Vincent Scott was convicted of robbery under 22 D.C. Code § 2901 (1967) and sentenced to prison for five to fifteen years. The proceedings preceding his conviction were, we conclude, free from error.1 The events surround*266ing his sentencing, however, present thorny questions concerning what factors the trial judge may properly consider at that stage. We affirm the conviction, but remand for a resentencing in accordance with the principles announced in this opinion.
I
In reviewing the appellant’s claims concerning the events at his sentencing hearing, we do not find it necessary to combat the massed precedent forbidding appellate modification of sentences.2 The objections raised center not upon the “duration or severity of this sentence,”3 but upon the reasons for which it was imposed. The trial judge has wide freedom in the information that may be considered in imposing a sentence.4 The result, the specific sentence selected, may be beyond the ken of the appellate court. But the appellate court must scrutinize the sentencing process to insure that the trial judge has considered the information available with some regard for its reliability,5 and has evaluated the information in light of the factors relevant to sentencing. This Court has remanded cases for resentencing where the trial judge failed to avail himself of the proper aids for sentencing, such as mental examinations or presentenee reports.6 We have also refused to accept a trial judge’s determination of sentence where the record demonstrated that an improper factor had been considered, such as a statutory sentencing provision since repealed.7
In many cases, of course, the appellate court does not know whether the sen-*267fencing judge has performed his task thoroughly or well. The frequent blankness of the record has led to suggestions that the trial judge should be required to set forth his reasoning in announcing the sentence decided upon.8 But that problem is not before us. Here the trial judge explained in some detail the reasons for which the sentence was imposed. He stated repeatedly throughout the hearing that he did not believe the exculpatory testimony the appellant had given at trial. And at one point the judge indicated that he was influenced as well by the fact that the appellant had insisted upon a trial in the first place:
Now the Court didn’t believe your story on the stand, the Court believes you deliberately lied in this case. If you had pleaded guilty to this offense, I might have been more lenient with you.
A few moments later, the sentencing hearing flew off on a revealing tangent when the trial judge read a letter submitted by the-appellant. The letter, from his attorney, referred to a visit the lawyer had made to the judge’s law clerk. In it/ he reported that in the clerk’s opinion “there was only one way to get a light sentence from Judge- and that was to confess that you did the robbery, to apologize four or five times and to say that you were willing to turn over a new leaf.” The trial judge then called his clerk to the witness stand and interrogated him concerning his conversation with the attorney. The clerk affirmed that the letter fairly reflected the substance of his comments to the lawyer. He stated, “It has always been my opinion that you view sentencing differently when someone admits guilt rather than maintaining innocence.” He added, however, “This has nothing to do with private conversations we have had in chambers. It is from things I have heard while sitting in that seat during sentencing hearings.”
The judge himself then commented upon his reactions to defendants found guilty by the jury who continued to assert their innocence at allocution. He went on, “I hope sometime I hear some defendant say, ‘Judge, I am sorry, I am sorry for what I did.’ That is what I have in mind.”
II
The appellant argues that he was denied due process of law by the trial judge’s consideration of the “inflammatory” letter and by the pressure placed upon him to confess his guilt after trial. The federal courts have examined this issue but seldom, perhaps because trial judges rarely announce a reliance upon this consideration. The Tenth Circuit has upheld a sentence of 18 months that the trial judge imposed instead of probation because the appellant continued to assert his innocence after trial.9 It did so, however, not because the court approved of the practice, but because “the matter is one entirely for the trial court.”10 To the extent that the court concluded that it lacked as an appellate court not only the power to modify a sentence but also the authority to review the sentencing process, we disagree.
The Fifth Circuit has recently reached precisely the opposite conclusion. The appellant in Thomas v. United States11 rejected the trial court’s invitation to “come clean and make a clean breast of this thing for once and for all” at his sentencing hearing. The trial judge then imposed the maximum sentence permitted by statute. Judge Rives reasoned for the court,
It must be remembered that, at the time of his allocution, Thomas had not been finally and irrevocably adjudged guilty. Still open to him were the *268processes of motion for new trial * * *, appeal, petition for certiorari, and collateral attack. * * * If he chose [to confess guilt] * * *, he would elect to forego all of the abovenoted post-conviction remedies [and] to confess to the crime of perjury * * *. Moreover, he would abandon the right guaranteed by the Fifth Amendment to choose not to be a witness against himself, not only as to the crime of which he had been convicted, but also as to the crime of perjury.12
The opinion concluded,
When [the appellant] received harsher punishment than the court would have decreed had he waived his Fifth Amendment rights, he paid a judicially imposed penalty for exercising his constitutionally guaranteed rights.13
This analysis equates a confession of guilt at allocution with a waiver of the self-incrimination privilege, and thereby imports the same strict standard of voluntariness that Judge Rives would also require for a plea of guilty.14 In struggling with the problems associated with plea bargaining, the Fifth Circuit sitting en banc has settled upon a less stringent standard of voluntariness which rejects such reliance upon the self-incrimination privilege.15 But whatever the reasons why a guilty plea may differ from a confession for self-incrimination purposes, the Fifth Amendment standards demanded by the Thomas case seem clearly required when a confession at allocution is involved, at least when the defendant has testified at trial and denied the crime. For, as Judge Rives points out, not only would a confession of guilt at the sentencing hearing be, in literal terms, just that — a confession — but also such a statement would constitute an admission of perjury.
It might be argued that the dilemma between waiving Fifth Amendment rights and receiving a harsher sentence might be elminated by a rule making a defendant’s statement at his allocution inadmissible in other proceedings.16 But the appellant had no such assurance in this case.17 The trial judge was eager for an acknowledgment of guilt presumably because a confession might indicate repentance. But the appellant could reasonably have believed that he could show penitence — real or affected —only at the price of prejudicing his appeal, if not worse. Whether the trial judge intended or not “to bargain thus with the defendant,” we conclude that “the court was without right * * * to put a price on an appeal.”18
III
Since a resentencing will be necessary, two other considerations the trial judge relied upon in sentencing the appellant require comment. The first problem arises from the judge’s repeated statements that he believed the appellant had committed perjury on the witness stand in denying that he had participated in the robbery. There are two arguments why this belief would properly influence the choice of a sentence: (1) that additional punishment should be imposed for the independent substantive offense of perjury; (2) that the commission of perjury reflected adversely upon the appellant’s prospects for re*269habilitation, and therefore justified a lengthier sentence for the crime of robbery.19
The first argument deserves emphatic rejection. The Government could if it wished prosecute the appellant for perjury.20 In such a proceeding, the appellant would have all the protections of a criminal trial. If the trial judge in fact imposed additional punishment upon the appellant for the supposed commission of perjury, he plainly denied the appellant the trial upon that offense to which Scott was entitled.
As for the second argument, the peculiar pressures placed upon a defendant threatened with jail and the stigma of conviction make his willingness to deny the crime an unpromising test of his prospects for rehabilitation if guilty. It is indeed unlikely that many men who commit serious offenses would balk on principle from lying in their own defense. The guilty man may quite sincerely repent his crime but yet, driven by the urge to remain free, may protest his innocence in a court of law. This realization, indeed, unquestionably accounts for the extreme infrequency with which convicted criminals are in fact prosecuted for perjury committed at their trials.
The Government argues that the appellant “has no constitutional right to lie.” The beguiling simplicity of this proposition misphrases the question. Of course a defendant has no constitutional right to lie, however much we may sympathize with his too human temptation. But the defendant does have a right to testify in his own defense. In doing so, he risks the jury’s disbelief. If he in fact fails to convince the jurors, conviction and punishment will follow. If the Government for whatever reason concludes that prosecution for perjury is appropriate, he risks punishment for that as well. To allow the trial judge to impose still further punishment because he too disbelieves the defendant would needlessly discourage the accused from testifying in his own behalf.
IV
The trial judge also stated at the sentencing hearing, “If you had pleaded guilty to this offense, I might have been more lenient with you.” The stark import of this comment is that the defendant paid a price for demanding a trial. In view of the prohibitions the Supreme Court has laid down against making the exercise of Fourth 21 Fifth,22 and Sixth23 Amendment rights costly, the pricetag thus placed on the right to a fair trial which these amendments guarantee would, on first impression, seem clearly impermissible.
And yet, despite the startling incongruity, empirical evidence supports the proposition that judges do sentence defendants who have demanded a trial more severely.24 At least one Court of Appeals has taken approving “judicial notice of the fact that trial courts quite generally impose a lighter sentence on pleas of guilty than in cases where the accused pleaded not guilty but has been found guilty by a jury.” 25 An advisory com*270mittee of the American Bar Association has concluded that “it is proper for the court to grant charge and sentence concessions to defendants who enter a plea of guilty * * * when the interest of the public in the effective adminstration of criminal justice would thereby be served.” 26
Much of this adulation for differential sentencing has been rationalized without frank recognition of the fact that whatever its advantages, the practice does exact a price from those who insist upon a trial. But the arguments in favor of differential sentencing cannot be dismissed by a wooden insistence that the exercise of constitutional rights can never be made costly. Some rights may be so vital that no deterrence to their free exercise can be tolerated. The Supreme Court has accorded such preeminent status to the self-incrimination privilege.27 But in other areas, the Court has suggested the need for a less truncated analysis. In United States v. Jackson,28 the Court held that Congress could not provide for a death penalty “applicable only to those defendants who assert the right to contest their guilt before a jury.” 29 But in doing so the majority did not rely upon the summary argument that the exercise of such a right could in no way be made costly. The Court rather asked “whether that effect is unnecessary and therefore excessive.” 30
In such cases as Sherbert v. Verner31 the Court has also suggested that some constitutional interests are not inviolate. After concluding there that a disqualification for unemployment benefits burdened the exercise of her religion by a Seventh Day Adventist who refused to work on Saturdays, the Court phrased the relevant test as “whether some compelling state interest * * * justifies the substantial infringement of appellant’s First Amendment right.” 32
The Supreme Court has offered little guidance concerning which constitutional rights can tolerate some chilling effect and which cannot. Perhaps the right to a trial, like the self-incrimination privilege but apparently unlike the right to a .jury, belongs in the latter camp. But until the Supreme Court speaks, the practice of differential sentencing should be evaluated with some attention paid to the nature of the price exacted from those .who plead innocent and why it is exacted.
Two arguments inevitably appear whenever differential sentencing is discussed. The first is that the defendant’s choice of plea shows whether he recognizes and repents his crime. One difficulty with this argument is that no court or commentator has explained why a defendant’s insistence upon his self-incrimination privilege is not also evidence of a lack of repentance. Or his insistance that evidence unconstitutionally seized should not be admitted.
Repentance has a role in penology. But the premise of our criminal jurisprudence has always been that the time for repentance comes after trial. The adversary process is a fact-finding engine, not a drama of contrition in which a prejudged defendant is expected to knit up his lacerated bonds to society.
There is a tension between the right of the accused to assert his innocence and the interest of society in his repentance. But we could consider resolving this conflict in favor of the latter interest only if the trial offered an unparalleled opportunity to test the repentance of the ac*271cused. It does not. There is other, and better, evidence of such repentance. The sort of information collected in presen-tence reports provides a far more finely brushed portrait of the man than do a few hours or days at trial. And the offender while on probation or in prison after trial can demonstrate his insight into his problems far better than at trial.
If the defendant were unaware that a proper display of remorse might affect his sentence, his willingness to admit the crime might offer the sentencing judge some guidance. But with the inducement of a lighter sentence dangled before him, the sincerity of any cries of mea culpa becomes questionable. Moreover, the refusal of a defendant to plead guilty is not necessarily indicative of a lack of repentance. A man may regret his crime but wish desperately to avoid the stigma of a criminal conviction.33
The Supreme Court was careful to point out in Sherbert v. Verner that “no showing merely of a rational relationship to some colorable state interest would suffice” to justify an infringement of First Amendment rights.34 Even if we assume that the right to a fair trial may in some circumstances be made costly, the required justification here also must be a paramount goal achievable in no other way. The supposed value of a guilty plea in demonstrating repentance does not meet this test.
The second argument for differential sentencing is necessity. Most convictions, perhaps as many as 90 per cent in some jurisdictions, are the product of guilty pleas.35 Unless a large proportion of defendants plead guilty, the argument runs, the already crowded dockets in many jurisdictions would collapse into chaos. Since most defendants are indigent, the only price they can be forced to pay for pleading innocent is time in jail. Ergo, differential sentences are justified for those who plead guilty and those who plead innocent.
When approached from this perspective, the problem inevitably becomes entwined with that of plea bargaining. And the difficulties that practice presents are exceeded only by its pervasive-, ness. In many areas such bargaining dominates the criminal process. Its format may vary. The prosecutor may agree to reduce the charge in exchange for a guilty plea, or he may agree to recommend a lighter sentence. The judge may be aware of the agreement or he may not. If aware that a bargain has been struck, the court may or may not ratify the agreement before a plea is offered and accepted.
When a defendant pleads guilty in exchange for the promise of the prosecutor or court, a subsequent challenge to the voluntariness of his plea raises a recognized constitutional issue. When the accused refuses to plead guilty and subsequently receives a heavier sentence, the invisibility with which the system operates in individual cases too often conceals the constitutional issue. But the problem is the same in both contexts. Whether the defendant surrenders his to a trial because of a bargain with aourt or prosecutor, or exercises his right at the cost of a stiffer sentence, a price has been put on the right.
The two sides of this coin are related in a practical sense as well. At least when only a single charge is involved, the effectiveness of plea bargaining depends upon the willingness of the court *272to impose a lighter sentence when a defendant pleads guilty. If such is the custom within a jurisdiction, the prosecutor enjoys credibility. Indeed, if the custom is sufficiently well known, actual bargaining may be unnecessary: enough defendants will be cowed into guilty pleas simply by the force of their lawyers’ warnings that defendants convicted after demanding a trial receive long sentences.36
Thus, to the extent that the appellant here received a longer sentence because he pleaded innocent, he was a pawn sacrificed to induce other defendants to plead guilty. Since this is so, to consider the price he paid for the exercise of his right without regard for the process of which it is but one instance would be to ignore reality.
The debate surrounding differential sentencing and plea bargaining involves two issues that should be separated. The first concerns what inducements for guilty pleas should be permitted; the second involves the question of who should be allowed to offer these inducements to the accused. Both questions must be resolved to articulate a comprehensive standard of voluntariness. Unfortunately, the difficulty of the first question has tended to lure courts toward a one-sided focus upon the second.
The saga can best begin with the vicissitudes of J. Paul Shelton in the Fifth Circuit Court of Appeals and the Supreme Court. The maneuverings of that crafty litigator pro se have created “much of the current uncertainty about the legal status of plea discussions and plea agreements,” 37 as the Fifth Circuit has itself acknowledged.38 Shelton pleaded guilty to transporting a stolen vehicle. He later moved to vacate the conviction on the ground that his plea was involuntary, having been induced by a promise of the prosecutor to recommend the one-year sentence he in fact received. The District Court denied the motion. The Court of Appeals reversed, holding that a plea induced by a promise of leniency was involuntary as a matter of law, whether or not the promise was kept.39
The opinion by Judge Rives for the majority equated a plea of guilty with a confession in open court, and concluded that the same strict test of voluntariness should be applied to guilty pleas as to confessions. Judge Tuttle dissented, arguing that the majority position “would considerably impede the administration of justice” by rendering prosecutors unable to offer sufficient “inducements for any person to plead guilty.” 40
When the case was reheard en banc, the dissenting opinion of Judge Tuttle became the majority position.41 The test of voluntariness that he had formulated for guilty pleas,42 and that the full court adopted,43 was essentially one requiring an informed choice, although perhaps one between unattractive alternatives:
[A] plea of guilty entered by one fully aware of the direct consequences, including the actual value of any commitments made to him by the court, prosecutor, or his own counsel, must stand unless induced by threats (or promises to discontinue improper harassment), misrepresentation (including unfulfilled or unfulfillable promises), or perhaps by promises that are by their *273nature improper as having no proper relationship to the prosectuor’s business (e. g. bribes).
This test would tolerate some “commitments” by the prosecutor to induce a guilty plea, and apparently some by the court itself as well.
The Solicitor General confessed error when the Shelton case reached the Supreme Court on the ground that “the plea of guilty may have been improperly obtained,” and the Court reversed the en banc decision of the¡ Court of Appeals in a memorandum opinion.44 The decision did not specify why or in what way the guilty plea might have been “improperly obtained,” and therefore did not necessarily overrule the reasoning of the en banc decision below. Such, at least, has been the interpretation of the Fifth Circuit, which has continued to apply essentially Judge Tuttle’s test of voluntariness.45
In the Shelton opinions the question of what bargains shall be permitted was squarely confronted, although inconclusively resolved. Recently the attention of the Fifth Circuit has moved to the second question involved in a determination of whether a plea is voluntary, that of who may offer the inducements or “commitments” permitted by the informed-choice model of voluntariness. Relying heavily upon the distinction drawn between the role of the trial judge and that of the prosecutor by the American Bar Association Project on Minimum Standards for Criminal Justice,46 the court concluded in Brown v. Beto.47 that (1) “Properly safeguarded plea discussions and plea agreements between an accused and a prosecutor are consistent with the fair administration of justice;”48 but (2) “‘The trial judge should not participate in plea discussions’ ” 49
If inducements are to be offered for guilty pleas, there are strong reasons why the court should not be the party to offer them. The trial judge may sacrifice his ability to preside impartially at trial by becoming too involved with pre-trial negotiations. Even if he does not, it may so appear to the defendant. It is important not only that a trial be fair in fact, but also that the defendant believe that justice has been done. The accused may fairly doubt this if he thinks the judge begrudges him the exercise of his right to trial. Moreover, the defendant’s uncertainty concerning the expectations or wishes of the judge will prevent his exercise of the best judgment in deciding upon a plea.
Judge Weinfeld has concluded in two careful opinions that whatever the propriety of plea bargaining between prosecutors and defendants, the peculiarly sensitive position of the trial judge renders involuntary any guilty plea induced by a commitment from the bench.50 His vivid portrayal of the “unequal positions of the judge and the accused, one with the power to commit to prison and the other deeply concerned to avoid prison”51 presents a compelling brief for demanding *274that the judge not become a participant in the bargaining process.
In this case the trial judge did not bargain with the defendant. Indeed he did not even point out that he might be more lenient with a defendant who pleaded guilty until after trial. But in so stating at the sentencing hearing he announced to all future defendants the guidelines in his court room. We cannot approve of these guidelines for the same reasons that we could not condone actual plea bargaining by a trial judge. The policy announced by the trial judge may not endanger his actual impartiality at trials as much as his participation in plea bargaining sessions might. And we certainly do not criticize the impartiality displayed by the experienced trial judge in this case. But we cannot ignore the impact of such a policy on the appearance of justice to criminal defendants and their ability to choose wisely between a plea of guilty and an exercise of their right to trial.
In recognizing that an announced policy of differential sentencing presents some of the same dangers as conventional plea bargaining, we do not, as the concurring opinion suggests, equate the two practices. Overt plea bargaining by a trial judge places direct and immediate pressure upon the defendant to forego his right to a trial. An announced policy of differential sentencing which distinguishes between defendants who demand a trial and those who do not presents an inducement to plead guilty which may be less coercive in the individual case, but which nevertheless must affect the decision to exercise the constitutional right to a trial.
V
In announcing the rule that trial judge should neither participate directly in plea bargaining nor create incentives for guilty pleas by a policy of differential sentences, we must at the same time point out that the trial judge cannot ignore the plea bargaining process. A guilty plea must be not only voluntary, but also knowing and understanding.52 If the defendant has decided to admit his guilt because of a commitment from the prosecutor, it is essential for the validity of his plea that he have a full and intelligent understanding of nature and extent of that commit-ment. To fulfill the requirement of Rule 1l53 and insure that a plea is made only “after proper advice and with full understanding of the consequences,”54 the trial judge must make certain that the defendant has made a knowing appraisal of the alternatives open to him.55
There is no justification for a courtroom charade in which the judge asks whether a plea has been induced by any promises, and the defendant replies that it has not, when all the actors realize that quite the contrary is true.56 Rule *27511 requires a full scrutiny of the circumstances surrounding a plea, and promises from the prosecutor do count. Moreover, if it is indeed proper for the prosecutor to make commitments to the defendant in exchange for a plea of guilty, there seems no reason why the judge should not be allowed to permit the defendant to know the value of the prosecutor’s promises before entering his plea. The Second Circuit has recognized the force of this argument. In United States ex rel. Rosa v. Follette,57 the trial judge had forthrightly told the defendants that the sentence recommended by the prosecutor would be accepted by the court if they pleaded guilty. The court noted that since "it is clear that the plea discussions were between the prosecutor and [the defendants] * * * [,] in no meaningful sense can [the judge] be said to have participated in the negotiations.” 58 Refusing to conclude that the mere “tangential” participation of the judge represented by his ratification of the agreement tainted the plea, the appellate court concluded, “In the instant case, [the defendants were] fortunate in being given the security of the Judge’s beneficence by learning immediately what most defendants are tortured over, can only hope for and anticipate — that the trial judge will follow the prosecutor’s recommendation.” 59 The Sixth Circuit has relied upon essentially the same reason in declining to set aside a sentence in Waddy v. Heer,60 distinguishing Judge Weinfeld’s decisions in Tateo and, Gilligan as cases where the judge took part in actual negotiations rather than merely ratifying an agreement between the prosecutor and defendant.
The line may well be a fine one be-" tween a trial judge “participating” in the plea bargaining process and a judge merely “ratifying” an agreement already reached between the accused and the prosecutor. But the likelihood that the judge will overawe the defendant or surrender his impartiality is at least sharply reduced if he does not play a role in the negotiations leading to the formation of a bargain.
The fact that the trial judge must be aware of any bargain made before accepting a plea of guilty, and perhaps may ratify the agreement in appropriate circumstances, imposes an obligation upon the judge to supervise the fairness of the bargain. Since the trial judge determines the sentence a convicted defendant will receive, he cannot escape this responsibilty if plea bargaining takes the form of a promised sentence recommendation. But the functional problem is the same if plea bargaining takes the form of a prosecutorial offer to reduce the charge, as seems to *276be the practice in the District of Columbia.61 By deciding what charge or charges to proceed with, the prosecutor can often effectively control the sentence the defendant is likely to receive. To the extent this decision is made independently by the prosecutor before the defendant is called upon to plead, there is no danger that the defendant will be deterred from exercising his right to trial. When, however, the prosecutor makes his decision by offering to reduce the charge in exchange for a plea of guilty, the situation is quite different. If the trial judge is to so supervise the bargain, the law must resolve the troubling issue left in limbo by the Shelton opinions: what inducements to plead guilty are permissible? Since this case does, not present that issue, it would be inappropriate to essay a comprehensive answer to the question. Because the problem is. so closely entwined with the policy of differential sentencing, however, several brief comments are in order.
First, the prosecutor clearly cannot have carte blanche to apply whatever tactics he wishes to induce a guilty plea. A policy of deliberately overcharging defendants with no intention of prosecuting on all counts simply in order to have chips at the bargaining table would, for example, constitute improper harassment of the defendant.
Second, there may be circumstances under which the prosecutor may bargain. with the defendant without raising the constitutional question of whether the exercise of the right to trial can be made costly. When there is substantial uncertainty concerning the likely outcome of a trial, “each side is interested in limiting these inherent litigation risks.” 62 The prosecutor may be willing to accept a plea of guilty on a lesser charge rather than chance an acquittal on the more serious. The accused may be similarly willing to acknowledge his guilt of the lesser charge rather than risk conviction on the more serious, or to accept the promise of a lighter sentence to escape the possibility of conviction after trial and a heavier sentence.
Superficially it may seem that even in such a case the defendant who insists upon a trial is found guilty pays a price for the exercise of his right when he receives a longer sentence than his less venturesome counterpart who pleads guilty. In a sense he has. But the critical disjunction is that the price he has paid is not one imposed by the state to discourage others from a similar exercise of their rights, but rather one encountered by those who' gamble and lose. After the fact, the defendant who pleads innocent and is convicted receives a heavier sentence. But, by the same token, the defendant whov pleads innocent and is acquitted receives no sentence. To the extent that the bargain struck reflects only the uncertainty of conviction before trial, the “expected sentence before trial” — length of sentence discounted by probability of conviction — is the same for those who decide to plead guilty and those who hope for acquittal but risk conviction by going to trial.
In determining who has or has not “paid a price,” it is essential to reason clearly concerning what class of defendants are being compared with what other class of defendants for what purpose and at what point in time. The danger presented by plea bargaining is that defendants deciding upon a plea will be deterred *277from exercising their right to a trial. The relevant vantage point is thus before trial, and the relevant comparison is between the expectations of those who decide to insist upon a trial and those who decide to eliminate the risk of trial by pleading guilty. If the sentence expectations of those two classes at that time are the same, then there will be no chilling effect upon exercise of the right to trial, and it is accurate to say that no “price” has been placed upon exercise of the right.
To determine the expectations of those defendants who insist upon a trial, we must consider the probability of conviction as well as the sentences received by those who plead innocent and are later convicted. The argument that the defendant who receives a heavier sentence after trial has “paid a price” because he receives a heavier sentence than the' defendant who is acquitted (and goes free) or pleads guilty (and receives a shorter sentence on the same or a reduced charge) errs on two counts: (1) the comparison is made at the wrong time — after trial, when the uncertainty of litigation has passed, rather than before trial — and (2) the comparison is made between the wrong categories of defendants — the class of defendants convicted after trial versus the class of defendants who plead guilty or are acquitted rather than the class of defendants who exercise their right to trial versus the class of defendants who do not.
The situation is quite different when the prosecutor engages in bargaining not because he is willing to take a sure half loaf rather than to await the outcome of a trial, but because his limited resources convince him he must deter defendants from demanding a trial.63 The divide between the two situations may be difficult to locate for even the best-intentioned prosecutor, and even more difficult for a trial judge to review. But the standards which guide prosecutors in the exercise of their discretion are as much a part of the law as the rules applied in court. Indeed, the impact of such standards is more decisive for many defendants than that of any other legal rules. If “it is procedure that spells much of the difference between rule by law and rule by whim or caprice,” 64 the same or more can be said of well-articulated, evenly-applied standards for prose-cutorial discretion. If we must as a practical matter rely upon prosecutors to apply these standards in good faith, the responsibility for articulating appropriate standards must belong to the courts, not prosecutors. The nature of the process may well prevent judges from actually reviewing specific instances of prosecutorial discretion in all but the most exceptional cases. But this reality imposes a special duty upon courts to provide what guidance they can for prosecutors entrusted with such discretion. In the area of plea bargaining, the lodestar must be the realization that our law solemnly promises each man accused his day in court. If a prosecutor enters plea and charge negotiations not with the purpose of adjusting the charge to reflect the uncertainties of litigation but with the goal of deterring the defendant from *278the exercise of his right to a trial, the chasm between promise and reality is no narrower because the trial court affects a righteous air of non-involvement. Perhaps the promise must be tempered if society is unwilling to pay its price. But that decision should be made in sunlight, and not in the shrouded mist of unguided prosecutorial discretion.
The arguments that the criminal process would collapse unless substantial inducements are offered to elicit guilty pleas have tended to rely upon assumption rather than empirical evidence. In many jurisdictions lacking sophisticated resources for criminal investigations, a large proportion of suspects apprehended are caught virtually red-handed.65 The argument “But what if everyone did not plead guilty?” has force only to the extent that a sizable proportion of defendants have some motivation to plead innocent. If the defendant does have some hope of acquittal, the right to a trial assumes overarching importance. If he does not, there is some presumption that most men will not indulge in a meaningless act.66 Moreover, the plea bargaining system itself may actually operate in some instances to burden the docket rather than to lighten it.67
If in fact more defendants wish to demand trials than our present system can process, there are two solutions: deter exercise of the right to trial, or commit more resources to the criminal process. This case does not require us to decide whether inadequate resources may constitutionally justify the former course. But we do think it important to phrase the question in the straightforward fashion of whether the cost of honoring a constitutional right can justify making the exercise of that right costly to the individual. The temptation is strong in the area of plea bargaining to assume that defendants convicted after trial receive a “normal” sentence while those who plead guilty and save the Government the cost of a trial receive special “leniency” in exchange. If this analysis were valid, some defendants would win and none would lose. But in reality there are winners and losers. The “normal” sentence is the average sentence for all defendants, those who plead guilty and those who plead innocent.68 If we are “lenient” toward the former, we are by precisely the same token “more severe” toward the latter.69
*279Conviction affirmed; remanded for re-sentencing.
. The appellant claims that the continuation of his trial after his co-defendant changed his plea to guilty at the completion of the Government’s case denied him due process of law. In the course of questioning the co-defendant to determine the voluntariness of his changed plea, the trial judge elicited a statement that implicated the appellant. At that time Scott’s attorney, moved for a mistrial on the ground that the trial judge, having heard the statement, would be prejudiced against the appellant. The motion was denied. On appeal Scott raises a somewhat different claim: that the jury must have realized that the co-defendant had changed his plea, and must have been improperly influenced by this realization in passing upon the appellant’s guilt or innocence.
We can agree with neither step in this reasoning. There is no necessary reason why the jurors should have concluded that because the co-defendant was absent on the second day, he must have changed his plea to guilty. The jury was, in the first place, instructed that they “must not conjecture, speculate, guess or surmise why the other defendant is not before you at this time.” Even if we assume that the jurors might have so speculated despite this admonition, there were other possible explanations for his disappearance: he might have jumped bond; he might have been granted a mistrial; the cases might have been severed.
If the jury did somehow conclude that the co-defendant had changed his plea, moreover, there is no reason why the appellant should have been prejudiced thereby. The theory of his defense was not that there had been no robbery, nor that he and his co-defendant were not present at the scene. His claim rather was that his co-defendant suggested and executed the crime, while he declined the invitation to participate and resolutely looked the other way. Since the co-defendant was by the appellant’s own argument guilty, the jury could hardly have been influenced if they did in fact conclude that the former had changed his plea after the first day of trial. In a similar factual situation, a District Court has recently found no constitutional in*266firmity in a case where, unlike the appellant’s, the change of plea by a co-defendant occurred before the jury. See United States ex rel. Iiuntt v. Russell, 285 F. Supp. 765 (E.D.Pa., May 23, 1968).
. For more than 70 years federal appellate courts have declared themselves impotent to modify sentences imposed within statutory limits. See, e.g., Gore v. United States, 357 U.S. 386, 393, 78 S. Ct. 1280, 2 L.Ed.2d 1405 (1958), Blockburger v. United States, 284 U.S. 299, 305, 52 S.Ct. 180, 76 L.Ed. 306 (1932), Leach v. United States, 118 U.S.App. D.C. 197, 203, 334 F.2d 945, 951 (1964). The Supreme Court has encountered the rule but seldom, and on those rare occasions has accepted it with only the most summary cement. For the most part, the rule has been announced and applied by the lower appellate courts. Occasional judges have expressed bewilderment concerning the origin of the rule, and have noted the strange contrast it presents to the role of appellate courts in reviewing abuses of discretion in other areas. See, e.g., United States v. Martell, 335 F.2d 764, 766-768 (4th Cir. 1966) (per Sobeloff, C. J.); United States v. Rosenberg, 195 F.2d 583, 604-605 (2d Cir. 1952) (per Frank, C. J.). It has been suggested, in those opinions and elsewhere, that the power given appellate courts to “affirm, modify, vacate, set aside or reverse any judgment” by 28 U.S.C. § 2106 (1964) would on its face extend to modification of sentences. And in unusual circumstances this Court has invoked that power to impose a sentence of life imprisonment in lieu of capital punishment. See Coleman v. United States, 123 U.S.App.D.C. 103, 357 F.2d 563 (1965) (en banc); Frady v. United States, 121 U.S.App.D.C. 78, 348 F.2d 84, cert. denied, 382 U.S. 909, 86 S.Ct. 247, 15 L.Ed.2d 160 (1965). But in those cases the scope of the power assumed was carefully limited to the factors there present.
. Townsend v. Burke, 334 U.S. 736, 741, 68 S.Ct. 1252, 1255, 92 L.Ed. 1690 (1948).
. See Williams v. People of State of New York, 337 U.S. 241, 69 S.Ct. 1079, 93 L. Ed. 1337 (1949); cf. Sprecht v. Patterson, 386 U.S. 605, 606-608, 87 S.Ct. 1209, 18 L.Ed.2d 326 (1967).
. See Townsend v. Burke, 334 U.S. 736, 68 S.Ct. 1252, 92 L.Ed. 1690 (1948).
. See Leach v. United States, 118 U.S. App.D.C. 197, 334 F.2d 945 (1964); Jones v. United States, 117 U.S.App.D. C. 169, 327 F.2d 867 (1963) (en banc); Peters v. United States, 113 U.S.App.D. C. 236, 307 F.2d 193.
. See Coleman v. United States, 123 U.S. App.D.C. 103, 357 F.2d 563 (1965) (en banc); cf. Verdugo v. United States, 402 F.2d 599 (9th Cir., May 16, 1968).
. See, e.g., Symposium, Appellate Review of Sentencing, 32 F.R.D. 257, 263, 274-275 (1962).
. Williams v. United States, 273 F.2d 469 (lOth Cir. 1959).
. Id. at 470.
. 368 F.2d 941 (5th Cir. 1966).
. Id. at 945.
. Id. at 946.
. See Shelton v. United States, 242 F. 2d 101 (5th Cir.), rev’d on rehearing, 246 F.2d 571 (1957) (en banc), rev’d 356 U.S. 26, 78 S.Ct. 563, 2 L.Ed.2d 579 (1958) (per curiam), discussed infra.
. See Shelton v. United States, 246 F. 2d 571 (5th Cir. 1957) (en banc), rev’d 356 U.S. 26, 78 S.Ct. 563, 2 L.Ed.2d 579 (1958) (per curiam), discussed infra.
. Cf. Simmons v. United States, 390 U. S. 377, 389-394, 88 S.Ct. 967, 19 L.Ed. 2d 1247 (1968).
. Cf. Gardner v. Broderick, 392 U.S. 273, 279, 88 S.Ct. 1913, 20 L.Ed.2d 1082 (1968).
. Worcester v. C. I. R., 370 F.2d 713, 718 (1st Cir. 1966) (per Aldrich, C. J.).
. See Note, The Influence of the Defendant’s Plea on Judicial Determination of Sentence, 66 Yale L.J. 204, 211-217 (1956).
. United States v. Williams, 341 U.S. 58, 62, 71 S.Ct. 595, 95 L.Ed. 747 (1951).
. See Simmons v. United States, 390 U. S. 377, 389-394, 88 S.Ct. 967, 19 L.Ed. 2d 1247 (1968).
. See, e.g., Gardner v. Broderick, 392 U. S. 273, 276, 88 S.Ct. 1913, 20 L.Ed.2d 1082 (1968); Spevack v. Klein, 385 U.S. 511, 515, 87 S.Ct. 625, 17 L.Ed.2d 574 (1967).
. See United States v. Jackson, 390 U.S. 570, 581-585, 88 S.Ct. 1209, 20 L.Ed.2d 138 (1968).
. See, e.g., Note, The Influence of the Defendant’s Plea or Judicial Determination of Sentence, 66 Yale L.J. 204, 206-209 (1956).
. Dewey v. United States, 268 F.2d 124, 128 (8th Cir. 1959).
. American Bar Association Project on Minimum Standards for Criminal Justice, Standards Relating to Pleas of Guilty § 1.8(a) (1967) [hereinafter Pleas of Guilty],
. See, e.g., Gardner v. Broderick, 392 U.S. 273, 276, 88 S.Ct. 1913, 20 L.Ed.2d 1082 (1968); Spevack v. Klein, 385 U.S. 511, 515, 87 S.Ct. 625, 17 L.Ed.2d 574 (1967).
. 390 U.S. 570, 88 S.Ct. 1209, 20 E.Ed. 138 (1968).
. Id. at 581, 88 S.Ct. at 1216.
. Id. at 582, 88 S.Ct. at 1216.
. 374 U.S. 398, 83 S.Ct. 1790, 10 L.Ed. 2d 965 (1963).
. Id. at 406, 83 S.Ct. at 1795.
. In fact, a colorable argument can be made that a glib willingness to admit guilt in order to “secure something in return” may indicate quite the opposite of repentance, and that a reluctance to admit guilt may in fact reflect repentance. See Alschuler, The Prosecutor’s Role in Plea Bargaining, 36 U.Chi.L. Rev. 50, 57 n. 24 (1968).
. 374 U.S. at 406, 83 S.Ct. at 1795.
. See The President’s Comm’n on Law Enforcement and Administration of Justice, Task Force Report: The Courts 9 (1967) [hereinafter Task Force Report] ; D. J. Newman, Conviction: The Determination of Guilt or Innocence Without Trial 3 n. 1 (1966).
. See Pilot Institute on Sentencing, 26 F.R.D. 231, 289 (1959).
. Pleas of Guilty 64.
. Brown v. Beto, 377 F.2d 950, 953 (5th Cir. 1967).
. Shelton v. United States, 242 F.2d 101, 112-113 (5th Cir.), rev’d on rehearing, 246 F.2d 571 (1957) (en banc), rev’d 356 U.S. 26, 78 S.Ct. 563, 2 L.Ed.2d 579 (1958) (per curiam).
. Id. at 115 (dissenting opinion).
. Shelton v. United States, 246 F.2d 571 (5th Cir. 1957), rev’d 356 U.S. 26, 78 S. Ct. 563, 2 L.Ed.2d 579 (1958) (per curiam).
. 242 F.2d at 115 (dissenting opinion).
. 246 F.2d at 572 n. 2.
. Shelton v. United States, 356 U.S. 26, 78 S.Ct 563, 2 L.Ed.2d 579 (1958) (per curiam).
. See, e.g., Cooper v. Holman, 356 F.2d 82 (5th Cir.), cert. denied, 385 U.S. 855, 87 S.Ct. 103, 17 L.Ed.2d 83 (1966); Busby v. Holman, 356 F.2d 75 (5th Cir. 1966); Sorrenti v. United States, 306 F.2d 236 (5th Cir. 1962), cert. denied, 373 U.S. 916, 83 S.Ct. 1306, 10 L.Ed.2d 416 (1963); Martin v. United States, 256 F.2d 345 (5th Cir.) (per Tuttle, C. J.), cert. denied, 358 U.S. 921, 79 S.Ct. 294, 3 L.Ed.2d 240 (1958).
. Pleas of Guilty 60-77.
. 377 F.2d 950 (5th Cir. 1967).
. Id. at 956.
. Id. at 957 (quoting Pleas of Guilty § 3.3(a)).
. United States ex rel. Elksnis v. Gilligan, 256 F.Supp. 244 (S.D.N.Y.1966); United States v. Tateo, 214 F.Supp. 560 (S.D.N.Y.1963); see also Euziere v. United States, 249 F.2d 293 (10th Cir. 1957).
. Elksnis, 256 F.Supp. at 254.
. See id. at 255 and n. 48.
. Rule 11 of the Federal Rules of Criminal Procedure, as amended in 1966, provides in relevant part:
The court * * * shall not accept * * * [a plea of guilty] without first addressing the defendant personally and determining that the plea is made voluntarily with understanding of the nature of the charge and the consequences of the plea.
. Kercheval v. United States, 274 U.S. 220, 223, 47 S.Ct. 582, 583, 71 L.Ed. 1009 (1927).
. See Bailey v. MacDougall, 392 F.2d 155 (4th Cir. 1968); United States ex rel. McGrath v. LaVallee, 348 F.2d 373 (2d Cir. 1965), cert. denied, 383 U.S. 952, 86 S.Ct. 1214, 16 L.Ed.2d 214 (1966).
. The somewhat bemused comments of one sociologist observing the legal process indulging its penchant for reality-denying ritual are worth repeating:
Apparently, however, there is some tension between the face of bargaining and moral expectations concerning justice. * * * A clear example of this tension is provided by an excerpt from a trial and Newman’s comments on it. The following questions were asked of a defendant after he had pleaded guilty to unarmed robbery when the original charge was armed robbery. This re*275duction is common, and the judge was fully aware that the plea was negotiated :
Judge: You want to plead guilty to robbery unarmed?
Defendant: Yes, sir.
Judge: Your plea of guilty is free and voluntary?
Defendant: Yes, sir.
Judge: No one has promised you anything?
Defendant: No.
Judge: No one has induced you to plead guilty?
Defendant: No.
Judge: You’re pleading guilty because you are guilty?
Defendant: Yes.
Judge: I’ll accept your plea of guilty to robbery unarmed and refer it to the probation department for a report and for sentencing Dec. 28.
[D. J. Newman, Conviction: The Determination of Guilt or Innocence Without Trial 83 (1966)].
* * * Newman’s comment on this exchange has an Alice-In-Wonderland quality:
This is a routine procedure designed to satisfy the statutory requirement and is not intended to disguise the process of charge reduction.
Scheff, Negotiating Reality: Notes on Power in the Assignment of Responsibility, 16 Social Problems 1, 5 (1968).
. 395 F.2d 721 (2d Cir. 1968).
. Id. at 725.
. Id. at 726.
. 383 F.2d 789, 793 (6th Cir. 1967). See also Task Force Report 12; Pleas of Guilty 71-77.
. For a partial account of the process in the District of Columbia which nevertheless suggests the complexity of the problem, see H. I. Subin, Criminal Justice in a Metropolitan Court 15-17, 34-35, 37-38, 42-50 (1966).
. See Task Force Report 10. One commentator well-versed in the realities of plea-bargaining has labeled the problem as one of “variable guilt”: “Justice does not flow readily from a computer where the survivor of a fight and witnesses at the taproom relate stories which indicate, in variable quantities, the facts of intoxication, provocation, malice aforethought and self-defense.” Specter, Book Review, 76 Yale L.J. 604, 606 (1967). See also Alsechuler, The Prosecutor’s Role in Plea Bargaining, 36 U.Chi.L.Rev. 50, 81 & n. 71 (1968); Weinberg & Babcock, Book Review, 76 Tale L.J. 612, 620-621 (1967).
. We do not, of course, suggest that a prosecutor can or should ignore resource limitations in deciding which cases to prosecute. Prosecutors cannot pursue every case to a jury verdict. In deciding which cases or charges to drop or reduce, prosecutors must and do consider such factors as the strength of the evidence, aggravating or extenuating circumstances surrounding the offense, and so forth. But it is one thing to use such criteria to tailor docket loads to available resources without consulting the defense, and quite another to combat crowded dockets by confronting individual defendants with the threat that they can demand a trial only at the cost of risking conviction for a more serious charge or, what is of course the same thing, foregoing the “leniency” of a “lighter” charge that would be available in exchange for a guilty plea.
. Joint Anti-Fascist Refugee Committee v. McGrath, 341 U.S. 123, 179, 71 S.Ct. 624, 652, 95 L.Ed. 817 (1951) (concur-ing opinion of Douglas, J.).
. See President’s Comm’n on Law Enforcement and Administration of Justice, The Challenge of Crime in a Free Society 96-97 (1967); see also, Comment, Interrogations in New Haven: The Impact of Miranda, 76 Yale L.J. 1519, 1588-1589 (1967).
. The counter argument, of course, is that the defendant has “nothing to lose” by demanding a trial. If this is true, and if enough defendants would demand a useless trial simply out of spite — or perhaps hope of remaining free on bail until trial — that only poses the question of why we should deter defendants from demanding a useless trial by threatening them with additional months or, more commonly, years in jail (whether through longer sentences or conviction on more serious charges). If the trial is truly useless, we should be able to induce guilty pleas with a much smaller stick— perhaps a small bribe for a guilty plea or fine for an innocent plea. But, of course, as the concurring opinion recognizes, a bribe would be an impermissible tactic to induce guilty pleas. And a fine imposed upon defendants who insist upon trial would be no more acceptable. The interesting question is why a literal price-tag upon the right to trial is offensive to our concepts of due process, while the figurative pricetag of time in jail is acceptable. The answer may well be in the tendency to define due process. by past process.
. See Alschuler, The Prosecutor’s Role in Plea Bargaining, 36 U.Chi.L.Rev. 50, 65 & n. 44, 103, 104 & n. 131 (1968).
. Eor a slightly more refined analysis, which does not affect the thrust of the reasoning here, see text supra at p. 23.
. The all-winners-no-losers analysis succeeds only if we “indulge an operating presumption of the defendant’s guilt.” Cf. Weinberg & Babcock, Book Review, 76 Yale L.J. 612, 617 (1967). If all defendants were guilty, we would not need to consider the possible chilling ef-*279feet upon exercise of the right to trial, since upon this assumption there are no innocent defendants for whom the right to trial is important. In this improbable world, if we make the additional assumption that the sentences meted out in the absence of a plea bargaining system would be the same as those meted out after trial in a system characterized by plea bargaining, it might be accurate to say that the defendant who is allowed to “cop” a plea to a less serious charge “wins,” while the (by hypothesis guilty) defendant convicted after trial is no worse off than he would be in the absence of a plea bargaining system. The assumption that all defendants are guilty is, of course, ludicrous if made in an absolute sense. But if one believes that as a factual matter a large majority of defendants are guilty, there may be a temptation to conclude that the lighter charges obtained by many guilty defendants in exchange for a plea outweighs by far the cost imposed upon the occasional defendant with a nonfrivolous defense who must risk a heavier sentence by going to trial. Regardless of whether one accepts the sort of calculus implicit in this reasoning, in evaluating the validity of the factual assumption upon which it is premised — that most defendants are guilty — -allowance should be made for the likelihood that the operation of a plea bargaining system will reduce the number of defendants actually innocent in fact or law who ever reach trial.