Dario Restrepo appeals his forty-six month prison sentence imposed after he was convicted of two counts of distribution of cocaine. A panel of this court issued an opinion on May 8, 1990. United States v. Restrepo, 903 F.2d 648 (9th Cir.1990). We granted Restrepo's petition to rehear the case en banc for the limited purpose of considering whether the preponderance of the evidence standard of proof for factors enhancing a sentence under Sentencing Guideline § 1B1.3(a)(2) satisfies due process. We conclude that it does, as a general rule. The facts, and parts one and three, of the earlier opinion remain unchanged and will not be repeated here. Part two of the earlier opinion is withdrawn and replaced with the following:

It is undisputed that in pre-Guidelines practice, a sentencing judge was free to consider or to decline to consider any and all information about a defendant's background and relevant conduct without a requirement that the information meet any particular standard of proof. See McMillan v. Pennsylvania, 477 U.S. 79, 91, 106 S. Ct. 2411, 2418, 91 L. Ed. 2d 67 (1986). Restrepo argues that because the Sentencing Guidelines severely reduce the judge's discretion and mandate the sentencing effect of uncharged crimes, due process requires something more than a preponderance of the evidence standard of proof for uncharged crimes under the Guidelines.

In commissioning and approving the Guidelines, Congress offered no guidance regarding a standard of proof for sentencing factors. Also, the Sentencing Commission expressly declined to decide the question presented in this case: "Among the legal issues that may have to be resolved are: .... What is the weight of the burden of persuasion (i.e., is it sufficient to prove the asserted factor by a preponderance of the evidence or is a higher degree of certainty required)?" Supplementary Report on the Initial Sentencing Guidelines and Policy Statements 47 n. 79 (1987). Consequently, we are left to determine the appropriate standard of proof for uncharged conduct used at sentencing based on relevant case law and due process principles.

The Supreme Court has not considered the precise question whether the preponderance of the evidence standard of proof for sentencing factors under the Guidelines satisfies due process. Every circuit that has considered the question has answered affirmatively, including a panel in an earlier case in this circuit. See, e.g., United States v. Wilson, 900 F.2d 1350, 1353-54 (9th Cir.1990); United States v. Frederick, 897 F.2d 490, 492-93 (10th Cir.), cert. denied, --- U.S. ----, 111 S. Ct. 171, 112 L. Ed. 2d 135 (1990); United States v. Alston, 895 F.2d 1362, 1372-73 (11th Cir.1990); United States v. Wright, 873 F.2d 437, 441-42 (1st Cir.1989); United States v. Guerra, 888 F.2d 247, 251 (2d Cir.1989), cert. denied, --- U.S. ----, 110 S. Ct. 1833, 108 L. Ed. 2d 961 (1990); United States v. McDowell, 888 F.2d 285, 290-91 (3d Cir.1989); United States v. Urrego-Linares, 879 F.2d 1234, 1237-38 (4th Cir.), cert. denied, 493 U.S. 943, 110 S. Ct. 346, 107 L. Ed. 2d 334 (1989); United States v. Silverman, 889 F.2d 1531, 1535 (6th Cir.1989); United States v. White, 888 F.2d 490, 499 (7th Cir.1989); United States v. Gooden, 892 F.2d 725, 727-28 (8th Cir.1989), cert. denied, --- U.S. ----, 110 S. Ct. 2594, 110 L. Ed. 2d 274 (1990).1

Although the Supreme Court has not addressed the precise question, an analogous pre-Guidelines case from the Court supports the currently unanimous conclusion by the Courts of Appeal that the preponderance standard generally satisfies due process under the Guidelines. In McMillan, 477 U.S. 79, 106 S. Ct. 2411, the Court considered whether, when a state legislature eliminates judicial discretion and sets a predetermined effect at sentencing for the visible possession of a handgun, due process requires something more than a preponderance of the evidence to establish the possession. The Court first determined that Pennsylvania did not transgress constitutional due process limits when it defined the visible possession of a handgun as a sentencing factor rather than as an element of a crime. Id. at 84-91, 106 S.Ct. at 2415-18. The Court reached this conclusion because the use of handgun possession as a sentencing factor does not (1) negate the presumption of innocence, or relieve the prosecution of the burden of proving guilt of the underlying crime, id. at 87, 106 S.Ct. at 2416, or (2) alter the maximum penalty available for the crime committed, or (3) create a separate offense calling for a separate penalty, id. at 88-90, 106 S.Ct. at 2417-18.

After the Court established that the visible possession of a handgun can be properly classified as a sentencing factor, it addressed the question whether due process might nevertheless require that the factor be proven by at least clear and convincing evidence at sentencing. The Court answered resoundingly in the negative:

We see nothing in Pennsylvania's scheme that would warrant constitutionalizing burdens of proof at sentencing.

... We have some difficulty fathoming why the due process calculus would change simply because the legislature has seen fit to provide sentencing courts with additional guidance.

Id. at 92, 106 S.Ct. at 2419 (footnote omitted). The result is not based on any peculiarities of the sentencing factor at issue in McMillan, but is based on the bifurcation of the criminal justice system into a conviction stage and a sentencing stage, with meaningful distinctions in the purposes, interests, and burdens involved in each stage. It was important to the Court that:

criminal sentencing takes place only after a defendant has been adjudged guilty beyond a reasonable doubt. Once the reasonable-doubt standard has been applied to obtain a valid conviction, the criminal defendant has been constitutionally deprived of his liberty to the extent that the State may confine him.

Id. at 92 n. 8, 106 S. Ct. at 2419 n. 8 (internal citation omitted). The Court emphasized that even a clear and convincing standard of proof at sentencing, rather than a preponderance standard, "would significantly alter criminal sentencing." Id.

McMillan holds that the preponderance standard for fact-finding generally is adequate to protect any interests a properly convicted defendant retains at sentencing. We believe that McMillan controls this case. The Guidelines do not differ from the Pennsylvania statute in any manner that significantly affects the due process conclusion in McMillan.2 As did the Pennsylvania statute,3 the Sentencing Guidelines have taken traditional sentencing factors and given them a predetermined effect. Like the Pennsylvania statute, the Guidelines, both as a general matter and specifically in the section at issue here (§ 1B1.3(a)(2)), do not (1) negate the presumption of innocence or the prosecutor's burden of proving guilt with regard to the underlying crime in the conviction stage of the criminal justice process, or (2) alter the maximum penalty available for the crime committed,4 or (3) create new offenses requiring separate punishment.

Obviously, the Guidelines herald an historic shift in sentencing procedure that is vastly different in scope from the statute at issue in McMillan. The dissent ably articulates the radical nature of this change: the sentencing judge's role, in large measure, no longer extends to controlling the effect of any of his or her findings on the sentence. However, we do not believe that this radical change has the effect on the constitutionally required standard of proof at sentencing that the dissent suggests.

Even before the Guidelines were enacted, a convicted criminal had a protected interest at sentencing.5 The criminal justice system assumed that the best way to protect that interest, as well as society's interest in just punishment, was to give the sentencing judge full discretion to consider all relevant information and to determine the effect, if any, the information ought to have on the sentence. The Guidelines are the product of a change of heart. Congress no longer believes that discretionary sentencing adequately serves the defendant's interest in a fair sentence and has decided to confer greater protection by providing uniform guidelines for sentencing. This change of heart, however, implicates the Supreme Court's warning against the erroneous reasoning that when a legislature undertakes to provide sentencing guidance, the standard of proof automatically increases at sentencing: "Pennsylvania's decision to [assign a predetermined effect to a sentencing factor] has not transformed against its will a sentencing factor into an 'element' of some hypothetical 'offense,' " thereby requiring a heightened standard of proof. McMillan, 477 U.S. at 90, 106 S.Ct. at 2418.

The bifurcation of the administration of criminal justice into a conviction stage and a sentencing stage has not changed under the Guidelines. The purpose of the Guidelines was not to make sentencing factors any more or less important relative to the offense of conviction in the sentencing process than they were under pre-Guidelines procedures. Rather, the purpose was to make the effect sentencing factors have on the sentence more uniform and predictable. Guidelines Manual, ch. 1, part A(3). Indeed, the Sentencing Commission made a studied attempt to conform the Guidelines to an empirical base that showed the effects judges traditionally gave to sentencing factors under pre-Guidelines procedures. Id. The Commission anticipated that some would criticize their approach "as overly cautious, as representing too little a departure from pre-guidelines sentencing practice," but it believed that the goal of reducing sentence disparity could be accomplished without a more radical change in philosophy. Id. at part A(5) (emphasis added). The intent to perpetuate the distinction between the interests and burdens of the trial stage and the sentencing stage under the Guidelines is also supported by the express direction of Congress and the Commission that the Guidelines not limit the evidence admissible at sentencing. 18 U.S.C. § 3661 ("No limitation shall be placed on the information concerning the background, character, and conduct of a person convicted of an offense which a court of the United States may receive and consider for the purpose of imposing an appropriate sentence."); U.S.S.G. § 6A1.3(a); see also U.S.S.G. § 1B1.4.

The preponderance of the evidence standard of proof is consistent with the greater formality in factfinding required by the Guidelines. We believe that the greater formality itself--such as fully adversarial factfinding procedures and the right to appeal sentencing findings--adequately protects the convicted defendant's interest without the need for a higher standard of proof.6 The Chairman of the Sentencing Commission, William Wilkins, and the Commission's General Counsel, John Steer, support this position in their article, "Relevant Conduct: The Cornerstone of the Federal Sentencing Guidelines":

The guidelines enhance procedural fairness by largely determining the sentence according to specific, identified factors, each of which a defendant has an opportunity to contest, through evidentiary presentation or allocation [sic], at a sentencing hearing. The advent of guideline sentencing thus presents no convincing reason to conclude that constitutional standards are somehow stricter when guidelines are used to assist in fashioning the appropriate sentence, or that policy considerations compel use of a higher standard. Hence, courts should apply the guideline adjustments within the realm of Relevant Conduct when those adjustments are established by the preponderance of the evidence.

41 S.C.L.Rev. 495, 519 (1990) (footnotes omitted). The Sentencing Commission, itself, sought to avoid too great an increase in the formality of sentencing procedures under the Guidelines:

A fact-finding process for sentencing decisions that has all the attributes of a formal trial could consume many times the resources devoted to the resolution of guilt or innocence. Ultimately, such an approach would render the sentencing process completely unworkable.

The Guidelines have made the defendant's interest in a fair sentence more defined and protectible. However, we emphasize that the convicted defendant's liberty interest is not an interest in the maximum guideline sentence set by the offense of conviction alone, as the dissent maintains. The Supreme Court has recognized that due process protects a defendant's interest in fair sentencing, but has emphasized in the same cases that the interest is not defined as a liberty interest in a sentence below the statutory maximum. McMillan, 477 U.S. at 92 n. 8, 106 S. Ct. at 2419 n. 8 ("Once the reasonable-doubt standard has been applied to obtain a valid conviction, the criminal defendant has been constitutionally deprived of his liberty to the extent that the State may confine him." (Internal quotation omitted)); Gardner, 430 U.S. at 358, 97 S.Ct. at 1204 ("[T]he sentencing process, as well as the trial itself, must satisfy the requirements of the Due Process Clause. Even though the defendant has no substantive right to a particular sentence within the range authorized by statute, the sentencing is a critical stage of the criminal proceeding....").

A convicted defendant has an interest in the accurate application of the Guidelines within statutory limits, nothing more, nothing less.8 That interest is a far cry in weight and importance from the liberty interest enjoyed by the defendant at trial. The statute for the offense of conviction sets the constitutional parameters of a possible sentence. Once those limits are established by a valid conviction on proof beyond a reasonable doubt, the defendant's liberty interest has been greatly reduced. However, factfinding is still necessary under some legislative schemes to set the sentence accurately within statutory limits, such as the Pennsylvania statute in McMillan and the Guidelines. The teaching of McMillan is that, as a general matter, due process is satisfied by a preponderance of the evidence standard of proof for that factfinding. 477 U.S. at 92, 106 S. Ct. at 2419.

The Supreme Court did recognize in McMillan that there may be an exception to the general rule that the preponderance standard satisfies due process when a sentencing factor has an extremely disproportionate effect on the sentence relative to the offense of conviction. 477 U.S. at 87-91, 106 S.Ct. at 2416-18. However, the extremely disproportionate effects that might lead the Supreme Court to require a higher standard of proof at sentencing have nothing in common with this case.

For example, in Mullaney v. Wilbur, 421 U.S. 684, 95 S. Ct. 1881, 44 L. Ed. 2d 508 (1975), the Court considered a Maine statute that required a defendant who otherwise might be convicted of murder to prove that he acted in the heat of passion on sudden provocation in order to avoid a murder conviction (which carried with it a mandatory term of life imprisonment) and, instead, be convicted of manslaughter (for which the maximum penalty was twenty years imprisonment). The Court pointed out that almost from the inception of the common law, the presence or absence of heat of passion has been the single most important factor in determining a homicide defendant's degree of culpability. Id. at 696, 95 S.Ct. at 1888. The Court also emphasized the significant difference in the stigma and sentences associated with murder and manslaughter and held that, consequently, due process requires the prosecution to prove at trial absence of heat of passion beyond a reasonable doubt. Id. at 698-700, 703-04, 95 S. Ct. at 1889-90, 1892.

Similarly, in Specht v. Patterson, 386 U.S. 605, 87 S. Ct. 1209, 18 L. Ed. 2d 326 (1967), the Court held that a defendant convicted under a statute carrying a maximum sentence of ten years could not be sentenced under another statute based on certain findings when the separate statute prescribed an indeterminate period of incarceration of from one day to life imprisonment, the necessary findings were not ingredients of the original offense, and the findings were made without procedural safeguards approaching those afforded at trial. Id. at 608-10, 87 S.Ct. at 1211-12. Indeed, the Court indicated that, in view of the consequences, the factors relevant in Specht amounted to a new criminal charge against the defendant. Id. at 610, 87 S.Ct. at 1212.9

The nature and effect of the disputed sentencing factors in this case differ markedly from the factors at issue in Mullaney and Specht. Under Guideline § 2D1.1, the quantity of cocaine involved in a defendant's uncharged conduct raises his or her offense level.10 In this case, the judge found that in addition to the 37.5 grams of cocaine involved in the two counts for which Restrepo was convicted, he was responsible for 65.83 grams associated with the conviction of his codefendant, DeMaldonado. Like the firearm possession in McMillan, 477 U.S. at 89, 106 S.Ct. at 2417, such a finding is not an uncommon sentencing factor. Nor does it drastically affect the length of his sentence, as in Mullaney and Specht. Furthermore, unlike Mullaney, the finding at issue here did not become relevant until after the defendant had been convicted. Cf. McMillan, 477 U.S. at 87, 106 S. Ct. at 2416. And, unlike Specht, the Guidelines did not expose this defendant, once convicted, to statutory penalties other than those originally specified for the offense of conviction. Indeed, as in McMillan, the statutory penalties to which the defendant was exposed have remained unchanged with the enactment of the Guidelines. Cf. 477 U.S. at 86, 89, 106 S. Ct. at 2416, 2417.11

The district court's finding at sentencing that Restrepo was responsible for an additional 65.83 grams of cocaine raised Restrepo's offense level from fourteen to eighteen, potentially increasing his sentence by a minimum of twelve months, or a maximum of twenty months, depending on his criminal history category. See Guidelines Manual, ch. 5, part A (Sentencing Table). The finding simply is not "a tail which wags the dog of the substantive offense." McMillan, 477 U.S. at 88, 106 S.Ct. at 2417. Whether such an offending tail will yet arise under the Guidelines in this circuit is a question for the future.12

We hold, as a general rule and as a rule appropriate for the facts of this case, that due process does not require a higher standard of proof than preponderance of the evidence to protect a convicted defendant's liberty interest in the accurate application of the Guidelines. We emphasize that the preponderance of the evidence standard is a meaningful one that requires the judge to be convinced "by a preponderance of the evidence that the fact in question exists." United States v. Streeter, 907 F.2d 781, 792 (8th Cir.1990); see also 9 J. Wigmore, Evidence § 2498 (Chadbourn rev. 1981). It is a " 'misinterpretation [of the preponderance test] that it calls on the trier of fact merely to perform an abstract weighing of the evidence in order to determine which side has produced the greater quantum, without regard to its effect in convincing his mind of the truth of the proposition asserted.' " Winship, 397 U.S. at 367-68, 90 S.Ct. at 1074 (quoting Dorsen & Rezneck, In Re Gault and the Future of Juvenile Law, 1 Fam.L.Q. No. 4, 26-27 (1967)).

It appears that in Restrepo's case, the district court found that the government had proved Restrepo's and DeMaldonado's participation in a common scheme by convincing evidence. The court said:

Now, in this case, the facts plainly support the conclusion and no other conclusion that the drugs which Judith Maldonado [sic] sold in Count Three and the drugs which she possessed in Count Four and the drugs that were found at her house late when she turned them in, were all part of the common scheme....

I'm satisfied certainly by a preponderance of evidence and in fact, I think, by clear and convincing evidence that Dario Restrepo supplied the cocaine that Judith Maldonado [sic] had at her house and that she sold on March 1, the other cocaine that he wasn't convicted of.

Transcript of Sentencing Proceedings at 64-65. We conclude that the district court applied, at a minimum, the preponderance of the evidence standard of proof and is, therefore, beyond reproach.

Restrepo's sentence is AFFIRMED.

I join the majority opinion, but write separately to point out what I believe is an important consideration. Specifically, the severity of penal consequences associated with a sentencing factor may in some cases tip the balance toward requiring heightened procedural protections in determining the applicability of that factor. Here, however, the penal consequences of the findings in question are not so severe as to merit additional safeguards when, as appears from Judge Wiggins's opinion, other relevant factors are taken into account.

In McMillan v. Pennsylvania, 477 U.S. 79, 106 S. Ct. 2411, 91 L. Ed. 2d 67 (1986), the Supreme Court observed:

Petitioners' claim that visible possession [of a firearm during the commission of certain criminal offenses] under the Pennsylvania statute is "really" an element of the offenses for which they are being punished--that Pennsylvania has in effect defined a new set of upgraded felonies--would have at least more superficial appeal if a finding of visible possession exposed them to greater or additional punishment, cf. 18 U.S.C. § 2113(d) (providing separate and greater punishment for bank robberies accomplished through "use of a dangerous weapon or device"), but it does not.

Id. at 88, 106 S. Ct. at 2417. Similarly, in the present case, the Sentencing Guidelines, and the findings made thereunder, have no effect on the maximum penalties authorized by statute.

But this does not conclude the inquiry regarding penal consequences. Also relevant is the quantitative effect of the sentencing factor on a defendant's potential sentence. Of course, the Supreme Court has "rejected the claim that whenever a State links the 'severity of punishment' to 'the presence or absence of an identified fact' the State must prove that fact beyond a reasonable doubt." Id. at 84, 106 S.Ct. at 2415 (quoting Patterson v. New York, 432 U.S. 197, 214, 97 S. Ct. 2319, 2329, 53 L. Ed. 2d 281 (1977)). But the quantitative effect of a sentencing factor does remain relevant in deciding what procedural safeguards are appropriate. See Mullaney v. Wilbur, 421 U.S. 684, 698, 700, 95 S. Ct. 1881, 1889, 1890, 44 L. Ed. 2d 508 (1975) ("The fact remains that the consequences resulting from a verdict of murder, as compared with a verdict of manslaughter, differ significantly.... [R]espondent here faces a differential in sentencing ranging from a nominal fine to a mandatory life sentence."); Specht v. Patterson, 386 U.S. 605, 607, 610-11, 87 S. Ct. 1209, 1211, 1212-13, 18 L. Ed. 2d 326 (1967) (concluding that "petitioner, having been convicted ... under one Colorado statute that carries a maximum sentence of 10 years" may not be "sentenced under [another Colorado statute] for an indeterminate term of from one day to life without notice and full hearing"); see also McMillan, 477 U.S. at 87, 106 S. Ct. at 2416.

Thus, in the present case, it is relevant not only that the Sentencing Guidelines did not alter the statutory maximum of the penalty to which Restrepo was subject, but also that the sentencing factor here in question did not drastically affect the Guidelines range to which Restrepo was otherwise subject. Although the factor in question raised Restrepo's Guidelines range, the increase is not within the realm to which the Mullaney and Specht decisions allude. Cf. United States v. Kikumura, 918 F.2d 1084, 1101-02 (3d Cir.1990) (concluding that, at a minimum, findings leading to a "twelve-fold, 330-month departure from the median of an applicable guideline range" must be made by clear and convincing evidence); United States v. Townley, 929 F.2d 365, 369-70 (8th Cir.1991) (recognizing that elevated standard of proof might apply where sentencing factor results in "18-level increase in [the] base offense level and a seven-fold increase in the permissible sentencing range"); United States v. St. Julian, 922 F.2d 563, 569 n. 1 (10th Cir.1990) (directing trial courts to consider the appropriate standard of proof when "the difference between the guideline range and the departure sentence is great").1 This realm may not be so remote when a prosecutor takes inordinate advantage of the "dubious invitation" presented by Guidelines section 1B1.3 by " 'indict[ing] for less serious offenses which are easy to prove and then expand[ing the charges] in the probation office.' " United States v. Ebbole, 917 F.2d 1495, 1501 (7th Cir.1990) (quoting United States v. Miller, 910 F.2d 1321, 1332 (6th Cir.1990) (Merritt, C.J., dissenting), cert. denied, --- U.S. ----, 111 S. Ct. 980, 112 L. Ed. 2d 1065 (1991)); see also Harmelin v. Michigan, --- U.S. ----, 111 S. Ct. 2680, 2718 & n. 6, 115 L. Ed. 2d 836, 853 & n. 6 (1991) (White, J., dissenting); United States v. Payne, 940 F.2d 286, 293 (8th Cir.1991) (Heaney, J., concurring in part and dissenting in part).

PREGERSON, Circuit Judge, with whom HUG, Circuit Judge, joins, dissenting:

I concur with Circuit Judge Norris but write separately to state my belief that the reasonable doubt standard ought to obtain.

Restrepo was convicted of violating a penal statute that provided for a maximum prison sentence of twenty years. The government contended at oral argument that Restrepo had no constitutionally-protected liberty interest in any sentence that was less than that statutory maximum.

The government's position may have been correct under discretionary sentencing schemes that operate in many states and that operated in federal courts before the Sentencing Reform Act of 1984 and the promulgation of the Sentencing Guidelines. Under such systems of discretionary sentencing, like the one reviewed in McMillan v. Pennsylvania, 477 U.S. 79, 106 S. Ct. 2411, 91 L. Ed. 2d 67 (1986), the defendant's conviction, by itself, furnishes sufficient grounds for the judge to impose the maximum sentence permitted by statute, regardless of what facts may be established at the sentencing hearing. See id. at 92 n. 8, 106 S. Ct. at 2419, n. 8 (citing Meachum v. Fano, 427 U.S. 215, 224, 96 S. Ct. 2532, 2538, 49 L. Ed. 2d 451 (1976) (prison officials may transfer prisoner for any reason or for no reason at all)).

The Supreme Court has now rejected the government's position that a defendant has no liberty interest in a sentence less than the statutory maximum. Now that the Sentencing Guidelines are law, the defendant does have a constitutionally-protected liberty interest in a sentence less than the statutory maximum. This liberty interest, created by the mandatory language of the Sentencing Guidelines, is protected by the due process clause. See Burns v. United States, --- U.S. ----, ---, 111 S. Ct. 2182, 2186-88, 115 L. Ed. 2d 123 (1991); id. at ----, 111 S.Ct. at 2190-92, 2196-97 (Souter, J., dissenting).

In Restrepo's case, the Guidelines created a constitutionally-protected expectation that Restrepo would receive a sentence of 27-33 months for the 37.5 grams of cocaine involved in the two counts for which he was convicted. The government sought to establish that Restrepo was actually responsible for an additional sixty-five grams of cocaine. If the government was successful, the judge was required to sentence Restrepo to an additional fourteen to eighteen months in prison.

Fourteen to eighteen months of Restrepo's freedom thus turned on whether the government could establish Restrepo's responsibility for the additional cocaine. When the government's power to deprive an individual of liberty or property turns on the correctness of a factual conclusion, due process protects the individual's stake in assuring that the determination is reliable and accurate. When the stakes are high, as they are when individuals stand to lose their freedom, we require a high degree of confidence that the disputed issues are determined reliably:The function of a standard of proof, as that concept is embodied in the Due Process Clause and in the realm of factfinding, is to "instruct the factfinder concerning the degree of confidence our society thinks he should have in the correctness of factual conclusions for a particular type of adjudication." In re Winship, 397 U.S. 358, 370 [90 S. Ct. 1068, 1075, 25 L. Ed. 2d 368] (1970) (Harlan, J., concurring). The standard serves to allocate the risk of error between the litigants and to indicate the relative importance attached to the ultimate decision.

Addington v. Texas, 441 U.S. 418, 423, 99 S. Ct. 1804, 1808, 60 L. Ed. 2d 323 (1979). "In the administration of criminal justice, our society imposes almost the entire risk of error upon itself" by requiring proof beyond a reasonable doubt. Id. at 424-25, 99 S.Ct. at 1808. The standard of proof "reflects the value society places on individual liberty." Id. at 425, 99 S.Ct. at 1809.

When someone's freedom depends on the factual conclusion whether or not he or she is responsible for a criminal offense, we have historically required proof beyond a reasonable doubt. In this case, Restrepo's freedom turned on the factual conclusion that he was responsible for an additional sixty-five grams of cocaine. Even though that conclusion was made after conviction, at the sentencing hearing, it still resulted in a mandatory additional loss of fourteen to eighteen months. To underscore the value we place on individual liberty, and to ensure the reliability of factual conclusions that result in a loss of that liberty, I believe the reasonable doubt standard should apply.

I cannot believe, as the majority does, that the Constitution permits the defendant to be deprived of his freedom and " 'imprisoned for years on the strength of the same evidence as would suffice in a civil case.' " In re Winship, 397 U.S. 358, 363, 90 S. Ct. 1068, 1072, 25 L. Ed. 2d 368 (1970). I dissent.

WILLIAM A. NORRIS, Circuit Judge, with whom HUG, PREGERSON, and D.W. NELSON, Circuit Judges, join, dissenting:

Appellant was charged with and found guilty of two separate sales of cocaine. His codefendant was charged with and pled guilty to two other sales of cocaine. At appellant's sentencing, the district court made findings of fact that appellant had participated in the two sales of cocaine charged to his codefendant. Under the Sentencing Guidelines, see U.S.S.G. § 1B1.3(a)(2), these sentencing findings triggered an automatic increase in appellant's sentence from the Guidelines range of 27-33 months for the sales proved at trial, to a Guidelines range of 41-51 months for the sales proved at sentencing.

The Guidelines do not specify the standard of proof applicable to factfinding at sentencing. Rather, Congress and the Sentencing Commission have left that issue for the courts to resolve. In upholding appellant's sentence, the majority announces a broad rule that all factfinding in Guidelines sentencing hearings is governed by the preponderance of evidence standard. The majority then applies the rule to this case, notwithstanding that the facts proved at sentencing constituted elements of a separate crime for which appellant was neither charged nor convicted. In so doing, the majority decides that once a defendant is convicted for one crime, his sentence may be increased for other crimes regardless of whether they are proved beyond a reasonable doubt at trial or merely by a preponderance of the evidence at sentencing.

I dissent because the majority's interpretation of the Sentencing Reform Act as adopting the preponderance standard for proving separate crimes at sentencing raises two serious constitutional questions that can and should be avoided. A higher standard of proof would not only avoid the constitutional questions, but would better serve the purposes of the Sentencing Reform Act.

In allowing, for the first time, separate crimes to be used as sentencing factors with mandatory penal consequences, the Guidelines encounter the due process mandate of In re Winship, 397 U.S. 358, 90 S. Ct. 1068, 25 L. Ed. 2d 368 (1970), that elements of a crime must be proved beyond a reasonable doubt. In McMillan v. Pennsylvania, 477 U.S. 79, 106 S. Ct. 2411, 91 L. Ed. 2d 67 (1986), the Court applied Winship in the context of sentencing proceedings. In determining the constitutionally permissible standard of proof for sentencing factors that "provide sentencing courts with additional guidance" without altering the statutory maximum for the offense of conviction, id. at 87-88, 92, 106 S. Ct. at 2416-17, 2419, the Court distinguished between traditional sentencing factors that may be proved by a mere preponderance and sentencing factors that must be proved beyond a reasonable doubt because they constitute elements of a crime. Id. at 89, 106 S. Ct. at 2417. Although the Court recognized that legislatures might attempt to circumvent the Winship requirement by reclassifying elements of a crime as sentencing factors, it upheld the Pennsylvania legislature's classification of "visible firearm possession" as a sentencing factor because visible firearm possession had not "historically been treated 'in the Anglo-American tradition' as requiring proof beyond a reasonable doubt." See id. at 90, 106 S. Ct. at 2418 (quoting Patterson v. New York, 432 U.S. 197, 226, 97 S. Ct. 2319, 2335, 53 L. Ed. 2d 281 (1977) (Powell, J., dissenting)).

The Guidelines provide us with a sentencing scheme that seems to be unprecedented in that it not only treats elements of a crime as sentencing factors, but also attaches mandatory penal consequences to proof of those facts. Congress has explicitly made each act of selling cocaine a separate crime. See 21 U.S.C. § 841(a). Congress has also required through the Guidelines that a convicted defendant receive a longer sentence if it is proved at sentencing that he has engaged in additional drug sales. The majority interprets the statute as permitting these additional sales to be proved at sentencing by a mere preponderance and finds that interpretation to be constitutional without addressing the constitutional line drawn in McMillan between elements of a crime and traditional sentencing factors. By contrast, I find no reason to resolve this serious constitutional question when construing the statute as adopting a higher standard would not be "plainly contrary to the intent of Congress." See, e.g., Burns v. United States, --- U.S. ----, 111 S. Ct. 2182, 2187, 115 L. Ed. 2d 123 (1991).

The majority's interpretation of the Guidelines raises a second due process question. The Guidelines represent an historic break from the practice of endowing trial judges with the virtually unfettered discretion to sentence up to the statutory maximum. Under the Guidelines, judicial discretion is all but eliminated. The Guidelines require a sentence determined by the offense level corresponding to the offense of conviction. The sentence changes if, and only if, the sentencing judge makes specified findings of fact. Judicial factfinding is substituted for judicial discretion as the currency of the new sentencing regime.

In so structuring judicial decisionmaking, the Guidelines create a liberty interest in a sentence within the relevant Guidelines range. Because the factfinding process profoundly impacts this newly created liberty interest, the majority's interpretation of the Guidelines as permitting all sentencing facts to be proved by a mere preponderance raises a serious due process question under Mathews v. Eldridge, 424 U.S. 319, 96 S. Ct. 893, 47 L. Ed. 2d 18 (1976).

Principles of judicial restraint dictate that we avoid resolving constitutional questions if a permissible alternative interpretation of the statute that is not "plainly contrary to the intent of Congress" is available. See, e.g., Burns, 111 S. Ct. at 2187. It is not plainly contrary to the intent of Congress to construe the Sentencing Reform Act as requiring that elements of a separate crime, when used as sentencing factors, be proved by a standard of proof higher than a preponderance of the evidence. Not only would such an interpretation avoid major constitutional problems, it would also better serve the Act's basic goals of reducing sentence disparity and promoting fairness in sentencing.

* Under traditional discretionary sentencing regimes, there is no rule, general or specific, for the standard of proof governing factfinding. In these sentencing regimes, all sentencing decisions, including the weight to be given conduct that would normally constitute an element of a separate crime, are entrusted to the virtually unreviewable discretion of the sentencing judge, so long as the sentence does not exceed the statutory maximum for the offense of conviction.1 Due process does not impose a minimum standard of proof under such regimes because factfinding has no mandatory penal consequences.

In McMillan, the Court dealt, for the first time, with a due process challenge to a sentencing regime in which sentencing factfinding had mandatory penal consequences within the statutory range authorized by the offense of conviction.2 The Pennsylvania statute at issue in McMillan provided for a mandatory minimum sentence of five years for a convicted defendant who was found, by a preponderance of the evidence, to have visibly possessed a firearm during the commission of particular felonies. McMillan, 477 U.S. at 81, 106 S.Ct. at 2413. The Court recognized that the minimum sentence provision encountered the due process mandate of Winship, 397 U.S. at 364, 90 S.Ct. at 1072, which "explicitly" held that due process requires that each element of a crime be proved beyond a reasonable doubt. McMillan, 477 U.S. at 84-85, 106 S.Ct. at 2415. Accordingly, the Court in McMillan faced up to the question whether the Winship rule applied to a sentencing factor that had mandatory penal consequences.

The Court upheld the Pennsylvania statute against a challenge that due process required the gun possession to be proved by a standard higher than preponderance of the evidence. Id. at 91, 106 S.Ct. at 2418. In so doing, the Court drew a constitutional line between facts that may be proved by a mere preponderance at sentencing (sentencing factors) and facts that must be proved beyond reasonable doubt at trial (elements of a crime). In McMillan, the Court announced that it would generally defer to a legislature's classification of a particular fact as "an element of a crime" or "a sentencing factor." Id. at 86, 106 S.Ct. at 2416. However, relying on Patterson v. New York, 432 U.S. 197, 97 S. Ct. 2319, 53 L. Ed. 2d 281 (1977), the Court added that "there are constitutional limits to the State's powers in this regard; in certain limited circumstances Winship's reasonable-doubt requirement applies to facts not formally identified as elements of the offense charged." McMillan, 477 U.S. at 86, 106 S. Ct. at 2416 (emphasis added).

In upholding the Pennsylvania statute, the Court relied on the fact that sentencing factfinding under the Pennsylvania statute "neither alters the statutory maximum for the crime committed nor creates a separate offense calling for a separate penalty." McMillan, 477 U.S. at 87-88, 106 S. Ct. at 2417. Accord Schad v. Arizona, --- U.S. ----, 111 S. Ct. 2491, 2504 n. 9, 115 L. Ed. 2d 555 (1991) (referring to McMillan as "relying on the fact that under Pennsylvania law possession of a weapon 'neither alters the maximum penalty for the crime committed nor creates a separate offense calling for a separate penalty' ") (quoting McMillan, 477 U.S. at 87-88, 106 S. Ct. at 2417). The Court noted that the Pennsylvania statute merely "rais[es] the minimum sentence that may be imposed by the trial court." McMillan, 477 U.S. at 89, 106 S.Ct. at 2418.

The Court acknowledged, however, that, even if a sentencing factor neither alters the statutory maximum for the offense of conviction nor creates a separate offense calling for a separate penalty, the preponderance standard might not pass constitutional muster if the legislature has attempted to evade the commands of Winship by reclassifying an element of a crime as a sentencing factor. Id. at 89, 106 S. Ct. at 2417. In so doing, Court drew a constitutional line between traditional sentencing factors that may be proved by a mere preponderance and sentencing factors that must be proved by a higher standard because they have historically been considered elements of a crime. Applying the constitutional line to the facts of the case, the Court upheld the Pennsylvania statute, reasoning that the use of the preponderance standard was permissible because firearm possession was a traditional sentencing factor and had not "historically been treated 'in the Anglo-American legal tradition' as requiring proof beyond a reasonable doubt." McMillan, 477 U.S. at 90, 106 S.Ct. at 2418 (quoting Patterson, 432 U.S. at 226, 97 S.Ct. at 2335 (Powell, J., dissenting)). The Court in McMillan noted that this distinction between elements of a crime and traditional sentencing factors is built in part around the legislative classification of the facts to be proved and not simply around penal consequences of such proof. Obviously, every element of a crime must be proved beyond a reasonable doubt regardless of how minimal the potential sentence might be. Just as a traditional sentencing factor does not become an element of a crime simply because it affects the sentence significantly, an element of a crime does not become a traditional sentencing factor simply because it affects the sentence slightly.

To apply this McMillan distinction between an element of a crime and a traditional sentencing factor to the facts of this case, we must spell out the sentencing process under the Guidelines. Under the Guidelines, the defendant's conviction at trial determines his "base offense level" which mandates a particular sentence. However, the Guidelines also require a sentencing judge to go through a lengthy factfinding process. If a judge finds that particular facts exist, she is obliged to enhance the sentence automatically. She can subsequently decrease that sentence only if she finds further particular facts.

Appellant was convicted of engaging in two cocaine sales. In determining his sentence, the sentencing judge made the following findings of fact: (1) the two sales for which appellant was convicted involved 37.5 grams of cocaine; (2) appellant was responsible for two additional separate sales; (3) the quantity involved in those sales was 65.83 grams. Under section 1B1.3(a)(2) of the Guidelines, the sentencing judge was required to aggregate all the cocaine and to sentence appellant on the basis of 103.33 grams. Since the Guidelines determine a drug-related sentence on the basis of the quantity of drugs sold, these sentencing findings of fact enhanced appellant's sentence automatically. McMillan requires that we evaluate the legislative classification of each mandatory sentence enhancing factor individually.

With respect to the first fact, Congress has classified the quantity of drugs involved in the particular sales for which defendant has been convicted as a sentencing factor, and not as an element of a crime. See 21 U.S.C. § 841(b) (quantity of controlled substance figures under subsection entitled "penalties"). I believe that, if Congress had adopted the preponderance standard for the proof of such a fact, Congress' classification might be permissible under the McMillan distinction because the quantity of drugs involved in a particular sale for which defendant has been convicted would seem to be a traditional sentencing factor. As the Seventh Circuit stated in interpreting section 841(b), "Section 841(b) is a penalty enhancement provision.... Section 841(b) has nothing to do with the substantive elements of the underlying offense [b]ecause the quantity of the controlled substance is a sentencing issue unrelated to a defendant's underlying guilt." United States v. Acevedo, 891 F.2d 607, 611 (7th Cir.1989); see also United States v. Wood, 834 F.2d 1382, 1388 (8th Cir.1987) ("Both the plain language and the structure of [21 U.S.C. § 841(b) ] indicate that it is a sentencing provision."). Under our criminal law, "[t]he quantity of the controlled substance is not an essential element of the crimes proscribed under sections 841(a)(1) and 846; rather, it is a sentencing issue to be raised after proof of a defendant's underlying guilt." United States v. McNeese, 901 F.2d 585, 600-01 (7th Cir.1990); see also Acevedo, 891 F.2d at 611; United States v. Brown, 887 F.2d 537, 540 (5th Cir.1989). Thus, due process might be satisfied if the preponderance standard were applied to determine the quantity of drugs involved in the two sales for which appellant was convicted.

The constitutional calculus is different for the second sentencing fact--that appellant was responsible for the additional separate sales. Unlike the Pennsylvania legislature in McMillan, which had classified "visible gun possession" as a sentencing factor and had not treated it as a separate substantive offense, Congress has declared each and every drug sale to be a separate substantive criminal offense. See 21 U.S.C. § 841(a)(1). The act of selling illegal drugs is a crime while the quantity of drugs sold determines the penalty. Compare 21 U.S.C. § 841(a) (declaring the manufacturing, distributing, dispensing, etc. of a controlled substance to be "unlawful acts") with 21 U.S.C. § 841(b) (declaring that the quantity of controlled substances involved in the unlawful acts determines the "penalt[y]"). In interpreting 21 U.S.C. § 841(a), the Fifth Circuit declared that "[t]he statutory language clearly defines the unit of prosecution to be the act of delivering controlled substances into the hands of another.... There is no ambiguity." United States v. McDonald, 692 F.2d 376, 379 (5th Cir.1982) (declaring two acts of distribution to be two distinct criminal acts) cert. denied, 460 U.S. 1073, 103 S. Ct. 1531, 75 L. Ed. 2d 952 (1983); see also United States v. Thompson, 624 F.2d 740, 742 (5th Cir.1980).

Further confirmation of the distinct criminal status of individual drug sales is provided by the fact that the prosecutor was free to charge appellant with all four counts of the indictment--where each count referred to a separate sale. See United States v. Smith, 757 F.2d 1161, 1165 (11th Cir.1985) ("The language of [21 U.S.C. § 841(a)(1) ] plainly indicates that the government may prosecute a person for each separate act of distribution even if one distribution facilitates the next one."); United States v. Weatherd, 699 F.2d 959, 962 (8th Cir.1983); United States v. Noel, 490 F.2d 89, 90 (6th Cir.1974). Since Congress has declared these drug sales to be separate crimes, McMillan strongly suggests that a Congressional reclassification of this element of a crime as a sentencing factor through the adoption of the Sentencing Guidelines is an impermissible attempt to evade the Winship due process mandate. See McMillan, 477 U.S. at 89, 106 S.Ct. at 2417. Thus, in this case, the Constitution would probably not permit the sentencing judge to find by a mere preponderance that appellant was responsible for these two additional sales.

The present case is easy to resolve, in part because Congress has declared drug sales to be separate and distinct crimes. Even if Congress had classified "a separate drug sale" as a sentencing factor that may be proved by a preponderance, however, there would have been little doubt that Congress' classification fell on the impermissible side of the constitutional line drawn in McMillan. See 477 U.S. at 91, 106 S. Ct. at 2418. Unlike the gun possession factor in McMillan, an additional drug sale plainly constitutes a separate crime because it has "historically been treated 'in the Anglo-American legal tradition' as requiring proof beyond a reasonable doubt." Id. at 90, 106 S. Ct. at 2418 (quoting Patterson, 432 U.S. at 226, 97 S.Ct. at 2335 (Powell, J., dissenting)).

The constitutional calculus for the third sentencing fact--the quantity involved in the additional sales--is determined by the constitutional calculus for the sales themselves. Before a sentencing judge can take the amount of drugs involved in those uncharged sales into account, she must make a finding of fact that these additional sales actually took place. Since this fact unequivocally constitutes an element of a crime, a serious constitutional question is raised if it may be proved by a mere preponderance. Of course, once uncharged sales are proved by a constitutional standard, the quantity involved might be permissibly proved by a mere preponderance.

The majority attributes to a silent Congress the intent to take the internally contradictory stand that the finding of a drug sale is both an element of a crime that must be proved beyond a reasonable doubt and a sentencing fact that may be proved by mere preponderance. Unlike the Pennsylvania legislature in McMillan, neither Congress nor the Sentencing Commission has taken a position on the standard of proof governing the sentencing factfinding process. Why then does the majority impute such an irrational and arguably unconstitutional intent to Congress when we have a perfectly sensible way of reconciling Congress' commands? When we consider the constitutional line drawn in McMillan, we are left with only one rational reading of Congress' intention: Congress intended to allow the government to prove certain elements of additional crimes at sentencing, but it did not intend that the government carry a lower burden of proof at sentencing than at trial. The majority's failure to adopt this rational and permissible reading is inexplicable in light of the serious constitutional question that the alternative reading raises.

The majority's interpretation of the Guidelines as adopting the preponderance standard raises a second serious due process question. Not only do the Guidelines require, for the first time, that a sentencing judge enhance a sentence based on findings of criminal conduct at sentencing, but they constitute a radical departure from traditional discretionary sentencing practices in an additional sense. See Burns, 111 S. Ct. at 2184 ("The Sentencing Reform Act of 1984 revolutionized the manner in which district courts sentence persons convicted of federal crimes."). The Guidelines have replaced unfettered judicial discretion with an elaborate factfinding process that structures judicial decisionmaking. See Senate Committee Report on the Sentencing Reform Act, S.Rep. No. 225, 98th Cong., 1st Sess. 65 (1983), U.S.Code Cong. & Admin.News 1984, pp. 3182, 3248 (discussing how the Guidelines "structure judicial sentencing discretion"); see also United States v. Lira-Barraza, 941 F.2d 745, 748 (9th Cir.1991) (stating that the Guidelines imposed "a single statutory sentencing structure guiding the discretion of the court in all cases"); United States v. Allen, 873 F.2d 963, 966 (6th Cir.1989) ("[T]he Act and Guidelines substantially circumscribe the discretion which sentencing courts formerly exercised.") (citations omitted). Indeed, to a large extent, the Guidelines have eliminated judicial discretion.

Under our due process jurisprudence, this revolutionary change in judicial discretion has created, for the first time, a liberty interest in a sentence below the statutory maximum. Indeed, in replacing a judgment-driven system with a data-driven system, the Sentencing Reform Act has created a liberty interest in a sentence within the relevant Guidelines range, initially determined by the "base offense level" corresponding to the "offense of conviction." Because an important liberty interest is at stake, we are compelled to carry out independently the relevant due process inquiry mandated by Mathews v. Eldridge, 424 U.S. 319, 96 S. Ct. 893, 47 L. Ed. 2d 18 (1976). Under our current constitutional jurisprudence, this inquiry leads to the inescapable conclusion that the indiscriminate use of the preponderance standard within the context of Guidelines factfinding raises serious due process questions. A higher standard not only avoids the constitutional question but better serves the goals underlying the statute.A

Central to any liberty interest analysis is the extent to which governmental discretion to take away liberty has been cabined through legislation. See, e.g., Connecticut Board of Pardons v. Dumschat, 452 U.S. 458, 465, 101 S. Ct. 2460, 2465, 69 L. Ed. 2d 158 (1981) ("The ground for a constitutional claim, if any, must be found in statutes or other rules defining the obligations of the [relevant] authority...."); Bishop v. Wood, 426 U.S. 341, 344, 96 S. Ct. 2074, 2077, 48 L. Ed. 2d 684 (1976). In the parole context, for instance, a convicted defendant may have a liberty interest in an early release date if the discretion of the parole authority is restricted by statute. See Greenholtz v. Nebraska Penal Inmates, 442 U.S. 1, 12, 99 S. Ct. 2100, 2106, 60 L. Ed. 2d 668 (1979). In Board of Pardons v. Allen, 482 U.S. 369, 381, 107 S. Ct. 2415, 2422, 96 L. Ed. 2d 303 (1987), the Court applied Greenholtz in holding that a Montana parole release statute created a liberty interest. The Montana Parole Board had argued that because the criteria for denying release were very broad, e.g., "detriment to the prisoner or to the community," the Board retained so much discretion that no liberty interest was created despite statutory language that "the board shall release on parole ... any person...." 482 U.S. at 376, 107 S.Ct. at 2420. The Court, rejecting this argument, held that mandatory language created a liberty interest even when the regulatory scheme "cannot be applied mechanically," as long as it requires release when designated findings are made. Id. at 375, 377-78, 107 S. Ct. at 2419, 2420.

In a traditional discretionary sentencing regime, a convicted defendant has no liberty interest in a sentence below the statutory maximum because the trial judge's discretion to sentence the defendant up to the statutory maximum is "unfettered." Cf. Dumschat, 452 U.S. at 465, 101 S.Ct. at 2464. McMillan, once again, provides meaningful guidance on this issue. In holding that the Pennsylvania sentencing scheme did not create a protected liberty interest in a sentencing range below the statutory maximum, the Court reasoned that the sentencing finding of visible gun possession did not restrict the defendant's liberty more than the defendant's conviction alone did. See McMillan, 477 U.S. at 88, 106 S.Ct. at 2417 ("[The Pennsylvania statute] neither alters the maximum penalty for the crime committed nor creates a separate offense calling for a separate penalty."). Under the Pennsylvania statute, the sentencing judge retained discretion to sentence a convicted defendant up to the statutory maximum based solely on the conviction itself. Thus, under that statute, a defendant had no legitimate basis for expecting a sentence less than the statutory minimum whether or not he was found to have visibly possessed a gun. Nor did he have a legitimate expectation of receiving a sentence below the gun enhancement minimum of five years because the gun enhancement provision did not affect the sentencing judge's discretion to impose a sentence up to the statutory maximum of twenty years.

In McMillan, all that defendant lost was a hope of a sentence of less than five years. A "unilateral hope" based on mere chance is not enough to create a liberty interest. See Dumschat, 452 U.S. at 465, 101 S.Ct. at 2464 (rejecting the idea that "a unilateral hope" based on statistical probability amounts to a legitimate expectation). Absent a statute-based expectation in a lesser sentence, a convicted defendant had no liberty interest at stake in the sentencing factfinding process. In other words, a finding of gun possession, by whatever standard of proof, could not change the "due process calculus." McMillan, 477 U.S. at 92, 106 S. Ct. at 2419.

The Guidelines change the due process calculus dramatically because they take an unprecedented step in our criminal law by creating, for the first time, an expectation of a sentence within a certain range that will usually be below the statutory maximum. This expectation creates a liberty interest protected by due process because the sentencing judge no longer has the discretion to impose a higher sentence up to the statutory maximum "for whatever reason or for no reason at all." Meachum v. Fano, 427 U.S. 215, 228, 96 S. Ct. 2532, 2540, 49 L. Ed. 2d 451 (1976); see also Lira-Barraza, 941 F.2d at 748 ("The problem was disparity. Its cause was unlimited judicial discretion. The remedy was a single statutory sentencing structure guiding the discretion of the court in all cases to the end that similar sentences would be imposed on similar offenders for similar offenses."). On the contrary, the judge can enhance the sentence only if she makes certain findings of fact specified in the Guidelines.

The Guidelines operate literally like a flow chart: The defendant's conviction at trial determines his "base offense level." See U.S.S.G. § 1B1.2(a). The Guidelines then require that the sentencing judge go through a long checklist of specified factors. See generally U.S.S.G. §§ 1B1.2(b)-(c) and 1B1.1(c)-(g). Once the judge makes prescribed findings of fact with respect to any one of the factors, she is required to increase or decrease the offense level automatically and by a specified amount. The offense level resulting from these compelled enhancements determines the "guideline range." U.S.S.G. § 1B1.1. Judicial discretion kicks in only at this point, and even so, in a structured manner: the judge can exercise an option to "depart" from the Guidelines range only if she finds certain further aggravating or mitigating circumstances that bear some resemblance to the sentencing structure established by the Act. See 18 U.S.C. § 3553(b); U.S.S.G. § 5K1.1-5K2.15; Lira-Barraza, 941 F.2d at 746, 748-51. Since, under Guidelines sentencing, judicial discretion has all but disappeared, the sentence authorized by statute is no longer any sentence up to the statutory maximum, but the sentence that is within the range determined by the Guidelines.

Under the Guidelines, judicial discretion kicks in only after the Guidelines range is set and then only in a limited way. At that point, the judge has discretion to depart from that range only if certain conditions are satisfied. Recently, the Supreme Court recognized that Guidelines sentencing may create a liberty interest even within the context of such discretionary departure decisions. Burns, 111 S. Ct. at 2187. Burns involved the question whether Federal Rule of Criminal Procedure 32 requires notice that the sentencing judge is considering departing upward from the "applicable Guidelines sentencing range." Id. at 2184. In interpreting Rule 32 to require such notice, the Court stated, "In this case, were we to read Rule 32 to dispense with notice, we would then have to confront the serious question whether notice in this setting is mandated by the Due Process Clause." Id. Implicit in that concern is the Court's recognition that Guidelines sentencing, even at its most discretionary phase, may create a liberty interest.

In dissent, Justice Souter reached the liberty interest question because he disagreed with the Court's interpretation of Rule 32. In his analysis, Justice Souter found unequivocally that the Sentencing Reform Act creates "an expectation subject to due process protection that [a defendant] will receive a sentence within the presumptively applicable range in the absence of grounds ... justifying departure." Burns, 111 S. Ct. at 2192 (Souter, J., dissenting); see also United States v. Lawrence, 918 F.2d 68, 73-74 (8th Cir.1990) ("[The Guidelines] give convicted defendants a protected liberty interest to a sentence within the Guidelines range appropriate for their conduct and circumstances.") (Bright, J., dissenting), cert. denied, --- U.S. ----, 111 S. Ct. 1399, 113 L. Ed. 2d 455 (1991). In his analysis, Justice Souter relied upon Greenholtz, supra, and its progeny, which recognized that mandatory parole statutes, by giving convicted defendants an "expectancy of release," create a liberty interest subject to "some measure of constitutional protection." Greenholtz, 442 U.S. at 12, 99 S.Ct. at 2106. Justice Souter reasoned that the Sentencing Reform Act similarly created a liberty interest by using mandatory language that a sentencing judge "shall impose a sentence of the kind, and within the range [set forth in the Guidelines,] unless the court finds that there exists an aggravating or mitigating circumstance of a kind, or to a degree, not adequately taken into consideration by the Sentencing Commission." Burns, 111 S. Ct. at 2192 (Souter, J., dissenting) (quoting 18 U.S.C. § 3553(b)).3

The reasoning of eight Justices in Burns,4 which involved a discretionary departure from the Guidelines range, applies with even greater force to the factfinding process that leads to the setting of that range. Judicial discretion to depart from the Guidelines range is relatively unfettered. In setting the range, by contrast, judges have no discretion. Because the constitutional protection accorded to an expectation turns on the degree to which the decisionmaker's discretion is cabined, a defendant's liberty interest in a sentencing range determined by the base offense level is necessarily greater than his liberty interest in a sentence within the applicable Guidelines range. This is so, even though the statistical probability of receiving a sentence determined purely by the base offense level might be less than the probability of receiving a sentence within the Guidelines range. See Dumschat, 452 U.S. at 465, 101 S.Ct. at 2465 ("[S]tatistical probabilities standing alone generate no constitutional protections.... The ground for a constitutional claim ... must be found in statutes or other rules defining the obligations of the [decisionmaker].").

In sum, the Guidelines create a liberty interest in a sentence determined by the base offense level corresponding to the offense of conviction unless required findings of fact are made. Thus, in the present case, appellant has a liberty interest in a sentence initially based upon his present conviction--offense level 14.

Because the Guidelines create a liberty interest in a sentencing range determined by the base offense level corresponding to the offense of conviction, we must determine what process is due before a defendant may be deprived of this interest. More specifically, the question is whether the preponderance standard satisfies the Due Process Clause by providing adequate protection against a deprivation of liberty through inaccurate factfinding.

Neither Congress nor the Sentencing Commission has taken a position on the standard of proof governing the factfinding process. The texts of the Sentencing Reform Act and of the Guidelines are silent on the standard of proof issue. Why Congress failed to address such an important issue is unclear, but one plausible explanation is that standard of proof questions have "traditionally been left to the judiciary to resolve." Woodby v. INS, 385 U.S. 276, 284, 87 S. Ct. 483, 487, 17 L. Ed. 2d 362 (1966).

The legislative history is instructive, however, because it demonstrates that Congress intended the Guidelines to change sentencing practices radically. See Burns, 111 S. Ct. at 2184. The law creating the Sentencing Commission makes clear that the changes are fundamental: The purpose of the Commission is to "establish sentencing policies and practices for the Federal criminal justice system" that comport with Congressional goals. 28 U.S.C. § 991(b)(1). The Commission is not meant to tinker; it is to establish sentencing policies and practices.

The Commission has stressed that factfinding under the Guidelines will require more formality. After noting that "[i]n pre-guidelines practice, factors relevant to sentencing were often determined in an informal fashion," U.S.S.G. § 6A1.3, the Commentary to the Guidelines states, "This situation will no longer exist under sentencing guidelines." Id. The prime reason is that "[t]he court's resolution of disputed sentencing factors will usually have a measurable effect on the applicable punishment," and, as a result, "[m]ore formality is therefore unavoidable if the sentencing process is to be accurate and fair." Id. (emphasis added). Indeed, "disputes about sentencing factors must be resolved with care." Id.

The Supplementary Report on the Initial Sentencing Guidelines and Policy Statements (1987) [hereinafter Supplementary Report ] also points out that "[a]ccurate fact-finding is essential to ensure that a proper sentence is imposed. The guidelines will not achieve their intended effect if sentencing procedures are unreliable." Id. at 45. "Sentencing ... can require attention to many more discrete factual issues" than do jury determinations of guilt, and these factual issues "receive increased emphasis in a guideline system." Id. And further:

Existing law addressing dispute resolution in the sentencing context remains to be developed fully. Current sentencing practice often is informal. Particular facts seldom have a formal sentencing consequence under current law. Under the guidelines, however, the resolution of disputed sentencing factors often will have a definite and often quite substantial impact on the sentence. As a consequence, greater formality than currently exists can be expected in many cases.

Id. at 47.

Although the Commission has focused our attention on the important new role played by factfinding in the sentencing process, it has left the standard of proof issue to the courts to resolve:

With respect to sentencing issues that are genuinely disputed, the Commission chose simply to emphasize the importance of accuracy and fairness. Especially in light of questions that have been raised regarding the Commission's power to prescribe enforceable rules for dispute resolution, most of the procedural details are left for resolution by the sentencing court in light of the nature and importance of the particular issue and the context in which it arises.

....

Among the legal issues that may have to be resolved are: ... What is the weight of the burden of persuasion (i.e., is it sufficient to prove the asserted factor by a preponderance of the evidence or is a higher degree of certainty required)?

Id. at 46, 47 n. 79 (emphasis added).

In the absence of clear guidance from either the Congress or the Sentencing Commission, we must determine the appropriate standard of proof with a paramount principle of judicial restraint in mind: we should favor a standard of proof that does not raise serious constitutional questions if one that is not plainly contrary to the intent of Congress is available. See, e.g., Burns, 111 S. Ct. at 2187 (quoting Edward J. DeBartolo Corp. v. Florida Gulf Coast Building & Construction Trades Council, 485 U.S. 568, 575, 108 S. Ct. 1392, 1397, 99 L. Ed. 2d 645 (1988) ("[W]here an otherwise acceptable construction of a statute would raise serious constitutional problems, the Court will construe the statute to avoid such problems unless such construction is plainly contrary to the intent of Congress.")).

It is settled law that "[t]he safeguards of due process are not rendered unavailable simply because a determination may already have been reached that would stigmatize the defendant and that might lead to a significant impairment of personal liberty." Mullaney v. Wilbur, 421 U.S. 684, 698, 95 S. Ct. 1881, 1889, 44 L. Ed. 2d 508 (1975). Having found that the Guidelines create a liberty interest in the "base offense level" corresponding to the "offense of conviction" and that Congress has specified no standard of proof, I turn to Mathews v. Eldridge, 424 U.S. 319, 96 S. Ct. 893, 47 L. Ed. 2d 18 (1976), to analyze whether the preponderance standard satisfies due process within this context.5 Mathews teaches that when a liberty interest is implicated, we must balance the individual's liberty interest against the interests of the government. In the Guidelines context, the individual's interest in avoiding excessive punishment must be balanced against the government's interest in imposing an appropriate sentence. See Mullaney, 421 U.S. at 699, 95 S.Ct. at 1890 (mandating "an analysis that looks ... to the interests of both the State and the defendant as affected by the allocation of the burden of proof"); cf. Addington v. Texas, 441 U.S. 418, 425, 99 S. Ct. 1804, 1809, 60 L. Ed. 2d 323 (1979) ("[W]e must assess both the extent of the individual's interest in not being involuntarily confined indefinitely and the state's interest in committing the emotionally disturbed under a particular standard of proof.").

The importance of the liberty interest at stake need not be labored. As Justice Souter put it in Burns, "The defendant's interest in receiving a sentence not unlawfully higher than the upper limit of the guideline range is ... clearly substantial." 111 S. Ct. at 2193 (Souter, J., dissenting). See also Allen, 482 U.S. at 373 n. 3, 107 S. Ct. at 2418 n. 3 ("At stake in the parole-release decision is a return to freedom, albeit conditional freedom; liberty from bodily restraint is at the heart of the liberty protected by the Due Process Clause."); Morrissey v. Brewer, 408 U.S. 471, 482, 92 S. Ct. 2593, 2601, 33 L. Ed. 2d 484 (1972) ("[T]he liberty of a parolee, although indeterminate, includes many of the core values of unqualified liberty and its termination inflicts a 'grievous loss' on the parolee and often on others."). Under the Guidelines, then, a defendant's stake in accurate factfinding at sentencing may be as great as his stake in accurate factfinding at trial.

Traditionally, a burden of proof greater than preponderance of the evidence has been required when "[t]he interests at stake in th[e] cases are deemed to be more substantial than mere loss of money." Addington, 441 U.S. at 424, 99 S.Ct. at 1808. Surely a defendant faced with a sentence longer than the sentence mandated for crimes proved at trial has a liberty interest at stake at least as great as that of an alien facing deportation, Woodby v. INS, 385 U.S. 276, 87 S. Ct. 483, 17 L. Ed. 2d 362 (1966), a citizen facing denaturalization, Chaunt v. United States, 364 U.S. 350, 353, 81 S. Ct. 147, 149, 5 L. Ed. 2d 120 (1960), a parent facing termination of parental rights, Santosky v. Kramer, 455 U.S. 745, 102 S. Ct. 1388, 71 L. Ed. 2d 599 (1982), a defendant in a civil action facing allegations of fraud, Addington, 441 U.S. at 424, 99 S.Ct. at 1808, or a mentally ill person facing civil commitment, id. at 427, 99 S.Ct. at 1810.

The government's interest in the preponderance standard of proof, on the other hand, is more attenuated. It goes without saying that the government has a compelling interest in protecting the community from crime. But the government also has an interest in returning convicted defendants to their families and their jobs. In other words, the government has an interest in avoiding sentences that are excessive as well as too lenient. In the last analysis, the government's interest is in a just sentence. Indeed, the spirit that drives Guidelines sentencing is fairness--fairness that is achieved by reducing sentencing disparity. See U.S.S.G. Ch. 1, Pt. A, intro. comment.

It is not clear how the preponderance standard, in allocating the risk of error nearly equally between the individual and the government, serves the government's interest in fairness and accuracy. Indeed, a higher standard better serves that interest because it assures "a higher degree of certainty" in the factfinding process that determines a defendant's applicable Guidelines range. Supplementary Report, supra, at 47 n. 79.

When the preponderance standard is coupled with the imbalance in litigation resources between the government and the individual, the result is the creation of "a significant prospect" of error. See Santosky, 455 U.S. at 762-64, 102 S.Ct. at 1399-1400 (disparity in litigation resources renders preponderance standard constitutionally inadequate in proceedings terminating parental rights). The imbalance in resources at sentencing hearings is striking. First, the government's ability to make its case vastly exceeds a defendant's ability to rebut it. The financial resources of the prosecution and the ability of prosecutors to access police, informants, and experts usually are much greater than those of defendants' counsel. Pope, How Unreliable Factfinding Can Undermine Sentencing Guidelines, 95 Yale L.J. 1258, 1268-70 (1986). Second, presentence reports often rely excessively on the prosecutor's files for their information, id. at 1277, and hearsay evidence in presentence reports or from confidential informants is difficult and costly to rebut. See United States v. Fatico, 458 F. Supp. 388, 398-99 (E.D.N.Y.1978), aff'd, 603 F.2d 1053 (2d Cir.1979), cert. denied, 444 U.S. 1073, 100 S. Ct. 1018, 62 L. Ed. 2d 755 (1980). The absence of equal litigation resources thus makes the preponderance of evidence standard generate even more errors--errors that fall primarily on the side of deprivations of liberty for individuals.

In other circumstances, the Supreme Court has held that a preponderance of evidence standard violated due process. Although these cases are not controlling authority on the due process question raised by proving sentence enhancing factors by a preponderance of evidence, they do illustrate the seriousness of that question.

In Addington v. Texas, the Court held that the state's legitimate parens patriae interest in providing care to citizens unable because of emotional disorders to take care of themselves did not outweigh the individual's interest in avoiding an erroneous involuntary commitment to a mental institution. 441 U.S. at 426-27, 99 S.Ct. at 1809-10. The Court pointed out that "[i]ncreasing the burden of proof is one way to impress the factfinder with the importance of the decision and thereby perhaps to reduce the changes that inappropriate commitments will be ordered." Id. at 427, 99 S. Ct. at 1810. The Court concluded that "[t]he individual should not be asked to share equally with society the risk of error when the possible injury to the individual is significantly greater than any possible harm to the state." Id.

Likewise, in Santosky v. Kramer, the Supreme Court held a parent's interest in avoiding erroneous termination of parental rights outweighed the state's interest in promoting the welfare of the child. The Court held that "[f]or the child, the likely consequence of an erroneous failure to terminate is preservation of an uneasy status quo," while for the parents, the consequence of an erroneous termination is "the unnecessary destruction of their natural family." 455 U.S. at 765-66, 102 S.Ct. at 1401. In those circumstances, "[a] standard that allocates the risk of error nearly equally between those two outcomes does not reflect properly their relative severity." Id. at 766, 102 S.Ct. at 1401. See also Woodby, 385 U.S. at 286, 87 S.Ct. at 488 (in a statutory interpretation case, federal government's interest in proving the deportability of resident aliens by a preponderance of the evidence held to be outweighed by the hardships of an erroneous finding of deportability).

In balancing the interests of the government against the interests of the individual, we are required by Mathews to consider the extent to which a higher standard of proof would burden the sentencing process. In attempting to justify the preponderance standard, some of our sister circuits have fretted about turning the sentencing hearing into a "second trial." See, e.g., United States v. Guerra, 888 F.2d 247, 251 (2d Cir.1989), cert. denied, --- U.S. ----, 110 S. Ct. 1833, 108 L. Ed. 2d 961 (1990). This concern is, however, more imaginary than real. The reality is that Congress, in discarding a system of discretionary, individualized sentencing in favor of a system that works by feeding facts into a grid, has made it inevitable that sentencing hearings would be endowed with the attributes of a trial regardless of which standard of proof applies. In Guidelines sentencing hearings, prosecutors and defense counsel are already pitted against each other in a genuinely adversarial factfinding process. The judge, instead of crafting an individualized sentence based on information from a variety of sources, now spends her time adjudicating issues of fact. See U.S.S.G. § 6A1.3; see also Burns, 111 S. Ct. at 2185 ("As amended by the Sentencing Reform Act, Federal Rule of Criminal Procedure 32 provides for focused, adversarial development of the factual and legal issues relevant to determining the appropriate Guidelines sentence."). When Congress passed the Sentencing Reform Act, it was fully aware that the new system it was creating would require formal factfinding. For the first time in the history of the federal courts, Congress gave defendants the right to appeal findings of fact that bear on their sentences. 18 U.S.C. § 3742(a)(2). The Commission also acknowledged that "[a]ccurate factfinding is essential to ensure that a proper sentence is imposed. The Guidelines will not achieve their intended effect if sentencing procedures are unreliable." Supplementary Report, supra, at 45. According to the Commission, "More formality is ... unavoidable if the sentencing process is to be accurate and fair." U.S.S.G. § 6A1.3, comment. Since Congress has made the policy choice in favor of a sentencing regime that is necessarily adversarial, it is not for the courts to sacrifice constitutional values for the sake of imagined cost-savings of the preponderance standard of proof. See Santosky, 455 U.S. at 767, 102 S.Ct. at 1402 ("[A] stricter standard of proof would reduce factual error without imposing substantial fiscal burdens upon the State").

In sum, the preponderance of evidence standard, when applied, in Guidelines sentencing, to factfinding that may enhance a sentence above the base offense level corresponding to the offense of conviction, raises a substantial due process question under the Mathews balancing test. Given that the government's interest in the mere preponderance standard of proof is so tenuous and the defendant's liberty interest is so real, we run a serious risk of violating due process if we require the individual "to share equally with society the risk of error." Addington, 441 U.S. at 427, 99 S. Ct. at 1810. We should, therefore, avoid the constitutional question and reject the majority's broad and indiscriminate rule that all factfinding at sentencing is governed by the preponderance standard. See, e.g., Burns, 111 S. Ct. at 2187.

A standard of proof higher than mere preponderance not only avoids the constitutional question, but is also not "plainly contrary to the intent of Congress." Edward J. DeBartolo Corp., 485 U.S. at 575, 108 S.Ct. at 1397. Indeed, a higher standard serves better the basic purpose of the Guidelines: fairness by reducing disparity in sentencing. See, e.g., Lira-Barraza, 941 F.2d at 748; United States v. Williams, 891 F.2d 962, 964 (1st Cir.1989) ("[T]o implement the guidelines properly, courts must bear [the statutory] goals in mind. Among other things, courts must remember the importance which Congress, and the Commission, attached to ensuring that like situations are treated alike.").6

A major, if not the major, goal of the Guidelines is the reduction of disparity in sentencing, disparity that is inherent in traditional discretionary sentencing systems. See 28 U.S.C. § 991(b)(1)(B) ("The purposes of the United States Sentencing Commission are to-- ... provide certainty and fairness in meeting the purposes of sentencing, avoiding unwarranted sentencing disparities among defendants with similar records who have been found guilty of similar criminal conduct."); U.S.S.G. Ch. 1, Pt. A, intro. comment. In the context of sentencing hearings under the Guidelines, the preponderance standard fails to serve the Guidelines' purpose of reducing sentencing disparity because it creates a new source of sentencing disparity--inaccurate factfinding.

A preponderance burden of proof, which allocates the risk of error nearly evenly, see Addington, 441 U.S. at 423, 99 S.Ct. at 1807, allows a fact to be considered true if the factfinder is convinced that the fact is more probably true than not, or to put it differently, if the factfinder decides there is a 50%-plus chance that it is true. The standard thus creates a much greater risk of error, far more error than a standard that would assure a "higher degree of certainty" in sentencing factfinding. Supplementary Report, supra, at 47 n. 79. The greater risk of error created by a preponderance standard disserves the Guidelines' goal of sentencing uniformity. No matter how accurate and statistically significant the Guidelines distinctions and sentencing ranges may be, they will not serve Congress' purpose unless defendants are assigned to the proper Guidelines range by accurate factfinding. As one commentator has said, "[I]n the war against [sentencing] disparity, the tacticians of the guidelines movement have paid insufficient attention to the procedures that develop the facts to which guidelines are applied.... Ironically, sentencing guidelines may entrench a different kind of disparity--factual disparity." Pope, supra, at 1260.

Judge Bright's dissent in Lawrence, 918 F.2d at 72-74 (Bright, J., dissenting), illustrates the cursory factfinding that the preponderance standard allows.7 He based his dissent on the ground that the sentencing factfinding was "plainly erroneous and constitutionally intolerable." Id. at 72. The Presentence Report description of the offense conduct "consisted of a random listing of drug sales in no apparent chronological or otherwise discernable order ... [and the Presentence Report] frequently made no attempt to determine when or to whom the drugs were distributed ... or whether the conspiracy provided the drugs or aided their distribution.... As a result, the Presentence Report may have double-counted drug quantities or included drug quantities that came from conspiracies not charged by the indictment." Id. at 73. Judge Bright noted:

Lawrence's plight is by no means unusual in these days of 'sentencing by the numbers.' To the contrary, the Sentencing Guidelines, despite their aura of objectivity, produce results that vary with the attitudes and judgments of human beings who interpret (and often misinterpret) them. Here, the district court accorded great weight to the numerous judgment calls in the probation officer's report. I see these administrative decisions as error.

Id. at 74.

Because the Guidelines are designed to reduce sentencing disparity, we should not attribute to Congress or the Commission the intent to employ a standard of proof that introduces a significant risk of unfair "factual disparity" into the sentencing process. A higher standard, by reducing the risk of sentencing disparity caused by factfinding errors, would better serve the goals underlying the statute.

In Burns the Court interpreted Rule 32 to require notice of a judge's intent to depart because it was not plainly contrary to the intent of Congress, served the purposes of the Guidelines and avoided a constitutional question. See Burns, 111 S. Ct. at 2187. For the same three reasons we should interpret the Guidelines to require a higher standard of proof for Guidelines sentencing factfinding. Because using the preponderance standard both disserves the Guidelines' purpose of reducing sentencing disparity and raises serious constitutional questions, we should not attribute to Congress the intent to use that standard in the absence of clear congressional direction that we should do so. See id. at 2187; NLRB v. Catholic Bishop of Chicago, 440 U.S. 490, 501, 99 S. Ct. 1313, 1319, 59 L. Ed. 2d 533 (1979); United States v. Won Cho, 730 F.2d 1260, 1273 n. 18 (9th Cir.1984) (en banc).

In this case, however, we need not determine the standard of proof applicable to all sentence enhancing factors. We deal only with an enhancement based on sentencing findings that appellant engaged in two additional sales of cocaine which were not charged in the indictment. These additional sales constitute elements of a separate crime, not traditional sentencing factors. Regardless of which standard of proof we decide in future cases to apply to other sentencing factors, we should interpret the Guidelines as requiring a higher standard to prove elements of a separate crime as "relevant conduct" under section 1B1.3(a)(2).

Interpreting the Guidelines as requiring a higher standard of proof is consistent with the McMillan distinction between "elements of a crime" and "traditional sentencing factors." Consequently, even if in future cases we decide that the preponderance standard satisfies due process when applied to Guidelines factfinding for traditional sentencing factors, we would have a principled constitutional basis for distinguishing elements of additional crimes from these other sentencing factors. After all, our holding may be limited to elements of separate crimes because the due process inquiry is a particularized one. As Justice Souter stated in Burns, "Due Process, unlike some legal rules, is not a technical conception with a fixed content unrelated to time, place and circumstances, ... but is flexible, calling for such procedural protections as the particular situation demands." Burns, 111 S. Ct. at 2192 (Souter, J., dissenting) (quoting Cafeteria Workers v. McElroy, 367 U.S. 886, 895, 81 S. Ct. 1743, 1748, 6 L. Ed. 2d 1230 (1961) and Morrissey, 408 U.S. at 481, 92 S.Ct. at 2600). The Sentencing Commission, too, envisaged a process of particularized analyses when it left such important procedural issues for the courts to resolve "in light of the nature and importance of the particular issue and the context in which it arises." Supplementary Report, supra, at 47 n. 79 (emphasis added).

In sum, regardless of the result we reach in future cases that deal with traditional sentencing factors under the Guidelines, there can be no doubt of the result in this case. Construing the Guidelines as allowing uncharged crimes to be proved by a mere preponderance raises a serious constitutional question and risks placing the Guidelines on the impermissible side of the constitutional line drawn in McMillan. An alternative construction of the statute as mandating a higher standard of proof would avoid this serious constitutional question.

In proposing that we reject the preponderance standard, I recognize I am asking our circuit to break ranks with virtually all of our sister circuits. With all due respect to them, I believe their opinions are generally infected with the same faulty analysis that infects the majority's opinion in this case. Circuit has followed circuit without, in some cases, even a vestige of independent analysis. See, e.g., United States v. Silverman, 889 F.2d 1531, 1535 (6th Cir.1989); United States v. Fonner, 920 F.2d 1330, 1333 (7th Cir.1990); United States v. Burke, 888 F.2d 862, 869 (D.C.Cir.1989).

There are signs, however, that the dam is finally cracking, signs of an emerging awareness of the truly profound implications of the changes the Guidelines have wrought.8 I had hoped that our court in this case would have chosen to lead rather than follow.