate the intuitive force of my argument. Moreover, the account of presumed legislative intent that counts only one violation despite multiple act-tokens of the statutory act-type leaves much room for courts to decide cases as they please. Lacking a comprehensive substantive defense of my concept of singular blameworthiness, I offer instead a procedural defense that follows from chapters 1 and 2.
The blameworthiness principle is nothing more than an inference about legislative intent, and it can always be rebutted by clear intent. If a court applied a blameworthy-act principle to find two statutory offenses to be the same, for example, and the legislature did not intend that result, clarifying legislation could easily be passed to replace the inference of intent with actual intent. Congress did precisely that in 1984 in reaction to the Supreme Court’s lenient interpretation of congressional intent concerning the application of the federal felony-firearm statute.114