Double jeopardy law made simple.

"[N]or shall any person be subject for the same offence to be twice

put in jeopardy of life or limb ...."(1)

Modern Supreme Court case law is full of double jeopardy double talk. Consider first the poetic phrase "life or limb." It seems sensible enough to read these words as a grim and graphic metaphor for criminal sanctions--and such an approach runs deep in American case law, to say nothing of English literature. This reading also makes the most sense of the precise location of the Fifth Amendment Double Jeopardy Clause, wedged as it is between two other provisions--the Grand Jury and Self-Incrimination Clauses--that apply only to criminal offenses. But can "life or limb" be stretched to encompass some civil suits involving only money? Today's Supreme Court seems to think so,(2) but how can this be squared with the text and structure of the Fifth Amendment? The Fifth Amendment Due Process Clause clearly applies to civil cases, but isn't its "life, liberty, or property" language obviously contradistinguished from the more narrow "life or limb" language of the Fifth Amendment Double Jeopardy Clause?

Consider next a far more egregious example of modern double jeopardy double talk. The Double Jeopardy Clause speaks of the "same" offense, and yet the Court casually applies the Clause to offenses that are not the same but obviously different. Premeditated murder is not the same as attempted murder or manslaughter; armed robbery is not the same as robbery; and yet under the so-called Blockburger test, the Court generally treats a greater offense as the same as each of its logically lesser-included offenses.(3) But on rare occasions, the Court rejects this test and reminds us that "[q]uite obviously the [greater] offense is not, in any common-sense or literal meaning of the term, the `same' offense as one of [its lesser-included] offenses."(4) How can we make sense of all this?

Finally, consider the question of precisely when "jeopardy" attaches. The modern Court claims that once the jury is sworn, a defendant is in "jeopardy."(5) Thus, defendant A cannot be tried a second time if, say, her first jury is dismissed because of gross prosecutorial misconduct. But if defendant B's first jury is dismissed for some other reason--because it cannot reach a unanimous verdict or because some jurors fall ill during trial--then B can indeed be tried again.(6) Why, on the Court's premises, doesn't such a retrial likewise place a person twice in jeopardy--and for the identical offense--in obvious violation of the bright-line rule of the Clause? And if a person is simply not "in jeopardy" until his jury is sworn, does this mean that if defendant C has won an acquittal in a fair and suitably error-free trial, C may nonetheless be reindicted for the same offense and held in pretrial detention until a second jury is sworn? If not--if jeopardy really attaches upon C's second indictment itself--does this mean that when a good faith prosecutor dismisses defendant D's first indictment pretrial, then the government is forever barred from bringing a new indictment on the same offense against D (based, say, on new evidence)?

Modern Supreme Court case law is also full of double jeopardy double takes. For example, in the 1975 Jenkins case,(7) the Court, per Justice Rehnquist, promulgated a double jeopardy test for identifying which erroneous trial court dismissals could be appealed by the government, reversed on appeal, and remanded for proper retrial.(8) But in the 1978 Scott case,(9) the Court, per Justice Rehnquist, explicitly overruled the 1975 Jenkins case and promulgated a new test.(10) (As we shall see, this new test is also flawed(11)--but never mind that for now.) Another example: In the 1990 Corbin case,(12) the Court, by a five-to-four vote, laid down a general test for identifying which formally different offenses should be treated as the "same" for double jeopardy purposes.(13) But in the 1993 Dixon case,(14) the Court, by a five-to-four vote, explicitly overruled the 1990 Corbin case.(15) And yet in the 1994 Kurth Ranch case,(16) the Court, by a five-to-four vote, ignored the 1993 Dixon test without explanation, and seemed sub silentio to apply some version of the overruled 1990 Corbin test.(17) As the modern Court itself has noted, "the decisional law in [the double jeopardy] area is a veritable Saragasso Sea which could not fail to challenge the most intrepid judicial navigator."(18)

What, in the end, are we to make of all this double jeopardy double talk, of all these double jeopardy double takes? In this Essay, I will suggest that the Double Jeopardy Clause is in fact rather simple and easy to apply. "Life or limb" connotes all criminal sanctions but never covers a mere civil suit about money. "Same offense" means just that--murder means murder, not attempted murder. And "jeopardy" begins with an indictment and ends with a suitably error-free verdict. However, the Double Jeopardy Clause itself does not exhaust the scope of constitutional principle involved in multiple prosecution and multiple punishment cases. Rather, the clean and simple rules of the Double Jeopardy Clause must be supplemented by several broader but more flexible commonsense principles protected by the Due Process Clause--and by certain other rules and principles rooted in the Sixth Amendment Jury Trial Clause. Rhetorically, the Court has tied itself into knots because it has failed to carefully disentangle the Double Jeopardy, Due Process, and Jury Trial Clauses. As a result, some defendants today are getting windfalls--needless and dangerous "get out of jail free" cards--while other defendants are getting less than they constitutionally deserve. Lawmakers, lawyers, citizens, and the Justices themselves are deprived of a clear account of exactly what the Constitution says, where, and why.

In this Essay, I seek to provide such an account.(19)

I. "LIFE OR LIMB"

Before we analyze in detail the precise meaning of the double jeopardy rule, let us consider its scope, its domain: What proceedings are covered by the rule? A hyperliteralist might insist that the Clause applies only to cases where the threatened punishment involves death or dismemberment--"life" or "limb." Although Justice Joseph Story once came close to saying as much on circuit to wriggle out of a tight spot,(20) the Supreme Court rejected this gambit long ago in Ex Parte Lange,(21) holding that the Clause applies to all criminal cases.(22) "[A]t the time this [double jeopardy] maxim came into existence," explained the Lange Court, "almost every offence was punished with death or other punishment touching the person."(23) And so the phrase "life or limb" should be understood as a vivid and poetic metaphor for all criminal punishment.

"A vivid and poetic metaphor?" our hyperliteralist might sneer. "In the Constitution? What about plain meaning?" Yes, I would respond, what about it? Surely the nuances of words and phrases are part of their plain meaning. The obviously alliterative, monosyllabic coupling of two good Old English words in the "life or limb" clause; their tethering to another fine poetic word, "jeopardy" (deriving from the French jeu-perdre, a "game" that one might "lose," and the Middle English iuparti, an uncertain game);(24) and the prominence of the unitary phrase "life or limb" in poetic and literary texts stretching back at least 500 years before the Bill of Rights(25)--all this is part of the plain meaning of the phrase. The hyperliteralist tries to play divide and conquer, splitting one phrase into two parts--"life" or "limb"--but in fact, the plain meaning yokes these words together in a single unitary phrase, "life or limb," whose whole is greater than its parts. Fidelity to constitutional text requires that we pay close heed to what the Constitution is trying to tell us; and to miss all the poetic notes here is to be not faithful, but tone-deaf. (Not literally tone-deaf, of course, but metaphorically--that's the point.)

Since words are sometimes used rather literally, and other times more metaphorically, faithful textualism must also attend to the apparent purpose and logic of a given clause. To me at least, it is hard to see why the double jeopardy principle should apply when the state wants to chop off my toes, but not when it seeks to slit my nose, or brand my skin, or gouge my ears, or flay my back. (All these, of course, were actual punishments meted out in seventeenth-century England.)(26) These are punishments that, to borrow from Lange, quite literally "touch[ ] the person";(27) and these are "games" that the government should not be able to keep playing until it wins. Imprisonment may not literally deprive me of my limbs, but chains and bars do deprive me of free use of my limbs--and so here too it seems that the spirit and purpose of the Clause obviously apply.

Nor is any of this mere special pleading on behalf of a pet clause. Consider, for example, the outlandish results that hyperliteralism would seem to demand elsewhere in the Constitution. Shall we say that because Article I addresses only "land and naval Forces"--"Armies" and "a Navy"(28)--that the Air Force is unconstitutional? Or that photographs can never receive copyright protection because the Copyright Clause speaks only of "Authors" and "Writings"?(29) Or that a handwritten private letter lacks all First Amendment protection because it is neither an oral "speech" nor the product of a printing "press"?(30) Or that a defendant may not subpoena and introduce reliable physical evidence that proves his innocence because the Sixth Amendment gives him only a right to compel the production of "witnesses" in his favor?(31)

But once we go beyond literal lives and limbs, where shall we stop? Were the Double Jeopardy Clause freestanding, I would think its life or limb imagery should obviously apply to imprisonment and all serious criminal punishments and should probably apply to petty criminal punishments (including criminal fines which, by their inherently stigmatic nature as "criminal" sanctions, rob a man of his good name as well as his purse by branding him a "criminal"),(32) but should never apply to mere civil suits involving only money. (It is, after all, precisely the obvious contrast between "life or limb," on the one hand, and mere "money," on the other, that drives the drama of Shakespeare's great meditation on law, lawyers, and hyperliteralism, The Merchant of Venice.)(33)

Happily, the Double Jeopardy Clause does not stand alone, but rather shoulder to shoulder with other clauses of the Fifth Amendment. Faithful textualism attends to the context as well as the text of a clause, and here context confirms our freestanding textual analysis. A faithful reader hearkening to nuance should not fail to note the obvious contrast between "life or limb" in the Double Jeopardy Clause and the more sweeping "life, liberty, or property" phrasing of the Due Process Clause, which clearly applies to a mere civil suit about only money. More generally, we should note the grand arc of ever-greater inclusion, the subtle spiraling outward of clausal scope, in the sequence of Fifth Amendment clauses. First comes the Grand Jury Clause, limited to "capital and otherwise infamous crime"; next comes the broader poetry of "life or limb"; then comes the perhaps broader--or at least clearer--scope of the Self-Incrimination Clause, applicable in "any criminal case"; and finally comes the still broader Due Process Clause encompassing all civil and criminal proceedings implicating "life, liberty, or property."

Of course, to say that the Double Jeopardy Clause does not reach civil cases about money(34) is not to leave defendants in these cases defenseless against state assault. Rather, it is to give them the shield of the Due Process Clause. Even if the Double Jeopardy Clause, strictly speaking, does not apply, if its rules make sense where they do apply, analogous rules could be assimilated into more general due process protection.(35) And the root commonsense idea underlying double jeopardy is generalizable beyond criminal cases: Government should not structure the adjudication game so that it is "heads we win; tails let's play again until you lose; then let's quit (unless we want to play again)."

The problem, alas, is that in its domain, the Double Jeopardy Clause has been read far beyond this commonsense norm; and so long as this is so, the damage should and can, in a principled way, be cabined to criminal cases. But a better approach would be to solve the problems at their root in double jeopardy doctrine proper. First, conviction of a lesser-included offense should not always bar a second prosecution for a greater offense. Second, an erroneous acquittal should not be immune from correction on appeal followed by a fair retrial. To see this more clearly, we must turn to the two other intriguing phrases in the Double Jeopardy Clause, "same offence" and "twice in jeopardy." Let us start with the words "same offense."

II. SAME OFFENSE

In contrast to the metaphoric and somewhat fuzzy phrase "life or limb," the phrase "same offense" feels exact, precise, mathematical. If X is the same as Y, and Y is the same as Z, then X is the same as Z by the clear and hard logic of transitivity. Rose is a rose is a rose.(36) Same means same; and done is done--this is the obvious meaning of the Double Jeopardy Clause. Thus, after a defendant has been tried and "finally" "acquitted" or "convicted" of attempted murder in a "fair" and "suitably error-free" trial--and we shall later refine the concepts in quotes(37)--he cannot be retried for the same attempted murder.

But can he later be tried for premeditated murder? The Supreme Court has put forth a test that, applied with a straight face, would always bar this second trial. Under the so-called Blockburger test, attempted murder is the same as premeditated murder.(38) More generally, a greater offense, under Blockburger, is treated as the same as any logically lesser-included offense with some but not all of the formal "elements" of the greater offense--in other words, Blockburger treats two offenses as different if and only if each requires an element the other does not.(39) Sometimes, Blockburger's rule makes sense: As we shall see, if a defendant were acquitted in a fair attempted murder trial, there would be something wrong and unconstitutional about retrying him for the "different" offense of murder.(40) But other times, Blockburger's rule makes far less sense: Suppose a defendant is convicted in a fair attempted murder trial, and the victim thereafter dies from injuries proximately caused by the initial attack. Taken seriously, Blockburger's rule would bar a murder trial here too--even if the state stood ready to offset against the murder sentence any penalty already imposed for the attempt (so as to avoid any kind of doublecounting). But as early as 1912, the Supreme Court held in the Diaz case(41) that a Blockburger-like test should not be taken seriously in this scenario, and that a second trial should not be barred.(42) Moreover, in the 1985 Garrett case,(43) the Court applied Diaz-like logic to uphold a trial for a greater continuing criminal enterprise offense after the defendant had already been convicted of a lesser-included predicate offense.(44) But other Supreme Court cases--the 1977 Brown v. Ohio case,(45) for example--have blandly ignored the deep logic of Diaz and have instead idolized Blockburger.

All this is, to put it mildly, confusing. If greater and lesser-included offenses really are the "same" offense within the meaning of the Double Jeopardy Clause, by what right do judges carve out Diaz/Garrett exceptions to clear textual commands? And if the answer is that of the Diaz and Garrett Courts--that quite obviously a greater offense is not in any commonsense or literal way the same offense as one of its lesser-included offenses(46)--then by what right do judges continue to apply Blockburger? If Diaz and Garrett are right, then isn't Brown wrong--at least in reasoning and perhaps also in result?

Now move beyond these legal puzzles to logical puzzles. As we have seen, transitivity is a basic property of sameness: If one thing is the same as each of two other things, each of those things must be the same as the other. But the Blockburger rule of sameness flunks this elementary test. Under Blockburger, murder is the "same" as the lesser-included offense of involuntary manslaughter, and is also the "same" as the lesser-included crime of attempted murder, but these two lesser offenses are not the same as each other. (Only one requires a dead body; only the other requires an intent to kill.) Under Blockburger, the lesser-included offense of robbery is the "same" as the greater offenses of armed robbery and bank robbery, but these greater offenses are not the same. (One requires a gun, the other a bank.)

Blockburger, it seems, is a mess, legally and logically. Can we do better than this? I think so.

A. Same Means Same (and Offense Means Offense)

When we consider the entire text of the Double Jeopardy Clause, and its history, a same-means-same approach makes perfect sense. At common law, the double jeopardy idea encompassed two basic pleas in bar, prior acquittal and prior conviction--in law French, autrefois acquit de meme felonie and autrefois convict de meme felonie.(47) The obvious idea here is that if a person has, on a prior occasion (autrefois) been acquitted or convicted of the exact same crime (la meme felonie) with which he is now charged, he can plead the previous judgment as a bar to the second indictment.

The logic behind this rule is simple. If Atticus has already been convicted of, say, attempted murder, in a fair trial that has rendered a final judgment, it is not right to try to heap a second punishment upon him in a second trial for what was after all only a single legal offense. A second punishment is a kind of double-counting, in which courts might end up imposing more punishment than the legislature authorized, in obvious violation of basic principles of the rule of law. If Atticus received the maximum sentence allowable at his first trial--say, ten years--then to subject him to a risk of another ten years is obviously wrong and unlawful. And, if this second trial were allowed, when would things stop? If two trials and two ten-year sentences are okay, why not three, or four, or . . . ? Anyone guilty of anything could end up being infinitely punished in defiance of statutes prescribing maximum penalties.

Here we see how the Double Jeopardy Clause, via the constitutionally guaranteed plea of autrefois convict, protects even the guilty. But the Clause is more precious for its protection of the innocent, via the constitutionally grounded plea of autrefois acquit. If Amanda is acquitted in a fair and error-free trial for attempted murder, the state may not ignore this final judgment and force her into a second trial for the same attempted murder. One obvious reason here is to protect the innocent from erroneous conviction.(48) If two trials are okay, why not three, or four, or . . . ? Eventually, the state may be able to wear an innocent Amanda down, and find one statistically aberrant or quirky jury that would erroneously convict.(49) Even if the state fails to convict Amanda in later trials, these judicial ordeals themselves can impose crushing psychic and financial costs on an innocent defendant.(50) Basic notions of symmetry are also offended by this "heads-we-win, tails-let's-play-again" scheme. If the state wins in an initial fair trial for attempted murder, it does not give the defendant the right to ignore the verdict and demand a new trial on a clean slate. Why should the defendant be placed in a lesser position when she wins? When the game is over, it's over. The winner is the winner; that's that; done is done.

And "same offense" is "same offense."(51) Even if a murder and robbery spring from the same crime spree, or the same underlying factual "transaction," they remain different legal offenses in statute books and decisional law--with different formal legal elements. They are not la meme felonie.(52) And so the Double Jeopardy Clause does not, for example, give Robert the right to plead autrefois convict to a first-degree murder charge merely because he has already been convicted of a robbery that grew out of the same set of events. Indeed, at the time of the Founding, the murder and the robbery would ordinarily not have been tried together.(53) Today, joinder rules might allow the two offenses to be litigated in a single trial, but the Double Jeopardy Clause does not demand that they be so tried.

As we shall see, general due process principles can be construed to require that factually related crimes be tried together, but this command cannot easily be crammed into the Double Jeopardy Clause in light of the syntax, grammar, purpose, and history of the Clause.(54) The Clause lays down a clear bright-line rule enforced by a special plea in bar before the second trial has even begun. And this enforcement scheme works perfectly where "same offense" means same legally defined offense; but not where "same offense" means factually related action or same overall transaction. Sometimes, the witnesses and evidence needed to prove one crime will materialize before the witnesses and evidence materialize for a factually related but legally distinct offense. But the precise amount of evidentiary overlap and divergence between "related" crimes will be highly case- and fact-specific and will often not be apparent before trial.(55) The Due Process Clause can flexibly accommodate all these concerns, and can be invoked during or after the second trial, when it becomes clear in retrospect that the second trial could easily have been folded into the first.(56) But the Double Jeopardy Clause is designed to spare Atticus and Amanda from the burden of the second trial itself: The second trial itself impermissibly "put[s]" the defendant "twice in jeopardy" in violation of the plain meaning of the Clause.(57) And so the language of the Clause is limited to the "same offense"; its grammar is absolute, not flexible and case-specific; its purpose is to prevent the second trial entirely; and its history shows it to be a paradigmatically pretrial plea.

As Robert's murder/robbery hypothetical illustrates, two offenses cannot be the same if they have different legal elements. But under the Double Jeopardy Clause, an offense must not only be the same in law--it must also be the same in fact. Even if Robert is convicted of robbery in an earlier trial, he may later be charged with and tried for robbery so long as the second indictment concerns a factually different robbery--committed, say, on a different day against a different victim. (Of course, to the extent that the two robberies grew out of the same factual "transaction" with a high degree of evidentiary overlap and a small amount of evidentiary divergence, the Due Process Clause could, as noted, be construed as a constitutionally based mandatory joinder principle.)(58) Even if the identity of the robbery victim and the day of the robbery are not formal elements of the offense of robbery--in other words, even if the legal elements in Robert's two trials are identical--Robert would have no good double jeopardy defense. He simply broke the same law twice, and thus he may be tried twice and punished twice. He may be "twice put in jeopardy of life or limb" because he committed two offenses, not one.

At times, nice "unit of prosecution" questions will arise.(59) Are two pulls of the trigger one attempted murder or two? (Does it matter if the two pulls are aimed at two different persons; or aimed at the same person, but on different days, as parts of different schemes?) Is an ongoing course of continuous conduct one offense or several? Is marrying four women one bigamy or three? (Does it matter whether the marriages occur simultaneously or sequentially?) Is a liquor store stickup in which the robber takes money from two persons one armed robbery or two? (Does it matter whether the two are both clerks, or are instead one clerk and one store patron?) These questions are both fascinating and difficult, but they are ultimately questions of substantive law, questions on which the Double Jeopardy Clause is wholly agnostic. The Clause takes substantive criminal law as it finds it; it is outlandish (and judicially unworkable) to suppose that hidden deep in the word "offense" lies some magic metatheory of substantive criminal law, telling legislators in all times and places what can and cannot be made criminal. And so it is up to the legislature to decide whether planting and exploding a bomb should be one crime or two (because the bomb was first planted, then exploded) or fifty (because fifty people died) or 500 (because 450 more were at risk) or 1,000,500 (because the bomb also destroyed one million dollars of property and each dollar of bomb damage is defined as a separate offense).(60) The Eighth Amendment's Cruel and Unusual Punishment Clause might impose limits on the total amount of punishment that can be heaped upon a person for a single "act" or series of acts,(61) but the Double Jeopardy Clause imposes no limits on how the legislature may carve up con&et into discrete legal offense units.(62)

B. Rethinking Blockburger's Boxes

In light of all this, it seems obvious that a same-offense-means-same-offense approach to the Double Jeopardy Clause enjoys clear textual, historical, and logical advantages over the Blockburger test. It remains to see how a plain-meaning approach to the Double Jeopardy Clause, when properly supplemented by various due process rules and principles, is functionally superior to Blockburger. As we shall see, Blockburger's test makes sense insofar as it (1) prevents double-counting; (2) requires joinder of certain related charges to prohibit vexatious reprosecution; (3) affirms the concept of implicit acquittal; and (4) safeguards the defendant's right to collateral estoppel. But Blockburger achieves these functional results in a crude and imprecise manner that is both over- and underinclusive. A more functional approach would admit that same means same (and offense means offense) in the Double Jeopardy Clause, and would use the Due Process Clause directly to vindicate these four principles in cases where the Double Jeopardy Clause alone cannot bear the textual load. Such a direct approach would avoid Blockburger's overprotective rigidity in cases where its test makes no sense--Diaz,(64) Brown,(64) and Garrett(65)--would also avoid Blockburger's underprotective narrowness in cases that fail to pass the Blockburger test for sameness but that nevertheless implicate one or more of the four principles. To see this more clearly, let us analyze the four logical categories--boxes--where Blockburger goes beyond true sameness: (1) prior conviction of a greater offense; (2) prior conviction of a lesser-included offense; (3) prior acquittal of a greater offense; (4) and prior acquittal of a lesser-included offense.(66)

1. Blockburger Box One: The Anti-Double-Counting Principle

Consider first a case where Robin has been convicted of armed robbery and is now charged, in a second prosecution, with robbery for the same transaction. Suppose that armed robbery carries an eight-year sentence and robbery a five-year sentence. To try Robin in two trials, and impose the maximum sentence each time, is to risk a kind of double-counting. Ordinarily, it is sensible to presume that the initial armed robbery statutory sentence already included five years for the logically lesser-included robbery, and then tacked on an additional three years for the use of a gun. And so to convict Robin of armed robbery and robbery and to cumulate the maximum sentences is in effect to double-count the robbery: (robbery plus gun) plus robbery--(five plus three) plus five. By preventing the robbery trial after the armed robbery conviction, Blockburger's rule prevents this double-counting.

But Blockburger's rule does not achieve this result very directly or precisely. For starters, the Supreme Court has now made clear that Blockburger's rule is overbroad to the extent that it irrebuttably bars cumulative punishment for logically related but distinct offenses such as robbery and armed robbery. In this doctrinal context, at least, Blockburger is now seen as simply a rule of construction creating a rebuttable presumption of sameness.(67) A legislature could, if it wanted (and spoke clearly), define armed robbery as a three-year crime to be added to a five-year robbery conviction rather than as an eight-year crime to be punished instead of a five-year robbery. But once we admit that Blockburger is only a presumption here, it is hard to insist that it furnishes a true constitutional definition of sameness.

Blockburger also underprotects the anti-double-counting principle in a number of ways. If Blockburger is indeed simply a Double Jeopardy Clause test, it applies only where a prior conviction or acquittal exists. But suppose a state prosecutes Robin in a single trial for five-year robbery and eight-year armed robbery, and announces its intention to cumulate maximum sentences. This too should be (presumptively) unconstitutional as a (presumptive) double-counting of Robin's robbery. It is as if the state simply tried Robin for robbery in one trial, and announced in advance it would double the statutory robbery punishment. This is indeed a kind of "double jeopardy"--like the "double jeopardy" round of the TV game show, when all dollar values are doubled--but one that requires some mild but real straining of words to fit into the Twice in Jeopardy Clause. Robin is being placed in a kind of "doubled" jeopardy, a kind of "two-fold" jeopardy; but is she really being placed twice in jeopardy--on two occasions, deux fois? At common law could she plead a prior (autrefois) conviction? Of course the claim here is not that doubled punishment in a single trial is constitutional, or even legal--of course not--but that the Double Jeopardy Clause fails to fit snugly here.

Even if we crammed this hypothetical into the text of the Double Jeopardy Clause, other double-counting scenarios cannot be so crammed, under Blockburger. Suppose Roberta is charged in a single trial with eight-year armed robbery and nine-year bank robbery for a single act in which she robbed a bank with a gun. Under the Blockburger test, armed robbery is not the same as bank robbery--and so the maximum penalties can be cumulated under Blockburger. But this cumulation ends up double-counting the commonpredicate robbery: (robbery plus gun) plus (robbery plus bank)--(five plus three) plus (five plus four). Notwithstanding Blockburger, this double-counting should be treated as presumptively violative of due process. If Blockburger would (presumptively) prohibit double-counting the robbery in a robbery-plus-armed-robbery trial, or in a robbery-plus-bank-robbery trial, surely the true logic at work here should (presumptively) bar the similar double-counting of the robbery in an armed-robbery-plus-bank-robbery trial.

Of course, here too, courts should apply this test only as a presumptive guide to legislative intent. Perhaps a legislature thinks that an armed bank robbery as a whole is worse than the sum of its parts. But if so, it should say so clearly in statute books--for example, by creating an offense of armed bank robbery explicitly punishable by more than twelve years (five for the robbery, three for the gun, four for the bank). If the legislature does not do so, courts should, as a simple matter of due process, presumptively deduct five years from the cumulative sentences for armed robbery and bank robbery, whether those sentences are imposed in a single trial or in two successive prosecutions.(68)

Alas, in many cases the presumptive math will not be so easy. When two offenses overlap, and the overlapping area is not itself a distinct crime with an identifiable sentence, a court might have to improvise to avoid double-counting the overlapping predicate area. (Sentencing guidelines may be a great boon here, since they may assign a specific number to the overlapping element, even if it is not itself a stand-alone crime.) And when more than two offenses are involved, the (presumptive) math can get quite tricky indeed.(69) Perhaps, in some cases, concerns about justiciability and the need for clear lines will result in some underenforcement of the anti-double-counting principle--as was evident in the famous case of Gore v. United States,(70) where various overlapping statutes (none logically greater or lesser-included than another) were brought to bear on a single transaction. But however courts ultimately address the double-counting problem, at least they should be asking the right due process question--would cumulative punishment risk double-counting, and if so, how much?(71)--rather than the mismatched, over- and underinclusive Blockburger double jeopardy question: Does each crime contain at least one element the other lacks?

2. Blockburger Box Two: The Anti-Vexation Principle

Consider next the case where a defendant seeks to plead a prior conviction of a logically lesser-included crime as a bar to prosecution on a greater offense. Here, too, Blockburger tells courts to bar the second prosecution, and in so doing both over- and underprotects the true constitutional principles at stake.

Suppose Robin is first prosecuted and convicted for a five-year robbery, and then the state seeks to prosecute her for eight-year armed robbery. Of course, if the state sought to cumulate punishment, the issue would be the same as Blockburger Box One: five plus eight is no different, or constitutionally less offensive, than eight plus five. But suppose the state stands ready to offset the five-year robbery sentence against the eight-year armed robbery sentence; suppose, in other words, that the state in the second trial is merely seeking an incremental three-year punishment for the use of the gun.

Blockburger purports to ban this second trial. Sometimes this ban makes sense; sometimes it does not. In many cases, the state may have no particularly good reason for splitting a single criminal transaction or episode into two separate trials. And in some cases, perhaps the state may have particularly bad--illegitimate--reasons. Perhaps bifurcation reflects a systematic attempt to vex or harass a defendant, by wearing her down in successive proceedings, draining her financial resources, and forcing her witnesses to appear twice on her behalf. Such vexation can create a real risk that a defendant, though innocent of the greater offense, will be erroneously convicted in the second trial.(72) Or perhaps the prosecutor is trying to evade statutory limits on prosecutorial discovery by forcing a defendant to tip her hand in the first trial--a preliminary round--so that the state, with the benefit of this "cheat peek," has an edge in the second trial, the main event.(73) Or perhaps the prosecutor is angry and vindictive after the first trial because, although the state won a conviction, the defendant largely prevailed in the sentencing, and the prosecutor seeks to punish this success--and send a message to future defendants--via a new round of charges.(74)

But once we see that these are the relevant constitutional principles at stake in Blockburger Box Two, it is clear that the rigid Blockburger test itself ill fits these principles. To the extent that Blockburger seeks to prevent Robin's harassment by, in effect, mandating joinder of robbery and armed robbery, it seems odd that Roberta's case (involving armed robbery and bank robbery growing out of the same factual transaction) and Robert's case (involving robbery and intentional murder arising out of the same factual transaction) are treated categorically differently. Why is Blockburger's mandatory joinder rule limited to greater and lesser-included offenses?(75) If the issue is the possible vexation and harassment and vindictiveness of bifurcated proceedings, why is Robin's case so different from Roberta's and Robert's?

One possible answer is that in Roberta's (bank robbery/armed robbery) case and Robert's (robbery/murder) case, the state may have good nonvexatious reasons for bifurcation. Perhaps the first-degree murder Robert committed did not occur at the precise moment of the robbery; and the two crimes involve different groups of witnesses. Perhaps it is clear that Roberta robbed a bank, but less clear that she actually used a gun, and the key witnesses and evidence about the gun are not available at the same time as the bank evidence. Because of this, whatever constitutional mandatory joinder rules are applicable in Roberta's and Robert's case must be flexible, fact- and case-specific rules of due process, rather than global, rigid, bright-line rules of double jeopardy.

But the same thing is true of Robin's (robbery/armed robbery) case! (And here we see how Blockburger, taken seriously, not only underprotects Roberta and Robert, but also overprotects Robin.) Despite due diligence, the government may not be able to prove--or even know--at the time of Robin's robbery trial that she used a gun. Suppose, for example, the armed robbery statute requires that the "gun" not be a toy gun, and at the first trial, the state lacks good evidence of a real gun.

Indeed, at the time of the first trial on a lesser-included offense, the greater offense may not even have occurred. Recall, for example, the attempted murder trial where the victim dies after the trial: Is the state forever barred from prosecuting the premeditated murder (with a sentencing offset to avoid double-counting)? The Blockburger test says yes; but the Supreme Court now says no. As early as 1912, in a case arising in the Philippines and decided under a statutory double jeopardy clause, the Diaz Court held as follows:

The provision against double jeopardy, in the Philippine Civil

Government Act, is in terms restricted to instances where the second

jeopardy is "for the same offense" as was the first. That was not the

case here. The homicide charged against the accused in the Court of

First Instance and the assault and battery for which he was tried

before the justice of the peace, although identical in some of their

elements, were distinct offenses both in law and fact. The death of the

injured person was the principal element of the homicide, but was no

part of the assault and battery. At the time of the trial for the latter

the death had not ensued, and not until it did ensue was the homicide

committed. Then, and not before, was it possible to put the accused

in jeopardy for that offense. Besides, under the Philippine law, the

justice of the peace, although possessed of jurisdiction to try the

accused for assault and battery, was without jurisdiction to try him for

homicide; and, of course, the jeopardy incident to the trial before the

justice did not extend to an offense beyond his jurisdiction .... It

follows that the plea of former jeopardy disclosed no obstacle to the

prosecution for homecide.(76)

Diaz was decided long before the Court proclaimed Blockburger as a general test of double jeopardy sameness; but recent cases have made clear that Diaz still lives.(77) But what exactly does Diaz mean, and how can it be squared with Blockburger? Read narrowly, Diaz applies only when the greater offense has not even occurred at the time of the first trial. In the 1977 case of Brown v. Ohio,(78) the Court, in a footnote, admitted that Diaz might sweep somewhat more broadly: "An exception [to Blockburger] may exist where the State is unable to proceed on the more serious charge at the outset because the additional facts necessary to sustain that charge have not occurred or have not been discovered despite the exercise of due diligence. See Diaz...."(79) But even this is far too narrow a reading of Diaz--and not merely because the Court acknowledged only that an exception may exist. Read most sweepingly, Diaz contradicts the very claim in Brown's text accompanying its Diaz footnote: "The greater offense is therefore by definition the `same' for purposes of double jeopardy as any lesser offense included in it .... Whatever the sequence may be, the Fifth Amendment forbids successive prosecution . . . for a greater and lesser included offense."(80)

Without this sweeping--and obviously wrong--textual claim, Brown's holding is hard to defend: The Brown Court precluded a trial on auto theft because of an earlier conviction on the lesser-included offense of joyriding, in a case where there was no real hint that bifurcation was a result of prosecutorial manipulation, harassment, vindictiveness, or bad faith.(81)

Eight years after Brown, the Court took a great, if somewhat unselfconscious, step away from Blockburger and Brown, and back towards a sweeping reading of Diaz. In Garrett v. United States,(82) the Court upheld a continuing criminal enterprise (CCE) prosecution against a man who had already been convicted of a lesser-included predicate drug felony. (The CCE statute required proof of three predicate offenses as evidence that a person was engaged in a continuing criminal enterprise.)(83) The Garrett Court noted that in some cases, a trial for a predicate offense could take place before the other predicate offenses--and thus the CCE offense--had even occured.(84) This tracks the narrowest reading of Diaz. The Court went on to suggest that even if the other crimes had occurred, the government, despite due diligence, would not always have evidence of this at the time of the first trial.(85) This tracks the Brown footnote's somewhat broader reading of Diaz. The Court also hinted that even if the government had such evidence, it should not necessarily have to tip off the defendant about ongoing undercover operations in order to prosecute for an initial predicate offense.(86) (A still broader reading of Diaz.) And most strikingly, the Court began its entire analysis by saying: "Quite obviously the CCE offense is not, in any commonsense or literal meaning of the term, the `same' offense as one of the predicate offenses."(87) This is the broadest reading of Diaz--and a reading which, if taken seriously, demolishes Blockburger, and leaves us instead with a more supple fact- and case-specific due process analysis asking when mandatory joinder makes functional good sense to protect defendants against prosecutorial manipulation and harassment.

3. Blockburger Box Three: The Implicit Acquittal Principle

Consider next the third Blockburger quadrant, where a defendant seeks to plead a prior acquittal on a greater offense. Suppose that Robin is tried for armed robbery and is acquitted by the jury. May the state turn around and try her for robbery?

Blockburger says no, and usually this makes good sense. The proper reasoning here is not that of avoiding double-counting or double punishment--for the first trial resulted in no punishment. Rather, the reasoning in this Blockburger quadrant rests on the idea of implicit acquittal.

At an armed robbery trial, a jury can ordinarily decide to convict the defendant of any lesser-included offense supported by the evidence. In the federal system, for example, Federal Rule of Criminal Procedure 31(c) provides that "[t]he defendant may be found guilty of an offense necessarily included in the offense charged or of an attempt to commit either the offense charged or an offense necessarily included therein."(88) Thus, at Robin's first (armed robbery) trial, she was in actual peril--in jeopardy--of a conviction for robbery. Under the well-established rule of Green v. United States,(89) when a jury renders an explicit verdict on one count in an indictment, but says nothing about another count, its silence is often best construed as an implicit acquittal. And so, ordinarily, Robin can plead autrefois acquit in her second trial because she really was tried for robbery--the same offense, la meme felonie--in the first trial, and was acquitted.(90)

Lest it seem that Blockburger always gets it right--in this quadrant at least--we must note that here, too, Blockburger's mindless equation of greater and lesser offenses is both over- and underinclusive. In a few rare situations, a trier of fact is not permitted to convict on a lesser-included charge, and in these contexts a prior acquittal should not always stand as a bar, where the acquittal was based on the government's failure to prove elements unique to the greater charge.(91) As the Court noted in the prominent 1889 case, In re Nielsen:(92)

[I]n order that an acquittal may be a bar to a subsequent indictment

for the lesser crime, it would seem to be essential that a conviction of

such crime might have been had under the indictment for the greater.

If a conviction might have been had, and was not, there was an

implied acquittal. But, where a conviction for a less crime cannot be

had under an indictment for a greater which includes it, there it is

plain that ... an acquittal would not or might not be a bar ....(93)

Conversely, suppose Roberta is explicitly acquitted in her first trial of bank robbery, and the state then seeks to prosecute her for armed robbery arising from the same events. Bank robbery and armed robbery are not the same under Blockburger--and so Blockburger would allow a second trial--but surely this second trial should be barred. If Roberta's first jury implicitly acquitted her of robbery, she cannot later be charged with the same robbery; and a fortiori she cannot later be charged with the same robbery with a gun.(94) (To think otherwise would be like saying that a person cannot be twice put in jeopardy; but thrice is just fine.)

The key point in this quadrant is not that greater and lesser-included offenses are somehow the same. Rather, it is that when a defendant is implicitly acquitted of, say, robbery, that acquittal bars all further charges that necessarily travel through that robbery. To see this more clearly still, let us consider the final Blockburger box, and the all-important constitutional right of collateral estoppel.

4. Blockburger Box Four: The Collateral Estoppel Principle

If Robin is first tried for robbery and explicitly acquitted, can she later be tried for armed robbery? Surely not, says Blockburger--and here, at last, Blockburger is surely right.

But even though Blockburger is surely right in this quadrant--it never overprotects Robin--it does underprotect other defendants. Consider, for example, defendant Robert whom the government claims was involved in a robbery and then a closely related intentional murder. The government tries Robert for robbery, and Robert claims mistaken identity: He was off on a sixmonth trek in Nepal at the time of the crime. If the jury acquits, may the government then bring a murder prosecution?

Blockburger would say yes--first-degree murder and robbery are not the same offense under its test--but we should say no. Under the constitutional principle of collateral estoppel, once a criminal defendant has prevailed against the government on an issue of ultimate fact, he should not be forced to continue to relitigate it criminally. As the Court made clear in the famous Ashe v. Swenson case in 1970,(95) the Constitution "surely protects a man who has been acquitted from having to `run the gantlet' a second time."(96) Without this principle, the government could keep prosecuting an acquitted defendant over and over on slightly different charges, wearing him down and increasing the chance that one odd jury might convict him of something. This conviction could occur even though the defendant is innocent, and prevailed on a key factual issue--such as his whereabouts during a crime spree--in a previous trial.

Ashe claimed that its collateral estoppel rule came from the Double Jeopardy Clause, not the Due Process Clause,(97) but this cannot quite be taken seriously; for Ashe applied its collateral estoppel rule to bar a second trial for an offense that was, strictly speaking, not the same offense in fact and law, nor even a greater or lesser-included offense. Rather, the Court's theory was that Bob Ashe's first jury acquitted him because it accepted his claim of mistaken identity, and that therefore he could not be charged with other offenses arising from the same set of events.(98)

Several points need to be made about the deep due process principle of collateral estoppel. First, as with Winship's asymmetric allocation of the burden of proof in criminal cases,(99) the principle is openly asymmetric in a prodefendant direction. The Ashe collateral estoppel principle bars the government from relitigating an issue of fact resolved in the defendant's favor; but it does not bar a defendant from later contesting an issue once decided in the government's favor.(100) Second, the principle is a special boon to the innocent. Whereas the formal rules of the Double Jeopardy Clause apply equally to autrefois acquit and autrefois convict, the collateral estoppel principle aids a defendant who is in effect acquitted on some contested issue.(101) Third, as a result of these two features of collateral estoppel, most prosecutors will be powerfully discouraged from attempting to bifurcate litigation in search of strategic advantage or to vex defendants.(102) If a prosecutor wins the first trial, she will have to prove everything all over again in a second criminal case; but if she loses on any issue, she loses that issue forever (in criminal cases, at least) against the defendant. Thus, the Ashe due process principle, properly understood, puts real teeth into our earlier due process principle frowning on manipulative bifurcation.

Finally, to get maximum protection from the Ashe principle, a defendant should be given the right to request a specific verdict from the first trier of fact so as to make clear precisely which issues--which elements, which ultimate facts--were decided in his favor. In light of a strong constitutional tradition upholding both the defendant's and the jury's right to a general jury verdict(103)--a simple "guilty" or "not guilty"--the most sensible approach would be to allow (but not oblige) the defendant to request specific findings from the jury after the jury has rendered its general verdict. The jury would then be allowed (but not obliged) to provide these findings, which would then furnish a solid basis for any Ashe claim a defendant might later assert.(104)

With this regime in place, an acquitted (alleged robber/murderer) Robert could benefit just as much as an acquitted (alleged robber/bank robber) Robin from the true constitutional principle at issue here. The key constitutional idea in this quadrant is not that two formally different offenses are somehow the same under the Double Jeopardy Clause--two different offenses are never the same, and no amount of judicial gibberish can make them so. Rather, the key idea is that, as a matter of due process, once a defendant has prevailed in a criminal case on a certain factual issue, he need not prove it all over again in a subsequent criminal case.

5. Beyond the Boxes: Applying the Framework

We have seen that Blockburger's test often gets it wrong in the four boxes. Moreover, no single test can work across all four boxes, given that each box contains a different cluster of functional issues. A better framework of analysis, I suggest, would insist that "same offense" means just that under the Double Jeopardy Clause, and would directly vindicate residual functional concerns using the Due Process Clause. With this revised framework of analysis in place, we are now in a position to sort out the real issues at stake in the Supreme Court's major double jeopardy decisions in the 1990s.

The decade began on an off note with the 1990 case of Grady v. Corbin.(105) Thomas Corbin was a drunk driver whose criminal behavior killed one person and seriously injured another. After pleading guilty to a couple of traffic tickets before the Town Justice Court, Corbin claimed that the Double Jeopardy Clause barred a later prosecution for reckless manslaughter and criminally negligent homicide. The Supreme Court, by a five-to-four vote, agreed.

To most people--myself included--the idea that the traffic offenses of driving on the wrong side of the road and driving while drunk are the same offense as criminally negligent homicide seems outrageous. So too does the idea that, by paying his $360 traffic tickets, Corbin deserved a "get out of jail free" card immunizing him from all serious prosecution. How on earth did five Justices (four of whom now no longer sit) reach this outlandish result?

Not by strictly applying Blockburger, since even under its decidedly expansive definition of sameness, Thomas Corbin would have lost. (Being drunk and on the wrong side of the road were not formal elements of the key offenses charged in the second trial; and dead bodies were surely no part of the traffic tickets.) But perhaps Blockburger's strained approach to the words "same offense," equating small crimes with bigger ones, emboldened the Corbin Court: One good stretch deserves another. The key fact, in the Court's mind, was that in the second trial, the state would "prove conduct that constitutes an offense for which the defendant has already been prosecuted"(106)--would prove, in other words, that Corbin was drunk and on the wrong side of the road as evidence (not elements) of his criminal negligence and recklessness. But this cannot be a Double Jeopardy Clause test because, as we have seen, double jeopardy is a pretrial plea; and before the trial, we often do not know exactly what factual conduct the prosecution will ultimately point to, and what evidence the state will ultimately invoke, to prove the formal elements of the offense.(107) (Suppose, for example, the state planned to prove Corbin's negligence and recklessness by pointing to his speeding, and not his drinking.) Thus, the Double Jeopardy Clause does not sensibly apply to factual conduct and evidence that will be proved in a second trial, but rather to the legal offense and elements that have been formally alleged in a second indictment, a second jeopardy. (Blockburger, whatever its other flaws, at least focuses squarely on legal offenses themselves, as defined in statute books and alleged in indictments.)

Of course, the Constitution does protect a defendant in a conduct-based as well as an offense-based inquiry, but only where a second trial tries to "prove conduct that constitutes an offense for which the defendant has already been [acquitted.]"(108) This is the key principle of Ashe v. Swenson;(109) but as we have seen, this is a due process principle designed to protect the innocent. The Corbin Court repeatedly invoked Ashe and various other prior acquittal cases,(110) and tried to apply symmetrically their lessons to the prior conviction context, substituting "prosecuted" for "acquitted" as the key test. The Double Jeopardy Clause is indeed symmetric--in Corbin's words, it bars "[s]uccessive prosecutions... whether following acquittals or convictions"(111)--but the collateral estoppel principle at the heart of Ashe is not. The obvious difference between Bob Ashe and Thomas Corbin is that Bob Ashe's first trial proclaimed him innocent of the conduct the state later tried to prove against him.(112)

Finally, the Corbin Court raised the specter of manipulative government bifurcation and vexation:

If Blockburger constituted the entire double jeopardy inquiry in the

context of successive prosecutions, the State could try Corbin in four

consecutive trials: for failure to keep right of the median, for driving while

intoxicated, for assault, and for homicide. The State could improve its

presentation of proof with each trial, assessing which witnesses gave the

most persuasive testimony, which documents had the greatest impact, and

which opening and closing arguments most persuaded the jurors. Corbin

would be forced either to contest each of these trials or to plead guilty to

avoid the harassment and expense.(113)

This argument fails on several counts. To begin with, it is quite far afield from the facts of Corbin, which involved not a manipulative, strategic, or vexatious prosecutorial juggernaut, but a bumbling bureaucracy. The state prosecutor had no clue in the Town Justice Court proceedings, and so this case was miles from any constitutional cap on prosecutorial bad faith. Second, even if Blockburger--or something less, like same-means-same--"constituted the entire double jeopardy inquiry," the state would not be free to try Corbin four times without constraint so long as other constitutional principles applied. The most important one here is Ashe, for it gives the state a strong disincentive to split one case into four for strategic reasons. Whenever the state prevails on a contested issue in the first trial, it will have to prove it all over again in later trials, but if it loses once on an issue, it loses forever. As we have seen, however, the Ashe principle is asymmetric, designed to protect defendants (like Bob Ashe) who prevail on contested issues and not defendants (like Thomas Corbin) who lose on those issues.

The striking injustice of Corbin is highly reminiscent of the claim rejected in Diaz, where a defendant pled guilty to a petty offense in a petty court, paid a petty fine, and then tried to use this to shield himself from a fair trial on his overall guilt. Corbin tucked Diaz away in a casual footnote,(114) a la Brown, paying no heed to Diaz's same-means-same language and ignoring its interesting argument that because a petty court simply had no jurisdiction over a major offense like homicide, the defendant could not possibly have been previously in jeopardy on the greater offense. One wonders whether Town Justice Traffic Courts in New York are equipped to try serious criminal charges; surely there are good reasons for trying traffic offenses in special traffic courts, under special procedures--perhaps without juries. Thus, the mandatory joinder demand implicit in Corbin--try everything together in one proceeding--seems hard to defend as an unyielding constitutional command on the facts of this case.

Three years later, the Supreme Court overruled Corbin, by a five-to-four vote in United States v. Dixon.(115) Four Justices (only two of whom remain on the Court) proclaimed that Corbin was right;(116) but the other five Justices insisted that Blockburger constituted the sole test of sameness. However, these five could not agree on what Blockburger meant. Three thought that by no stretch of imagination could a fifteen-year drug offense be seen as a logical lesser-included offense of a six-month contempt of court offense;(117) two Justices disagreed, and thus voted in the end with the Corbin four to immunize Alvin Dixon from drug prosecution because he had earlier been convicted of contempt for violating a court order not to commit crimes while on bail.(118) The mind boggles here: Contempt of court and drug pushing are somehow the same; a fifteen-year offense is somehow lesser than a six-month offense; because someone is found guilty of a small crime he must be set free on his big crime--and all this from Justice Scalia, who usually claims he believes in plain meaning and common sense.(119)

Even good Justices have bad days. Here is how Justice Scalia should have analyzed Dixon in keeping with his commitment to text and common sense: Alvin Dixon has no good core Double Jeopardy Clause argument since he had never been previously convicted or acquitted of the same fifteen-year drug offense with which he is now charged. Nor does he have any good claim against double-counting since Congress clearly intended contempt of court and drug offenses to be punished cumulatively (and even if not, this would argue only for a six-month setoff in the second case, not a fifteen-year windfall). Nor does he have any good implicit acquittal or Ashe-type argument, since he has never been acquitted of anything. Finally, to try his drug conduct in two separate proceedings is not constitutionally vexatious because a single hybrid proceeding should not be constitutionally required here under mandatory joinder principles. It is legitimate for the government to litigate contempt in a separate proceeding from other crimes, since a minor contempt is subject to special procedural rules (no jury, for example).(120) In this sense, the juryless tribunal that adjudged Dixon's contempt did not even have "jurisdiction" over his big drug crime--a scenario strikingly reminiscent of Diaz.

The following year, the Court again split five-to-four in a major double jeopardy case, Department of Revenue v. Kurth Ranch.(121) Montana criminally prosecuted and convicted members of the Kurth family on various drug charges. After these convictions became final and sentences were imposed, Montana continued to press civil charges against the Kurths based on a special, highly punitive civil law imposing special taxes on illegal drugs. The Court, holding that the second, civil trial, was aimed at "the same conduct"(122) as the first, sided with the Kurths and blocked the civil suit on double jeopardy grounds.

Kurth Ranch is exceedingly hard to square with Dixon, decided only a year earlier.(123) Dixon overruled Corbin and insisted on Blockburger as the sole test of sameness. But Kurth nowhere applied the Blockburger test, under which the civil and criminal charges seem obviously different.(124) In its key passages--indeed in its opening sentence--Kurth stressed that the two trials focused on "the same conduct,"(125) but this is precisely the Corbin-five test rejected by the Dixon five in favor of Blockburger's focus on legal offenses and elements rather than factual conduct. Kurth was authored by Justice Stevens--the only member of the Corbin five who still sits on the High Court--and his opinion was joined by only one of the (anti-Corbin) Dixon five, Justice Kennedy.

What Kurth said makes little sense. Following a case from 1989,(126) Kurth held that the Double Jeopardy Clause could apply to a civil trial involving only money. In light of the "life or limb" clause, this seems a real stretch.(127) If Kurth was in effect saying that the civil suit was really a criminal wolf disguised in civil sheep's clothing, well and good--but then the Double Jeopardy Clause was the least of the state's problems If the second suit was really criminal, it called for the entire panoply of applicable constitutional safeguards--appointed counsel, proof beyond reasonable doubt, and so forth(128)--and yet the Court seemed to think that none of these other protections applied. Since double jeopardy is formally symmetric between acquittal and conviction, does Kurth mean that if the Kurths had won in the criminal trial--because the state could not prove their drug conduct beyond reasonable doubt--then the state would be barred from trying to prove that same conduct, in a civil case, under a preponderance-of-the-evidence test?(129) Since the precise sequence of trials should make no difference under the Double Jeopardy Clause, does this mean that if the Kurths had rushed forward and paid their civil drug tax first, they would have been forever immune from all criminal prosecution?

But perhaps the Kurth Court--with Justice Kennedy as the key vote--was onto something after all, and just failed to package its intuition well. Even if the Double Jeopardy Clause is inapplicable to civil cases involving only money, the Due Process Clause surely applies to all proceedings involving life, liberty, or property. Due process protects a defendant's right to collateral estoppel in suits against the government, and frowns on vexatious and vindictive successive prosecutions, civil or criminal. But a particular problem exists where a state first criminally prosecutes, and then litigates in a civil suit after the criminal suit has ended. A vindictive prosecutor who wins a criminal conviction, but gets a lower punishment than she sought, or who loses a criminal suit, may be tempted to go after the defendant again in a civil suit.(130) Ordinarily, this temptation to bifurcate strategically and to double-dip is constrained ex ante by Ashe v. Swenson.(131) A prosecutor knows that if she wins on a given issue, she must always prove it all over again in another trial, but if she loses once, she loses forever; therefore she has strong ex ante incentives to try to consolidate everything into one round. But where the state sues first criminally and then civilly, Ashe may have less bite. If the state loses the first, criminal trial, it can claim that this is only because it failed to prove something beyond reasonable doubt; this failure, it can claim, should not estop it from trying to prove that same thing in a later civil case by a preponderance of the evidence.(132)

As a result of the relative weakness of Ashe in deterring vindictive multiple proceedings ex ante where criminal suits are followed by civil suits, courts must be especially vigilant to protect against vindictiveness ex post. The best analogy here is to the North Carolina v. Pearce(133) line of cases constraining vindictiveness in the resentencing context, a line in which the Court has explicitly rejected a rigid double jeopardy analysis in favor of a more flexible due process approach focused squarely on vindictiveness.(134)

If this is indeed the best thing that can be said on Kurth's behalf--and I think it is--it has profound implications for so-called reverse-Kurth cases now percolating in lower courts.(135) In these cases, citizens in various civil suits rush forward to pay their civil drug taxes and then claim that they may never be criminally prosecuted for the basic underlying drug conduct. Such cases are the "reverse" of Kurth because the civil suits come first here. If Kurth is indeed a true and pure Double Jeopardy Clause case, it is hard to deny the logic of clever drug dealers who insist that under this Clause, the sequence of trial matters not. But the basic equities in these cases--where plainly guilty people are never even tried for their crimes--feel rather different from the equities in Kurth itself, and a supple due process recharacterization of Kurth can help explain why these cases feel so different. When a civil case is litigated first, a defendant who prevails will be able to use Ashe to shield himself from an abusive criminal suit: If the state could not prove the drug conduct by a preponderance-of-the-evidence standard, it surely should be estopped from claiming the conduct occurred beyond reasonable doubt. Since Ashe works ex ante here to deter strategic bifurcation, much less ex post Kurth-like judicial protection is called for. And so even if Kurth is right (in result), it should not extend to reverse-Kurth cases.

Last term, a reverse-Kurth case did reach the Court, via the Sixth Circuit, in United States v. Ursery.(136) The Justices, however, managed to sidestep any real analysis of the key "same offense" issues.(137) At a couple of points, the Court came close to claiming that the civil suit at hand was merely an in rem action directed against a particular piece of drug-related property itself (here, a house)--and not against Guy Jerome Ursery himself, the owner and occupant of the house. On this in rem logic, Ursery could not claim that he had ever been previously convicted or acquitted in any way that would bar his criminal prosecution.(138) The in rem gimmick here is a plausible and ancient one;(139) but it can lead to striking injustices if we truly buy into the fiction that the rights of individual persons are not really at stake in in rem actions.(140) (Just ask Tina Bennis.(141)) And so the deepest lesson here is that when the Supreme Court overreads a clause in a prodefendant way, it often sets in motion a chain of events that can result in a rather drastic curtailment of citizens' rights somewhere else.(142)

Even if no curtailment ever occurred, the overreading of the Double Jeopardy Clause is of course wrong in itself--and has resulted in indefensible windfalls to various car thieves, drunk drivers, drug pushers, and other criminals. The solution, I have suggested, is to heed the words of the Clause; and, where necessary and appropriate, to supplement those words with due process principles to directly protect implicit acquittals, enforce collateral estoppel, thwart vexatious reprosecutions, and avoid double-counting. An alternative approach could label these issues as "double jeopardy principles" but apply these principles functionally and flexibly. This alternative approach has the virtue of highlighting connections between the Double Jeopardy Clause itself and the four functional concerns we have identified. But it also poses some distinctive risks. Some of those functional concerns radiate far beyond the Double Jeopardy Clause, however broadly construed. (Double-counting, for example, should be avoided in every case, civil or criminal.) Given the need to wheel in due process at some point, conceptual clarity is served if we directly apply due process whenever it fits best. And if we mean to have a genuinely flexible approach in at least some of the Blockburger boxes--to avoid the mindlessness and definitional foolishness of Brown and Corbin and Dixon--then the flexible phrasing of the Due Process Clause seems to fit better than the stiffer syntax of the Double Jeopardy Clause. Unless judges sharply distinguish between the core double jeopardy "rules"--autrefois acquit de meme felonie and autrefois convict de meme felonie--and various penumbral double jeopardy "principles," there is a real risk that the penumbra will be applied too rigidly (Brown, Corbin, and Dixon, for example) and/or that the core will be applied too mushily (the curtailment effect). Thus, the conceptually clearest approach is to insist that the Double Jeopardy Clause means just what it says, but to recognize that the principles underlying and radiating beyond the words of the Clause should inform due process analysis.(143)

III. "Twice in Jeopardy"

Let us, finally, turn to the remaining key phrase in the Double Jeopardy Clause: "twice in jeopardy." Even if a person clearly risks losing "life or limb" (in a poetic sense) for a single legal offense--"the same offense"--exactly how are we to decide whether he has been placed twice in jeopardy? The Court has approached this question by asking when jeopardy attaches, but a more precise analysis would bifurcate this inquiry. We must ask both when does a jeopardy begin and when does that same jeopardy end. Only then can we decide whether a particular event threatens a second jeopardy. Since the word jeopardy draws upon an obvious game metaphor--jeopardy is, etymologically, an uncertain game, a game that one might lose(144)--we might recast the question by asking when does the "game" begin and end. What exactly are the rules of the "game" when officials make bad calls, when the players cheat or commit fouls, or when external events like the weather intrude upon the game? As we think about these questions, and try to make them simpler to understand, perhaps various lessons and examples from other well-known games can help sort things out.

A. The Jeopardy Game

1. The Puzzle

Although courts often speak of when jeopardy attaches, this attachment metaphor misleads to the extent that it implies that there is one key moment rather than two. Jeopardy is a process--like any other game--and we thus must ask when it begins and when it ends. Consider the puzzles any other approach creates. In early American courts, judges often proclaimed that jeopardy did not attach until the jury rendered its verdict. Thus, in 1823, Justice Washington proclaimed on circuit that "jeopardy" means "nothing short of the acquittal or conviction of the prisoner, and the judgment of the court thereupon."(145) But textually, it seems a bit odd to say that one's jeopardy--one's risk, one's uncertainty--begins just at the point the uncertainty is resolved by a verdict. It is like saying that the game one might lose begins just as the final buzzer sounds. And if jeopardy connotes nothing less than a verdict (or a verdict and a sentence), does this mean that one is not put twice in jeopardy until a second verdict (or a second sentence)? Surely this cannot be right; for as we have seen, the Double Jeopardy Clause was clearly designed to be enforced by a pretrial special plea in bar to spare a person from the burden of the second trial itself (and indeed, from any pretrial impositions incident to a voidable indictment on the "same offense").

The motivation of the early American rule is clear: to allow a defendant whose first trial had not reached a final judgment of acquittal or conviction--say, because of a mistrial created by a hung jury or juror illness--to be retried. As Justice Story wrote in his classic treatise on the Constitution:

[The Double Jeopardy Clause means] that a party shall not be tried a

second time for the same offence, after he has once been convicted,

or acquitted of the offence charged, by the verdict of a jury, and

judgment has passed thereon for or against him. But it does not mean,

that he shall not be tried for the offence a second time, if the jury has

been discharged without giving any verdict; or, if, having given a

verdict, judgment has been arrested upon it, or a new trial has been

granted in his favor; for, in such a case, his life or limb cannot

judicially be said to have been put in jeopardy.(146)

This seems sensible enough except for the last line ("for, in such a case...."). Taken literally, this line would contradict Story's earlier statements by allowing a person to be tried after acquittal so long as the trial did not proceed to verdict and judgment: The defendant could "not be said to have been put in [a second] jeopardy." A friendly reworking of Story's last line would say that "although placed in jeopardy once, his first jeopardy has only begun, but has not ended in a final verdict of conviction or acquittal. His jeopardy may thus continue until a proper final verdict is reached, after which, of course, he may no longer be tried for the same offence."

More modern American courts have held that jeopardy attaches much earlier than Story and Washington thought: at the impaneling of the jury.(147) This solves some problems, leaves others unsolved, and creates new ones of its own. It seems more plausible to say that the game and the risk begin before the end, and that a person is in genuine jeopardy throughout his trial. But he is in peril even earlier, given that an indictment itself can trigger pretrial loss of liberty--loss of the free use of one's "limbs," loss of days, weeks, or even months of one's "life."(148) Surely we should not say to someone acquitted of robbery that he may be reindicted for the same robbery and held pretrial because, unless and until his second jury is seated, he will not be twice "in jeopardy."

Now consider the trial itself. The modern rule seems superior to the earlier rule in some mistrials and inferior in others. Suppose the prosecutor, near the end of a trial, senses that things are not going well for the state, and that the jury will likely acquit. Out of desperation, she engages in gross and intentional misconduct that forces the judge to declare a mistrial. She then argues that because the first trial never reached a verdict and final judgment, the defendant can be retried on a clean slate. Under the Washington/Story definition of "jeopardy" perhaps this gambit might work, but not under the modern definition: Jeopardy has already "attached." Under the modern definition however, why is a retrial permissible if a mistrial results from some other, wholly benign, cause--like a hung jury, or juror illness? The Double Jeopardy Clause seems to lay down a clear rule. By what right do judges balance away this right, saying that some second jury impanelings and thus second "jeopardies" are permissible and others are not?(149)

2. The Solution

The most sensible approach, I suggest, is as follows. The Double Jeopardy Clause in effect says that for any given offense, the government may play the adjudication game only once: No person shall be "twice put in jeopardy." The adjudication game--highlighted by the richly poetic word "jeopardy"--begins when a person is indicted. Before this event, of course, there is simply no criminal charge on the table, and thus no way to decide whether an offense is la meme felonie as the offense for which a person was autrefois convict or acquit. What's more, once an indictment does issue, a person's "life or limb," in the poetic sense, is very much at risk and in jeopardy. But indictment is only the moment when adjudication game begins; and the adjudication game does not end until there is a final winner. Until there is a final winner, the first jeopardy has not ended, and further proceedings do not place a person in jeopardy a second time--"twice."

Sometimes, a game can begin and not end; and so it must be replayed from the start later. Consider, for example, a baseball game that is called off after several innings because of rain and is then replayed on a blank scorecard at the end of the season. In a roughly analogous way, domestic law often allows a prosecutor to drop an indictment, with leave to refile--to reschedule the game--later.(150) Sometimes, a game that seemed to end must be replayed because a referee erred and that error compromised the fairness of the game. When a linesman in tennis makes an incorrect call on which the players nevertheless rely, the presiding official can overrule the call, and the point is simply played over. So too, domestic law should generally be allowed to provide that a trial judge error--even one in the defendant's favor--can be reversed, and a new trial ordered. The game need not end--jeopardy may continue--until an error-free result is obtained, and a fair winner emerges. Finally, a winner sometimes emerges because of the misconduct of the other player. A boxer who repeatedly hits his opponent with low blows is disqualified, and the other boxer is declared the winner. Some mistrials should be similarly understood: Because of the flagrant misconduct of the prosecutor, the defendant is simply declared the winner. Such a trial court ruling, if correct (and upheld on appeal), is in effect an acquittal, and thus jeopardy has indeed ended.

My approach has several major implications. First, it supports Justice Holmes's general theory of continuing jeopardy:(151) When a trial judge errs in favor of the defendant and the defendant thereby wins an alleged "acquittal" at trial, this "acquittal" need not be seen as the end of the game. Rather, an appeals court should be able to review and reverse the legal error, and remand for a new trial--a continuation of the game--until a suitably error-free result is reached. Second, my approach highlights the special role played by the Sixth Amendment in immunizing certain jury decisions from judicial review and reversals. In our system, a properly informed criminal jury given the proper evidence is always entitled to acquit a defendant; no other official can reverse that decision, no matter how "erroneous" it seems to nonjurors. On some issues, juries are the ultimate referees, who are infallible because they are final.(152) Because it is the Sixth Amendment that requires certain jury decisions to be treated as nonerroneous acquittals, the Fifth Amendment Double Jeopardy Clause here merely piggybacks, specifying the legal effect of these acquittals in future proceedings.(153) Finally, my approach helps to identify a range of penalties that can be imposed on players who cheat or commit fouls. The Due Process Clause, with its historic emphasis on fair play, is the obvious constitutional vehicle for implementing those penalties; and so the Double Jeopardy Clause itself is not really the formal source of these penalties--here too it merely piggybacks by specifying the consequences for future judicial proceedings when the form of a given penalty is, in effect, an acquittal.

B. Continuing Jeopardy

Consider first the idea of continuing jeopardy: Must a person go free because he wins something called an "acquittal" from a trial judge, even though appellate courts with general review power over the trial court deem the "acquittal" legally erroneous? Holmes thought not, and here Holmes was right.(154) To begin with, surely there is nothing magic in a name or a word. Surely it matters not at all whether a state calls a trial judge's decision an "acquittal" subject to appellate review or a "bacquittal" that only ripens into a final and true "acquittal" if upheld on appeal. In both cases, domestic law is in effect saying that a trial ruling is merely provisional, not final, and should be subject to correction if erroneous.

Holmes's basic idea is that a state could indeed structure its judicial system such that erroneous trial judge rulings are simply not final. When a defendant is convicted at trial because of a pro-state trial court error, an appellate court can review this error, reverse the conviction, and remand for a new trial. This new trial is not, constitutionally, a second jeopardy, but a continuation of the first, because the trial court "conviction" was not a true conviction. In effect, it was a "bonviction" that failed on appeal to ripen into a true "conviction." But if retrial is allowed after an erroneous trial court conviction, why not after an erroneous trial court acquittal?(155)

The usual answer will not do:(156) "When a defendant appeals a conviction, he waives any double jeopardy claim he might have against retrial; but this waiver argument does not apply when the government appeals an acquittal." The problem is that a defendant who appeals does not really "waive" anything: He does not voluntarily, knowingly, and freely relinquish his double jeopardy claim. He would love to win a reversal on appeal and be done with it--to face no further trial proceedings. But domestic law denies him this option: It forces him to "waive" his double jeopardy right in order to exercise his right to appeal. Nothing in logic requires this construction (and constriction) of options--unlike, say, the situations where logic "forces" one to forgo his right to complete silence at trial if he exercises his right to speak at trial, or where logic "forces" one to forgo her right not to have a lawyer if she exercises her right to have a lawyer. It is hard to see how a defendant could simultaneously do X and not-X; but it is easy to imagine how, in theory, a defendant could first appeal, and then (if successful) go free. England, for example, allows defendants to appeal without generally subjecting the successful appellant to retrial. To put the point another way, if we insist on saying that a defendant implicitly waives any claim against retrial when he appeals, we could likewise say that he implicitly waives any claim against a government-appeal-driven retrial when he demands the benefit of being tried under nonerroneous law in the initial proceeding: Unless a defendant renounces the right to appeal all erroneous decisions against him, he implicitly waives the right to object to any governmental appeal of any erroneous decisions in his favor.

One possible objection to continuing jeopardy is that although trial judges sometimes do err in prodefendant directions, some defendants would clearly have won anyway, and so it is unfair to subject these defendants to retrial. But this objection argues not against continuing jeopardy but for a kind of prodefendant "harmless error" doctrine, in which appellate courts should refuse to order retrials where the errors below likely did not affect the verdict. (Similarly, in football, if a pass was simply not catchable anyway, any pass interference made no difference, and so officials will not intervene.)

A second possible objection to continuing jeopardy is a variant of the first: The initial jury may have acquitted to "nullify" the prosecution altogether, and appellate judges have no right--under Sixth Amendment doctrine--to disturb this jury prerogative. But even if a jury has a shadow right to nullify(157) (a right that jurors are not told about), a jury also has a right not to nullify. The jury as an institution is not well respected if its decisions are taken as final even when those decisions are based on a trial judge's legally erroneous instructions or evidentiary rulings. No nullification prerogative would be threatened by correcting the trial judge's legal errors on appeal and remanding the case back to the same jury for retrial, this time with legally correct evidence and instructions.(158) If the jury still wants to acquit and nullify, it can then do so.

But the Constitution probably does not require even this. Even if different jurors sit in the retrial, these jurors would enjoy the same shadow right to nullify and acquit as would the original jurors in a retrial. Constitutionally speaking, a second jury simply stands in the shoes of the first--just as, in a single trial, jury alternates may be called to stand in the shoes of jurors who fall sick. Perhaps a defendant prefers his first jury, but the Constitution does not absolutely protect that preference in all contexts. When a defendant is erroneously convicted--solely because of progovernment error in the first trial that he vigorously opposed--he may be retried by a second jury even if he would prefer the first. It is hard to see why his preference should be constitutionally stronger in a retrial triggered by a successful government appeal; here, the very error that causes the retrial is one that he may have introduced rather than one that he opposed.(159)

A third objection to continuing jeopardy points to other prodefendant asymmetries in our criminal justice system, such as Winship's rule that the government bears a heavy burden of proof,(160) and Ashe's openly asymmetric rule of collateral estoppel.(161) One good asymmetry deserves another, the argument might go, and so it is constitutionally apt that defendants may appeal erroneous trial rulings but prosecutors may not. This is a bad argument chasing a bad analogy in fruitless search of a good reason. Other asymmetries in our criminal justice system are not random or purposeless; they exist to protect an innocent defendant from an erroneous conviction. Indeed, I have elsewhere suggested that this is one of the governing principles of the field of constitutional criminal procedure.(162) But no similar constitutional value is served by insulating what are--by hypothesis--incorrect judicial rulings of law from review and reversal. Insulation here is an arbitrary windfall to the guilty, not a carefully structured scheme to protect the innocent. It is as if we simply said that every third (randomly selected) defendant should go free without good reason.

A defendant has no vested right to a legal error in his favor.(163) For example, if a trial judge erroneously excludes evidence in a pretrial suppression hearing, a defendant has no constitutional right to prevent the government from seeking an interlocutory appeal to review and reverse this legal error. Subsequent proceedings are simply a continuation of the original jeopardy. But--constitutionally speaking--the matter is no different if appeal occurs after trial; here too, the subsequent proceedings are simply a continuation of a single jeopardy in search of a fair verdict based on the correct application of law. We do not demand that the defendant go free because the initial jurors catch fever and a new trial is ordered; why should things be any different if a fevered judge makes egregiously wrong legal rulings at trial?(164)

The final objection to continuing jeopardy is that it might allow a government to treat a trial court's prodefendant rulings more harshly than its pro-state rulings. In this nightmare scenario, a trial court's pro-state rulings are called "convictions," and are final and unreviewable, but a trial court's prodefendant rulings are called "bacquittals" that do not ripen into true acquittals until an endless series of onerous appeals. But a simple symmetry rule rooted in due process would prevent the state from structuring decisionmaking authority in any way that endowed pro-state rulings with more finality than prodefendant rulings.(165)

C. Jury Acquittals

So far, we see that nothing in the Constitution, properly read, bars the state from structuring its domestic law to give appellate judges plenary power to correct the errors--to "reverse the calls"--of trial judges. We must now ask, what bars the state from likewise giving appellate judges plenary power to reverse the calls of juries? The short answer is the Sixth Amendment. In civil suits, the Seventh Amendment privileges jury factfinding, but a judge can reverse the call of a jury when it is clear beyond doubt that the jury has erred in its task.(166) Under the Sixth Amendment, however, a criminal jury has the right to acquit a defendant even in the face of indisputable factual evidence of guilt.(167) And so when the jury is given the proper evidence and the proper instructions, it may acquit and no other official can overrule its call: In legal contemplation, this properly informed jury's acquittal is by definition nonerroneous. Of course, its general power to acquit with finality does not mean it must be final on all its decisions (convictions and sentencing, for example); by analogy, the home-plate umpire is the ultimate arbiter of balls and strikes generally, but not on check swings.

D. Mistrials and Misconduct

Under the continuing jeopardy principle, an innocently induced mistrial--due, say, to juror illness--can indeed be followed by retrial: Jeopardy has begun, but has not yet ended, and may resume on a blank scorecard at retrial. (The issue is rather precisely analogous to a baseball game called on account of rain after three innings.) In other contexts, where judges suspect that the prosecutor may have triggered a mistrial to avoid an impending defeat, the judges (whether at trial or on appeal) may simply declare the defendant the victor by penalty.(168) But the key principles at work in determining when judges should award an acquittal by penalty are not the hard and fast rules of the Double Jeopardy Clause--which formally come into play only after an acquittal occurs--but the flexible, case-specific fair-play ideals of due process. These due process principles determine whether a judge should, in effect, direct an acquittal; and prosecutorial motive may matter here, in a way that it does not generally matter under the Double Jeopardy Clause. Formally, all that the Double Jeopardy Clause does is specify the consequences for future proceedings if the Due Process Clause or domestic law compels the judicial award of an acquittal by penalty.(169)

If we view the misconduct/mistrial issue primarily through the prism of the Double Jeopardy Clause, we might think that the sole or most apt judicial response to prosecutorial misconduct is an award of acquittal; the Clause, as we have seen, is tightly linked to the idea of autrefois acquit. But once we see that a judicial award of acquittal is a kind of due process penalty for low blows, it becomes clear that this extreme sanction is only one of a whole set that could be devised.(170) Ordinary games from everyday life offer a series of analogies. Some minor misconduct could be ignored.(171) (Incidental contact under the boards, all part of the game.) Some more serious misconduct could be dealt with by sanctions outside the proceedings at hand, such as fines and contempt citations.(172) (A three-game suspension.) Still other misconduct may call for disgorgement of ill-gotten gain in the proceeding at hand.(173) (A touchdown called back because of offensive pass interference.) Some misconduct may trigger a defendant option--to continue a trial with a curative instruction, or to demand an immediate mistrial and new trial.(174) (An option to decline the penalty, or to accept it and replay the down.) Other types of misbehavior may justify special opportunities for one's opponent.(175) (A free penalty kick.) Still other forms of grave misconduct might be best addressed by allowing a defendant two shots at winning--either with the initial jury, or with a second jury summoned if the first jury convicts and the conviction is then set aside.(176) (A "free play" in football that gives the quarterback two chances to score.)

The broad range of possible alternative penalties suggests that judges should hesitate before awarding the extreme sanction of acquittal as a penalty. For such a sanction, of course, when cloaked with the special measure of finality of the Double Jeopardy Clause, can allow the guilty to escape without a full and fair trial of their guilt. If judges award an acquittal by penalty because the defendant was about to win his trial when the prosecution committed a low blow out of desperation, the award seems fully justified.(177) But to use such a sanction to punish or deter less egregious misconduct--especially where the defendant would likely have lost a fair first trial on the merits--is a very different thing. In many games, victory by penalty is not particularly troubling, since in the grand scheme of things it does not matter much who wins. But if the justice system is like a game in some ways,(178) it is also much more than a game in other, more important ways.(179) Lives and limbs--of victims as well as defendants--are serious business, and so it is hugely important in the criminal adjudication game to get it right: to acquit the innocent and convict the guilty.

And, as I have tried to suggest throughout this Essay, constitutional interpretation is also serious business; here too it is important to try to play by the rules, and to get it right. In this Essay, I have tried to do both.(180)

Akhil Reed Amar, Southmayd Professor of Law, Yale Law School. (1.) U.S. CONST. amend. V.

(2.) See, e.g., United States v. Ursery, 116 S. Ct. 2135, 2142-47 (1996); see also Department of Revenue v. Kurth Ranch, 114 S. Ct. 1937 (1994); United States v. Halper, 490 U.S. 435 (1989).

(3.) See Blockburger v. United States, 284 U.S. 299 (1932). In fact, the Blockburger case itself does not quite stand for the global test of sameness that later courts have attributed to it. See infra note 38. For an example of a modern-day application of the so-called Blockburger test, see, e.g., Brown v. Ohio, 432 U.S. 161, 166-69 (1977), which barred prosecution for auto theft on double jeopardy grounds because the defendant had been previously convicted of the lesser-included offense of joyriding.

(4.) Garrett v. United States, 471 U.S. 773, 786 (1985) (holding that continuing criminal enterprise (CCE) offense is not same as any of its lesser-included predicate crimes).

(5.) See Crist v. Bretz, 437 U.S. 28, 32-38 (1978).

(6.) See, e.g., id. at 34 n.10; Arizona v. Washington, 434 U.S. 497, 505-16 (1978); Wade v. Hunter, 336 U.S. 684, 688-92 (1949).

(7.) United States v. Jenkins, 420 U.S. 358 (1975).

(8.) See id. at 365-70 (generally barring any government appeal that would require retrial).

(9.) United States v. Scott, 437 U.S. 82 (1978).

(10.) See id. at 95-101 (allowing erroneous ruling to be appealed and case retried if ruling was not related to factual guilt or innocence).

(11.) See infra Part III (defending "continuing jeopardy" approach that would permit appeal of virtually all erroneous rulings by trial judge).

(12.) Grady v. Corbin, 495 U.S. 508 (1990), overruled by United States v. Dixon, 113 S. Ct. 2849 (1993).

(13.) See id. at 521 (holding that in second teal, government may not "prove conduct that constitutes an offense for which the defendant has already been prosecuted").

(14.) United States v. Dixon, 509 U.S. 688 (1993).

(15.) See id. at 703-12 (overruling Corbin and restoring Blockburger as sole test of sameness).

(16.) Department of Revenue v. Kurth Ranch, 114 S. Ct. 1937 (1994).

(17.) For more analysis, see infra text accompanying notes 105-35.

(18.) Albernaz v. United States, 450 U.S. 333, 343 (1981).

(19.) A roadmap may be helpful here. In Part I, I explore the scope of the Double Jeopardy Clause. Does it, for example, ever apply to civil cases? In Part II, I ask when two offenses should be deemed the same, and I identify the functional concerns that should animate analysis here. In Part III, I examine when jeopardy begins and ends--attaches--with particular emphasis on the issues of continuing jeopardy, erroneous jury verdicts, and mistrials. To put the point textually rather than functionally, Part I explores the phrase "life or limb"; Part II explains the words "same offense"; and Part III expounds the trope "twice in jeopardy."

(20.) See United States v. Gibert, 25 F. Cas. 1287, 1294-97 (C.C.D. Mass. 1834) (No. 15,204) (finding "life or limb" clause inapplicable to misdemeanors); see also OFFICE OF LEGAL POLICY, U.S. DEP'T OF JUSTICE, REPORT TO THE ATTORNEY GENERAL ON DOUBLE JEOPARDY APPEALS OF ACQUITTALS, No. 6 (1987) [hereinafter OLP REPORT], reprinted in 22 U. MICH. J.L. REFORM 831, 842 (1989) (presenting originalist evidence that "life or limb" was term of art for felonies); cf. id. at 862 n.108 (declining to press this point in light of well-established case law reading phrase more broadly to encompass misdemeanors too).

(21.) 85 U.S. (18 Wall.) 163 (1873).

(22.) See id. at 170-73.

(23.) Id. at 173.

(24.) See 8 THE OXFORD ENGLISH DICTIONARY 214 (2d ed. 1989).

(25.) See id. at 956 (quoting source from 1205 using phrases "lif & leomen" and "leome & vppe lif"; source from 1300 using phrase "leme or lif"; source from 1362 using phrase "his lyf and his leome"; source from 1480 using phrase "lyf and lymme"; and source from 1548 using phrase "lifes and lymmes").

(26.) See, e.g., 4 WILLIAM BLACKSTONE, COMMENTARIES *376-77.

(27.) See supra text accompanying note 23.

(28.) U.S. CONST. art. I, [sections] 8, cls. 12-14.

(29.) Id. cl. 8.

(30.) Id. amend. I. For an earlier exploration of this analogy, see Akhil Reed Amar, The Case of the Missing Amendments: R.A.V. v. City of St. Paul, 106 HARV. L. REV. 124, 134 & n.66 (1992).

(31.) See U.S. CONST. amend. VI. For more discussion and analysis of this point, see Akhil Reed Amar, Foreword: Sixth Amendment First Principles, 84 GEO. L.J. 641, 647-49 (1996).

(32.) Cf. WILLIAM SHAKESPEARE, OTHELLO, act 3, sc. 3 ("Who steals my purse steals trash .... But he that filches from me my good name/Robs me of that which not enriches him/And makes me poor indeed."). Note the metaphoric link here to physical branding. See supra text accompanying note 26.

(33.) For a brilliant meditation on these (and other) themes of the play, see Kenji Yoshino, The Lawyer of Belmont, 9 YALE J.L. & HUMAN. (forthcoming Summer 1997).

(34.) To be sure, a genuine criminal punishment masquerading as a "civil" sanction can and should be unmasked by courts; but this unmasking would summon up all the Constitution's criminal procedure protections, not merely the Double Jeopardy Clause. See infra text accompanying note 128.

It is also worth noting that not until the late 1980s did the Supreme Court ever embrace the novel notion that the Double Jeopardy Clause could be stretched to cover some civil suits about money. See United States v. Halper, 490 U.S. 435 (1989). For an analysis of how a Due Process Clause analysis would have better fit the facts of Halper, see infra text accompanying notes 130-35.

(35.) See infra text accompanying notes 130-43.

(36.) Cf. GERTRUDE STEIN, SACRED EMILY (1913).

(37.) See infra Part III.

(38.) See Blockburger v. United States, 284 U.S. 299, 304 (1932). On its facts, Blockburger propounded a test for identifying when two offenses could be charged in a single proceeding, resulting in two convictions and possibly cumulative sentences. This test thus addressed what I shall later call the "doublecounting" problem. See infra Subsection II.B.1. Later courts, however, applied Blockburger's test as a general measure of double jeopardy sameness for situations involving successive prosecutions. See, e.g., Brown v. Ohio, 432 U.S. 161, 168 (1977) (noting that greater and lesser-included offenses are "by definition the `same'" for double jeopardy purposes).

(39.) See Blockburger, 284 U.S. at 304.

(40.) See infra Subsections II.B.3-4 (discussing implicit acquittals and collateral estoppel).

(41.) Diaz v. United States, 223 U.S. 442 (1912).

(42.) See id. at 448-49. This case is discussed infra text accompanying note 76.

(43.) Garrett v. United States, 471 U.S. 773 (1985).

(44.) See id. This case is discussed infra text accompanying notes 82-87.

(45.) 432 U.S. 161 (1977). This case is discussed infra text accompanying notes 78-81.

(46.) See Garrett, 471 U.S. at 786; Diaz, 223 U.S. at 448-49.

(47.) See 4 BLACKSTONE, supra note 26, at *335-36; 2 MATTHEW HALE, THE HISTORY OF THE PLEAS OF THE CROWN *240-55; 2 WILLIAM HAWKINS, A TREATISE OF THE PLEAS OF THE CROWN 368-79 (Garland Publ'g, Inc. 1978) (1721). At common law, a third plea, autrefois attains complemented autrefois acquit and autrefois convict. Courts have also linked double jeopardy to the plea of prior pardon. See United States v. Scott, 437 U.S. 82, 87 (1978); United States v. Wilson, 420 U.S. 332, 340 (1975).

(48.) Elsewhere, I have suggested that the protection of the innocent from erroneous conviction is one of the root principles of our system of constitutional criminal procedure. See, e.g., Amar, supra note 37, at 642 46; Akhil Reed Amar, The Future of Constitutional Criminal Procedure, 33 AM. CRIM. L. REV. 1123 (1996). See generally AKHIL REED AMAR, THE CONSTITUTION AND CRIMINAL PROCEDURE: FIRST PRINCIPLES (1997).

(49.) Thus, there is a resonance between the autrefois acquit idea and the Winship principle that seeks to avoid erroneous convictions by placing a heavy burden of proof on the government. See In re Winship, 397 U.S. 358 (1970).

(50.) See Amar, supra note 31, at 658-62.

(51.) Cf. George Kannar, The Constitutional Catechism of Antonin Scalia, 99 YALE L.J. 1297, 1323 n.128, 1324, 1337 (1990) (tracing similar tropes in writings of Justice Scalia).

(52.) See, e.g., 4 BLACKSTONE, supra note 26, at *336 ("It is to be observed that the pleas of autrefois acquit and autrefois convict . . . must be upon a prosecution for the same identical act and crime.") (emphasis altered); 2 HALE, supra note 47, at *245-46 (stating that burglary and larceny committed "at the same time" are different offenses; prior acquittal of one will not bar prosecution of other); 2 HAWKINS, supra note 47, at 371 ("It seems that it is no Plea to an Appeal of Larceny, That the Defendant hath been found Not guilty in an Action of Trespass brought against him by the same Plaintiff for the same Goods; for Larceny and Trespass are entirely different."); id. at 376 (offering similar example). The foregoing sources, and many others, are discussed in Grady v. Corbin, 495 U.S. 508, 530-35 (1990) (Scalia, J., dissenting). That discussion, in turn, was heavily relied on in United States v. Dixon, 509 U.S. 688, 709-10 (1993).

(53.) See Ashe v. Swenson, 397 U.S. 436, 453 (1970) (Brennan, J., concurring) ("English common law . . . severely restricted the power of prosecutors to combine several charges in a single trial."); 1 JAMES FITZJAMES STEPHEN, A HISTORY OF THE CRIMINAL LAW OF ENGLAND 508 (William S. Hein & Co., Inc. 1980) (1883) ("In an indictment for felony one offence only can practically be charged.").

(54.) But see Ashe, 397 U.S. at 453-54 (Brennan, J., concurring) (arguing that Double Jeopardy Clause generally requires that all offenses arising out of "same transaction" be prosecuted in single proceeding).

(55.) See Corbin, 495 U.S. at 529-30 (Scalia, J., dissenting).

(56.) Cf. United States v. MacDonald, 435 U.S. 850, 858-59 (1978) (making similar point about Speedy Trial Clause).

(57.) See Corbin, 495 U.S. at 529-30 (Scalia, J., dissenting); Abney v. United States, 431 U.S. 651, 660-62 (1977). (58.) I place the word "transaction" in quotes because this word is hardly self-defining, as even the chief proponent of a "same transaction" rule has acknowledged. See Ashe, 397 U.S. at 454 n.8 (Brennan, J., concurring).

In light of the incentive effects created by other (relatively easily enforced) constitutional rules--namely, the asymmetric rule of collateral estoppel--a global mandatory joinder rule for offenses growing out of the same "transaction" is probably unnecessary and unwise as a matter of due process (though textually permissible). For more explanation, see infra text accompanying notes 102, 130-32.

(59.) For an excellent discussion, see Peter Westen & Richard Drubel, Toward a General Theory of Double Jeopardy, 1978 Sup. Ct. Rev. 81, 111-22.

(60.) See id. at 114 ("There is simply no way to make sense out of the notion that a course of conduct is `really' only one act, rather than two or three, or, indeed, as many as one likes."); Comment, Twice in Jeopardy, 75 Yale L.J. 262, 275-77 (1965) (making similar point).

(61.) See Nancy J. King, Portioning Punishment: Constitutional Limits on Successive and Excessive Penalties, 144 U. Pa. L. Rev. 101 (1995); Westen & Drubel, supra note 59, at 114.1 place the word "act" in quotes here because, like "transaction," "act" is hardly self-defining. Cf. supra note 60; see also supra note 58. By contrast, an "offense" on my view connotes a precise and discrete set of elements and penalties defined by statute or decisional law; and so the word is much easier to define and apply. Cf. King, supra, at 182 (arguing that Eighth Amendment prohibits excessive overall punishment for single act, and noting that "[t]he concept of punishment under the Eighth Amendment, unlike the textual reference to `offence' in the Fifth Amendment, is not limited by statutory elements or legislative intent").

(62.) See generally Westen & Drubel, supra note 59, at 111-22; see also cases cited infra note 67.

(63.) Diaz v. United States, 223 U.S. 442 (1912).

(64.) Brown v. Ohio, 432 U.S. 161 (1977).

(65.) Garrett v. United States, 471 U.S. 773 (1985). Diaz, Brown, and Garrett are discussed infra Subsection II.B.2.

(66.) In 1995, Jon Marcus and I sketched out a preliminary critique of Blockburger that set the stage for the more thoroughgoing critique I shall now offer. See Akhil Reed Amar & Jonathan L. Marcus, Double Jeopardy Law After Rodney King, 95 Colum. L. Rev. 1, 28-38 (1995). Consider the present Essay, if you will, my own double take on double jeopardy.

(67.) See Garrett, 471 U.S. at 778-79; Ohio v. Johnson, 467 U.S. 493, 499 n.8 (1984); Missouri v. Hunter, 459 U.S. 359, 366-07 (1983).

(68.) See King, supra note 61, at 194 n.269.

(69.) Imagine, for example, three crimes, A+B, A+C, and B+C. The first crime (A+B) is punishable by five years, the second (A+C) is also punishable by five years, and the third (B+C) is punishable by four years. Presumably, the legislature here deems A alone worthy of three years; B worthy of two; and C worthy of two. But exactly what is a judge to do if she finds two more crimes on the statute books, A+D, and B+D, punishable by four years apiece? The simple presumptive (additive) math breaks down.

(70.) 357 U.S. 386 (1958) (upholding six convictions and three consecutive sentences arising out of two narcotics sales, in violation of three statutes; first statute banned drug sale without "a written order," second banned drug sale without "stamped package," and third banned sale of illegally imported drug). The case is discussed in Comment, supra note 60, at 303-04.

(71.) Suppose two statutes share a common substantive legal element, but the element is described in slightly different language in the two statutes. (Say, one uses the word "intentionally" and the other, "purposefully.") The anti-double-counting principle turns on things, not words, and so two verbal formulas describing the same thing should (presumptively) be treated as defining the same legal element. Cf. Amar & Marcus, supra note 66, at 38-14 (putting forth similar argument for double jeopardy analysis in dual sovereignty context).

(72.) As we shall see, other constitutional principles that are--or at least should be--in place can severely constrain the prosecutor's incentive to strategically bifurcate her prosecution. See infra text accompanying notes 102, 130-32.

(73.) A similar "peek," of course, can lawfully occur when a defendant is tried, convicted, and then--because of impermissible progovernment error--the conviction is set aside and the defendant is retried. See infra Part III.

One obvious way to reduce the prosecutor's incentive to bifurcate litigation and thus evade statutory limits on discovery is to allow liberal discovery against the defendant in the first place. Renee Lettow and I have elsewhere explained why the Fifth Amendment Self-Incrimination Clause, rightly understood, poses a much smaller obstacle to pretrial discovery against defendants than is conventionally assumed. See Akhil Reed Amar & Renee B. Lettow, Fifth Amendment First Principles: The Self-Incrimination Clause, 93 Mich. L. Rev. 857 (1995) (arguing that defendant can be forced to tell all pretrial, with only narrow "testimonial immunity" granted, enabling prosecutor to use all testimonial fruit, but not defendant's words themselves, at trial). Here we see a way in which overprotecting some defendants' rights--in this case rights to resist discovery-may end up hurting defendants in other ways.

(74.) Cf. United States v. Goodwin, 457 U.S. 368, 372, 384 (1982) (affirming that due process prohibits prosecutors from bringing new vindictive charges to penalize defendant for exercising right to jury trial on initial charge, but declining to presume vindictiveness without specific proof); Blackledge v. Perry, 417 U.S. 21 (1974) (applying due process principles to constrain prosecutorial vindictiveness after successful defendant appeal); North Carolina v. Pearce, 395 U.S. 711 (1969) (developing due process test to protect against judicial vindictiveness when judge "loses" to defendant who successfully appeals judge's verdict).

(75.) See Westen & Drubel, supra note 59, at 162-63 (noting how Blockburger test establishes narrow mandatory joinder rule).

(76.) Diaz v. United States, 223 U.S. 442, 448-49 (1912) (citations omitted).

(77.) For recent favorable citations of Diaz, see, e.g., Rutledge v. United States, 116 S. Ct. 1241, 1250 n.17 (1996); Grady v. Corbin, 495 U.S. 508, 516 n.7 (1990); Garrett v. United States, 471 U.S. 773, 791-93 (1985); id. at 796-98 (O'Connor, J., concurring); id. at 803-04 (Stevens, J., dissenting); Brown v. Ohio, 432 U.S. 161, 169 n.7 (1977); and Jeffers v. United States, 432 U.S. 137, 151-52 (1977) (plurality opinion). See also Illinois v. Vitale, 447 U.S. 410, 420 n.8 (1980) (referring to Diaz principle, but not by name).

(78.) 432 U.S. 161 (1977).

(79.) Id at 169 n.7 (emphasis added).

(80.) Id. at 168-69 (emphasis added).

(81.) Nothing in the Brown Court's opinion suggests any governmental bad faith. Thus we may assume that the police in the county where Brown was caught and immediately charged with joyriding simply did not know of the prosecutorial plans of charging authorities from another county, where the defendant had stolen the car nine days earlier.

Brown also raised a double-counting issue: Should Brown have received a setoff for the one month he served for joyriding against his six-month car theft sentence? But this concern should not have precluded a second trial itself; it merely went to the proper amount of additional punishment to be meted out after a second conviction.

(82.) 471 U.S. 773 (1985).

(83.) See id. at 775.

(84.) See id. at 788-89, 791-93.

(85.) See id. at 789.

(86.) See id. at 789-90.

(87.) Id. at 786.

(88.) Fed. R. Crim. P. 31(C).

(89.) 355 U.S. 184, 190-91 (1957) (stating that when jury convicts on lesser-included offense without rendering explicit verdict on greater offense, "the great majority of cases in this country have regarded the jury's verdict as an implicit acquittal on the [greater] charge.... [W]e believe [this implicit acquittal assumption] legitimate.") (alternative holding); see also Grady v. Corbin, 495 U.S. 508, 533-34 (1990) (Scalia, J., dissenting) ("`Thus an acquittal on an indictment for murder will be a good bar to an indictment for manslaughter [because] had the defendant been guilty, not of murder but of manslaughter, he would have been found guilty of the latter offence upon that indictment ...."') (quoting Commonwealth v. Roby, 30 Mass. (12 Pick.) 496, 504 (1832)), overruled by United States v. Dixon, 113 S. Ct. 2849 (1993).

(90.) The attentive reader will note that here, at least, the Double Jeopardy Clause itself can bear the load--perhaps, however, with a little help from lenity principles construing the jury's arguably ambiguous verdict as an Implicit acquittal on robbery.

(91.) For further elaboration, see Amar & Marcus, supra note 66, at 33.

(92.) 131 U.S. 176 (1889).

(93.) Id at 189-90 (emphasis added); see also Grady v. Corbin, 495 U.S. 508, 531-32 (1990) (Scalia, J., dissenting) (unselfconsciously quoting similar observation of Charles Petersdorff in 1825), overruled by United States v. Dixon, 113 S. Ct. 2849 (1993).

(94.) Cf. Corbin, 495 U.S. at 531 (Scalia, J., dissenting) ("`But if one charge consists of the circumstances A. B. C. and another of the circumstances A. D. E., then, if [A] does not itself constitute a distinct substantive offense, an acquittal from the one charge cannot include an acquittal of the other."') (quoting 1 Thomas Starkie, Criminal Pleading 323 (2d ed. 1822)) (emphasis added).

(95.) 397 U.S. 436 (1970).

(96.) Id. at 446 (quoting Green v. United States, 355 U.S. 184, 190 (1957)).

(97.) This claim enabled the Ashe Court to sidestep an earlier case with virtually identical facts, decided against the defendant under the Due Process Clause. See Hoag v. New Jersey, 356 U.S. 464 (1958). But the Ashe Court could have distinguished away Hoag simply by reinterpreting due process in light of the intervening incorporation of the Double Jeopardy Clause against states in Benton v. Maryland, 395 U.S. 784, 793-96 (1969). This intervening event, the Court should have argued, properly exerted a new gravitational pull on due process principles. See infra text accompanying note 143.

(98.) Ashe was suspected of having robbed six poker players around a poker table. In his first trial, the government prosecuted him for robbing one of the players, Donald Knight, and Ashe won an acquittal on grounds of mistaken identity. In the second trial, Ashe was charged with a formally different robbery--of one of the other five players. Each robbery was a distinct offense for "unit of prosecution" purposes, see supra text accompanying note 59; in other words, Ashe could initially have been charged with six offenses, received six convictions, and served six cumulative sentences, see Ashe, 397 U.S. at 446.

(99.) See In re Winship, 397 U.S. 358, 362-64 (1970) (finding due process requires governmental proof beyond reasonable doubt in criminal cases).

(100.) See Ashe, 397 U.S. at 443 (noting "lack of `mutuality"') (citation omitted); see also Simpson v. Florida. 403 U.S. 384, 386 (1971) (per curiam) (reaffirming lack of "mutuality"; convicted defendant is not estopped in second trial from denying facts resolved against him in first trial).

(101.) This factor should be especially important for those who believe, as I do. that constitutional criminal procedure should protect the innocent from erroneous conviction without needlessly advantaging the guilty. See generally AMAR, supra note 48.

(102.) See United States v. Dixon, 509 U.S. 688, 710 n.15 (1993) (noting that because of Ashe asymmetry, prosecutors "have little to gain and much to lose from [a bifurcation] strategy").

(103.) The jury's right to a general verdict is intimately connected to the jury's right to acquit against the evidence. I do not challenge this right here--on the contrary. See infra text accompanying notes 152, 157-58, 167.

(104.) For earlier discussions of this point, see Amar & Marcus, supra note 66, at 33 n.166, and Cynthia L. Randall, Comment, Acquittals in Jeopardy: Criminal Collateral Estoppel and the Use of Acquitted Act Evidence, 141 U. Pa. L. Rev. 283, 317-25 (1992). Presumably, the voting rule in the jury for any special verdict would be the same one prescribed by domestic law for general verdicts. Note that, under a broad view of Ashe's asymmetry, a defendant who prevails on a given issue could have the right not only to prevent any future criminal prosecution based on that issue, but also to vacate any past criminal conviction so based.

(105.) 495 U.S. 508 (1990).

(106.) Id. at 521.

(107.) Justice Scalia has made this point well. See id. at 529-30 (Scalia, J., dissenting); see also supra text accompanying notes 54-57.

(108.) Cf. supra text accompanying note 106.

(109.) 397 U.S. 436 (1970).

(110.) See Corbin, 495 U.S. at 518 n.8, 518-19.

(111.) Id. at 518.

(112.) See id. at 528 (Scalia, J., dissenting) (affirming Ashe's conduct-based analysis where earlier adjudication resolved factual issue "in the defendant's favor") (emphasis added); United States v. Dixon, 509 U.S. 688, 705 (1993) ("[Ashe] may bar a later prosecution for a separate offense where the Government has lost an earlier prosecution involving the same facts. But this does not establish that the Government `must . . . bring its prosecutions . . . together.' It is entirely free to bring them separately, and can win convictions in both.").

(113.) Corbin, 495 U.S. at 520-21 (citation omitted).

(114.) See id. at 516 n.7.

(115.) 509 U.S. 688, 704 (1993).

(116.) See id. at 740 (White, J., concurring in judgment in part and dissenting in part); id. at 741 (Blackmun, J., concurring in judgment in part and dissenting in part); id. at 744 (Souter, J., joined by Stevens, J., concurring in judgment in part and dissenting in part).

(117.) See id at 718 (Rehnquist, C.J., joined by O'Connor & Thomas, JJ., concurring in part and dissenting in part).

(118.) See id. at 697-700 (Scalia, J., joined by Kennedy, J.)

(119.) See, e.g., California v. Acevedo, 500 U.S. 565, 581-85 (1991) (Scalia, J., concurring in judgment) (championing plain meaning and commonsensical approach to Fourth Amendment reasonableness); Minnick v. Mississippi, 498 U.S. 146, 166-67 (1990) (Scalia, J., dissenting) (arguing for plain meaning and commonsensical understanding of Fifth Amendment compulsion under Self-Incrimination Clause); see also Grady v. Corbin, 495 U.S. 508, 529-30, 540-43 (1990) (Scalia, J., dissenting) (rejecting majority analysis on textual grounds that "offense" in Double Jeopardy Clause means legally-defined crime, not factual transaction, and that "put in jeopardy" precludes second trial itself, thus calling for pretrial plea; and on commonsensical grounds that majority approach would be unworkable and would lead to absurd results), overruled by United States v. Dixon, 113 S. Ct. 2849 (1993).

(120.) See Dixon, 509 U.S. at 692-93 (describing bench trial in contempt proceeding). According to the Court, the Constitution does not require a jury in a "petty" criminal case where the authorized term of imprisonment is six months or less. See Baldwin v. New York, 399 U.S. 66 (1969); Bloom v. Illinois, 391 U.S. 194 (1968).

(121.) 114 S. Ct. 1937 (1994).

(122.) Id. at 1941.

(123.) This point is made well in King, supra note 61, at 120-22.

(124.) See id. at 121 (noting that civil and criminal statutes contained different elements, and that Montana legislature intended them as separate and cumulative sanctions). How, exactly, could Montana enforce both laws against a single person on the Kurth Court's view? By bringing a single hybrid civilcriminal case? It seems hard to say that such a virtually unprecedented proceeding could really be required by the Constitution.

(125.) 114 S. Ct. at 1941; see id. at 1943 ("the same criminal conduct"); id. at 1947 ("the precise conduct").

(126.) See United States v. Halper, 490 U.S. 435 (1989) (holding that double jeopardy bars civil penalty proceeding filed after criminal conviction).

(127.) See supra Part I.

(128.) See Kurth, 114 S. Ct. at 1960 (Scalia, J., dissenting).

(129.) But see infra note 132.

(130.) The facts of Kurth itself were rather more complicated than in Halper. In Halper, the civil penalty proceeding was initially filed after the conclusion of the criminal case; in Kurth, civil proceedings were initiated while the criminal suit was pending, and continued after the criminal suit ended. The sequence of events in Halper raised a stronger inference of prosecutorial vindictiveness than in Kurth. Cf. Halper, 490 U.S at 451 n.10 ("[W]hen the Government already has imposed a criminal penalty and seeks to impose additional punishment in a second proceeding, the Double Jeopardy Clause protects against the possibility that the Government is seeking the second punishment because it is dissatisfied with the sanction obtained in the first proceeding.").

(131.) 397 U.S. 436 (1970).

(132.) Cf. Dowling v. United States, 493 U.S. 342, 347-50 (1990). In this case, the Court declined to use Ashe to bar the government from introducing evidence that the defendant had in fact committed an earlier crime, for which he had been previously acquitted. The Court reasoned that the second jury could believe that even though defendant was not guilty of the first crime beyond reasonable doubt he was nonetheless probably guilty, and that this probable guilt was relevant to his guilt beyond reasonable doubt of the second crime. Put a different way, since the government did not need to prove in the second case that the defendant committed the first crime beyond a reasonable doubt, its theory was not logically inconsistent with the earlier acquittal. For similar discussions of the hole in collateral estoppel created by the gap between the reasonable doubt and preponderance standards, see United States v. One Assortment of 89 Firearms, 465 U.S. 354, 361-62 (1984); and One Lot Emerald Cut Stones v. United States, 409 U.S. 232, 235 (1972) (per curiam).

(133.) 395 U.S. 711 (1969).

(134.) See, e.g., id. (using Due Process Clause to constrain judicial vindictiveness toward defendant who successfully overturns judge's ruling on appeal and then faces stiffer sentence on retrial; judge must specifically justify sentence increase); Wasman v. United States, 468 U.S. 559 (1984) (further elaborating Pearce); cf. United States v. Goodwin, 457 U.S. 368, 372, 384 (1982) (using Due Process Clause to prohibit prosecutorial vindictiveness toward defendant who exercises right to jury trial, but requiring specific proof of vindictiveness); Blackledge v. Perry, 417 U.S. 21 (1974) (using Due Process Clause to constrain prosecutorial vindictiveness toward defendant who exercises right to appeal).

(135.) Several of these cases are discussed in King, supra note 61, at 123 n.70.

(136.) 116 S. Ct. 2135 (1996). At the Supreme Court, the Sixth Circuit case was consolidated with a case originating in the Ninth Circuit. See id. at 2138. Because the Sixth Circuit fact pattern presents a clearer example of reverse-Kurth issues, I shall focus on it here.

(137.) See id. at 2140 n.1.

(138.) See id. at 2140, 2145 (relying on in rem fiction of Various Items of Personal Property v. United States, 282 U.S. 577, 581 (1931)). But see id. at 2148 n.3 (claiming that Court's ultimate conclusion does not rest on "the long-recognized fiction that a forfeiture in rem punishes only malfeasant property rather than a particular person").

(139.) See id. at 2140-42 (invoking congressional statute from 1789, and several nineteenth- and twentieth-century Supreme Court cases).

(140.) See id. at 2151 (Kennedy, J., concurring) ("It is the owner who feels the pain and receives the stigma of the forfeiture, not the property.").

(141.) See Bennis v. Michigan, 116 S. Ct. 994 (1996) (using in rem gimmick to uphold hardhearted forfeiture of family car against wholly innocent wife of wrongdoing husband).

(142.) For many more examples of this phenomenon, see Amar, supra note 48, at 1138. (43.) See generally John Hart Ely, Democracy and Distrust 87-101 (1980) (recommending that open-textured constitutional clauses such as Due Process be informed by values underlying more specific constitutional clauses); Akhil Reed Amar, Fourth Amendment First Principles, 107 Harv. L. Rev. 757, 804-11 (1994) (following similar approach to help define "constitutional reasonableness" under Fourth Amendment).

It might be asked whether my approach leaves any independent bite to the Double Jeopardy Clause itself. If the Clause did not exist, couldn't the Due Process Clause itself do all the work? Perhaps, but this is hardly a forceful argument against a plain-meaning reading of the Double Jeopardy Clause. Many of the clauses of our Constitution, when best read, may simply textually clarify what would probably be the best reading of the Constitution, even without that clause. Many of the clauses of the Sixth Amendment, for example, clarify rules that might otherwise have been implicit in due process. See generally Amar, supra note 31. For a general discussion of the ubiquity of such clarifying clauses in our Constitution, see Akhil Reed Amar, Some Opinions on the Opinion Clause, 82 Va. L. Rev. 647, 648-51 (1996).

(144.) See supra text accompanying note 24.

(145.) United States v. Haskell, 26 F. Cas. 207, 212 (C.C.E.D. Pa. 1823) (No. 15,321).

(146.) Joseph Story, Commentaries on the Constitution of the United States [sections] 1781, at 659-60 (Fred B. Rothman & Co. 1991) (1833).

(147.) See. e.g., Crist v. Bretz, 437 U.S. 28, 32-38 (1978).

(148.) See Amar, supra note 31, at 658-61.

(149.) The Court has said that mistrials followed by retrials may be granted in cases of "manifest necessity." See Arizona v. Washington, 434 U.S. 497, 505 (1978). But the Court has hastened to add that "it is manifest that the key word `necessity' cannot be interpreted literally." Id. at 506. More double jeopardy double talk? For a superb account of the actual judicial standards at work in mistrial cases, see Westen & Drubel, supra note 59, at 85-106.

(150.) See Westen & Drubel, supra note 59, at 88, 97-99 (distinguishing between harassing and nonharassing indictment dismissals followed by reindictment).

(151.) See Kepner v. United States, 195 U.S. 100, 134 (1904) (Holmes, J., dissenting) ("[L]ogically and rationally a man cannot be said to be more than once in jeopardy in the same cause, however often he may be tried. The jeopardy is one continuing jeopardy from its beginning to the end of the cause.").

(152.) Cf. Brown v. Allen, 344 U.S. 443, 540 (1953) (Jackson, J., concurring in judgment) ("We are not final because we are infallible, but we are infallible only because we are final.").

(153.) Ordinarily, domestic law should define what counts as an "acquittal" just as it defines what counts as an "offense." And ordinarily, the Double jeopardy Clause itself should take these domestic law concepts as it finds them. But just as other constitutional clauses can constrain what domestic law can treat as a criminal "offense"--an "offense" for example cannot be ex post facto--so here the Sixth Amendment demands that domestic law must treat certain events as "acquittals." See also infra text accompanying notes 165, 168-69 (describing similar role for Due Process Clause).

(154.) See Kepner, 195 U.S. at 134 (Holmes, J., dissenting) (quoted supra note 151); see also Palko v. Connecticut, 302 U.S. 319 (1937) (upholding Connecticut continuing jeopardy regime under Due Process Clause). Palko, of course, was decided before the Court held that the Double Jeopardy Clause directly applied against states via Fourteenth Amendment incorporation. See Benton v. Maryland, 395 U.S. 784, 793-96 (1969).

(155.) See Kepner, 195 U.S. at 135 (Holmes, J., dissenting) (explaining that defendant "no more would be put in jeopardy a second time when retried because of a mistake of law in his favor, than he would be when retried for a mistake that did him harm").

(156.) My analysis in this paragraph is obviously heavily indebted to Professor Peter Westen's work which addresses the relevant issues with great penetration and power. I shall thus only summarize his arguments here; readers with any lingering doubts should directly confront Peter Westen, The Three Faces of Double Jeopardy: Reflections on Government Appeals of Criminal Sentences, 18 Mich. L. Rev. 1001, 1008-09, 1055-59 (1980); and Westen & Drubel, supra note 59, at 125-28.

(157.) Here, too, I stand on the shoulders of Professor Westen, who has thoughtfully addressed the interplay between double jeopardy doctrine and jury nullification theory. See Westen, supra note 156, at 1012-23; Westen & Drubel, supra note 59, at 129-32.

(158.) Here, I go one step beyond Professor Westen, who seems to overlook this possibility and its larger logical implications. See Westen, supra note 156, at 1017 n.58; Westen & Drubel, supra note 59, at 130 & n.230.

(159.) See Westen, supra note 156, at 1009; Westen & Drubel. supra note 59. at 127-28.

(160.) See In re Winship, 397 U.S. 358, 362-64 (1970).

(161.) See Ashe v. Swenson, 397 U.S. 436 (1970).

(162.) See supra note 48.

(163.) See Lockhart v. Fretwell, 506 U.S. 364, 366, 370-71 (1993) (stating that criminal defendant is "not entitled" to "windfall" of incorrect application of law or to "`the luck of a lawless decisionmaker'" when "`the state court make[s] an error in his favor'") (citations omitted); see also United States v. Scott, 437 U.S. 82, 91 n.7 (1978) (allowing appellate reversal of erroneous trial court acquittal, where result is to reinstate jury's guilty verdict).

It might be argued that, under Ashe, we immunize an "erroneous" jury finding of fact from reversal by a second jury; why shouldn't we likewise immunize an "erroneous" trial judge ruling of law? The answer is that, in legal contemplation, we have no good reason to systematically prefer the second jury's finding as generally more likely to be "true" or "accurate" than the first jury's finding. Perhaps it is really the second jury that is erroneous in thinking that the first erred. But the entire structure of judicial appellate review gives us good reason to think that an appeals court--typically with more judges and more time for careful legal study of difficult issues--is generally more likely to be legally correct than a trial judge. Thus, we have structured our system so that appellate judges can generally reverse trial judges; but one jury cannot generally reverse another.

In addition, as we saw in Part II, Ashe can be defended as creating a worthy and workable incentive structure discouraging prosecutorial vexation and strategic bifurcation. No such vexation exists in the typical case where a prosecutor seeks reversal and retrial to correct an erroneous prodefendant ruling by the trial judge; appeal and retrial here would typically occur because the defendant--over the prosecutor's objection--persuaded the trial judge to commit legal error. If we generally allow retrial after an erroneous conviction (where the state is to blame for the initial error, and retrial can be vexing), we should allow retrial after an erroneous acquittal a fortiori.

(164.) Note that I do not argue here that every American jurisdiction must or should provide for continuing jeopardy, but only that government should be free to do so if it chooses. Perhaps many legislatures will decide that in certain subcategories or across the board, the appeal and retrial game is not worth the candle. As with my work elsewhere, see supra note 48, here I address only questions of constitutional criminal procedure.

(165.) While finding the logic of continuing jeopardy compelling, the Office of Legal Policy (OLP) rejected the idea on originalist grounds in 1987. See generally OLP Report, supra note 20. The main argument of the OLP was not textual but historical: England at the time of the Founding rejected continuing jeopardy (and so did many later American courts).

This argument blurs the key difference between those legal rules that were effectively inscribed in constitutional text, and those that were not. The English cases, I submit, rested on the interaction of two distinct rules: (1) The double jeopardy principle barred a retrial after a true (that is, a suitably error-free and final) acquittal; and (2) domestic English law at the time did not provide for appeals of certain trial court error. Thus, trial court acquittals were suitably error-free and final in the sense that domestic English jurisdictional law chose to endow them with finality. But this second rule of English judicial structure was nowhere locked into the American Constitution; and if the American judicial structure were to depart from the English model, then the first principle--which was inserted into the text of the Double jeopardy Clause--would interact differently. The Double Jeopardy Clause demands that if a true acquittal occurs, retrial is barred, but it looks to current domestic law, not the law of England in 1789, to determine what should count as a true acquittal.

Analogously, we might ask whether double jeopardy today applies to RICO offenses. An OLP-like approach might lead us into the following absurdity: Since old England (and early America) did not view RICO as an offense, double jeopardy does not apply. The correct approach, of course, is to see two distinct rules interacting 200 years ago: (1) The double jeopardy principle applied to all criminal offenses; and (2) domestic law at the time did not make RICO an offense. Only the first principle was inscribed into the Double Jeopardy Clause. The Clause demands that if RICO is a criminal offense, retrial is barred; but it looks to current domestic law, not the law of England in 1789, to determine what should count as an offense. See supra text accompanying notes 51-62.

To put the point one final way, recall that England does not generally allow a defendant to be retried after he successfully appeals his conviction. The current American rule, of course, ordinarily permits retrial after an appellate court reverses a conviction tainted by legal error. But, as we have seen, this retrial is permissible only because it simply continues the initial jeopardy, not because defendant somehow "waives" his double jeopardy rights. If retrial after erroneous conviction is allowed, so must be retrial after erroneous acquittal: The formal rules of the Double Jeopardy Clause are symmetric between acquittals and convictions. (Due process is, of course, asymmetric, and protects acquittals more; but there is no strong reason to specifically protect legally erroneous acquittals.) England, in short, cannot be our guide here because its domestic judicial structure sharply depute from America's, and nothing in the Double Jeopardy Clause freezes in place England's domestic jurisdictional rules for appeals. Ct United States v. DiFrancesco, 449 U.S. 117 (1980) (allowing appellate court to increase sentence imposed by trial court where domestic law authorizes such appeals, and paying no heed to whether English appellate courts in 1789 had such power as matter of domestic law).

(166.) See Amar, supra note 31, at 685-86.

(167.) Yet again, I echo Professor Westen. See Westen & Drubel, supra note 59, at 129-32 (deriving this right of criminal juries from series of doctrinal rules concerning directed verdicts, special verdicts, verdict inconsistency, and collateral estoppel); Westen, supra note 156, at 1012-23 (providing similar derivation).

(168.) See Green v. United States, 355 U.S. 184, 188 (1957); Westen & Drubel, supra note 59, at 85-106.

(169.) Alternatively, we could see acquittals by penalty as penumbral double jeopardy "principles" designed to prevent purposeful evasion of the core double jeopardy rule: A prosecutor who intentionally derails a case she is about to lose should stand in no better position than a prosecutor who in fact does lose (and is thus barred from retrial by the Double Jeopardy Clause's formal rule of autrefois acquit).

The labeling issue here--are acquittals by penalty double jeopardy or due process issues?--parallels the issues discussed supra Part II.

(170.) Thus, my analysis of the Double Jeopardy Clause complements my analysis of other constitutional criminal procedure clauses. Elsewhere I have argued for alternative remedies to various kinds of exclusionary rules--under the Fourth Amendment, the Fifth Amendment Self-Incrimination Clause, and the Sixth Amendment Speedy Trial Clause--that needlessly free the guilty. See Amar, supra note 48.

(171.) Cf. Chapman v. California, 386 U.S. 18, 22 (1967) (propounding harmless error rule).

(172.) This type of sanction is of course quite common today.

(173.) In theory, this is what happens when a trial judge strikes from the record, and instructs the jury to ignore, counsel's improper question and any answer it elicited.

(174.) See Westen & Drubel, supra note 59, at 102 nn.114 & 116, 104-05 (discussing this remedial possibility).

(175.) Cf. Rosemary Nidiry, Note, Restraining Adversarial Excess in Closing Argument, 96 Colum. L. Rev. 1299, 1319-23 (1996) (discussing tit-for-tat "invited response" doctrine).

(176.) In fact, this is apparently current doctrine's approach to certain types of mistrials. See Westen & Drubel, supra note 59, at 101-102, 106. Once we see this theoretical possibility, it is conceptually possible to argue that the most apt remedy in a particular case would be a requirement that the penalized party win three times in a row, or four, or more.

Query whether this penalty could ever be imposed against a defendant for his low blows, or the low blows of his counsel. A judge could never award a victory by penalty to the government--such a ruling would be tantamount to a directed verdict and would violate the Sixth Amendment right of a jury to acquit against the evidence. But even accepting this asymmetry between the government and the defendant, must we also say that penalties less severe than victory by default can never be assessed against defendants who cheat in the adjudication game?

(177.) See supra note 169.

(178.) See Arthur Allen Leff, Law and, 87 Yale L.J. 989, 998-1005 (1978) (describing "Ludic Metaphor" and "The Game of the Trial" and concluding that if Leff's fictional "Trial is not a game, it is not not a game either").

(179.) See Akhil Reed Amar, Hits, Runs. Trial Error: How Courts Let Legal Games Hide the Truth, Wash. Post, Apr. 16, 1995, at C1 ("A criminal trial is not a football game, even if it stars O.J. Simpson.").

(180.) Playing by the rules of constitutional interpretation requires fidelity to constitutional text (among other things), but there are different ways that one can be faithful to a text. This Essay has featured three slightly different approaches. In Part I, fidelity to the poetic phrase "life or limb" called for literary sensitivity; in Part II, fidelity to the mathematical word "same" and the distinctly legal word "offense" called for a more legalistic and logic-chopping analytic framework; and in Part III, fidelity to the game metaphor at the root of the word "jeopardy" invited comparisons and contrasts with other familiar games.