Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
Opinion concurring in part and dissenting in part filed by Chief Judge MIKVA.
STEPHEN F. WILLIAMS, Circuit Judge:A jury found appellant Thomas Todd Jones guilty of possession with intent to distribute fifty grams or more of a substance containing cocaine base. 21 U.S.C. §§ 841(a), 841 (b)(l)(A)(iii). On appeal Jones argues that the judge erred in admitting evidence obtained after an allegedly illegal search, in admitting testimony alleged to be hearsay, and in using Jones’s failure to plead guilty as a reason for choosing a sentence stricter than the low point of the guidelines range. We disagree, and affirm.
* * * * * sis
Early in the morning of May 2, 1990, Jeffrey Huffman, a plain-clothes detective, watched Jones step off a bus at the Greyhound station in Washington, D.C., go to a restaurant in the station, buy some food, and turn to go back to the bus. Huffman approached Jones, identified himself as a police officer, and asked Jones whether he could speak to him. Jones said yes, and in response to Huffman’s questions, told him that he had come from New York and was headed toward Norfolk, Virginia, where he was a student, and that he had left his ticket and student identification on the bus. Jones offered to retrieve these papers, and started toward the bus ostensibly to do so. Huffman followed, either in response to an “invitation” to come along (according to Huffman) or on his own initiative (according to Jones). These events took place in an open area of the bus station and no one blocked Jones’s path. Upon reaching the platform Jones did not enter his bus, but instead “appeared to get some passengers between [Huffman] and him[self]”, and then “took off running” across L Street.
*930Sergeant John Brennan saw Jones run off, and, with Detective Kim Oxendine, pursued him in a police car. After initially-losing contact, they spotted him walking in a nearby alley. Jones ran into a football field, and did not stop when Brennan got out of the car and yelled, “Police. Stop.” Oxendine continued the chase by car, Brennan on foot. Finally, Oxendine stopped Jones in the alley. When Brennan caught up, he ordered Jones to the ground, handcuffed him, and put him in the police car. Brennan asked Jones why he had run, and Jones responded that it was because he had thrown a gun and was afraid. The gun, if there was one, was never found.
Brennan and Oxendine drove Jones back to the bus station. Meanwhile, another officer, Detective Beard, who had not seen Jones take off, was told of his flight. Beard boarded the bus on which he had seen Jones arrive and had the passengers claim their on-board luggage. No one claimed a green tote bag. After the driver told Beard that all passengers scheduled to continue south had reboarded, save one, Beard took the bag from the bus. When Brennan and Oxendine arrived with Jones, the officers showed him the bag. (According to Jones, officers first showed him another bag, asked him whether he owned it, and, after Jones denied ownership, determined that it belonged to someone else.) Jones disclaimed ownership. The officers opened the bag and found in it a substance that later tested positive for cocaine, a sweatshirt with the words “Norfolk University” and a t-shirt with the words “Norfolk State Express” emblazoned on them, a New York-Norfolk bus ticket, and a Norfolk-New York airline ticket in the name of “Todd Jones”.
* * # * * *
Jones argues that Huffman illegally seized him by accompanying him to the bus platform; that Brennan and Oxendine illegally seized him by handcuffing him and putting him in the police car; and that Beard illegally searched the tote bag. None of the three contentions succeeds.
We have repeatedly held, following United States v. Mendenhall, 446 U.S. 544, 554, 100 S.Ct. 1870, 1877, 64 L.Ed.2d 497 (1980) (opinion of Stewart, J.), and Michigan v. Chesternut, 486 U.S. 567, 108 S.Ct. 1975, 100 L.Ed.2d 565 (1988), that officers who approach a person and ask him questions effect a seizure only when “a reasonable person would conclude from the circumstances, and the show of authority, that he was not free to leave the officer’s presence.” United States v. Brady, 842 F.2d 1313, 1314 (D.C.Cir.1988). Here, as in the typical case where we have found no seizure, the officer wore plain clothes, identified himself, did not show his weapon, and refrained from asking intimidating questions. Neither Huffman nor any other officer blocked Jones’s exit; nothing unusual about the time or place would have intimidated Jones; and, although Huffman did not tell Jones that he could refuse to answer questions and could leave, this omission is not significant. See, e.g., United States v. Tavolacci, 895 F.2d 1423 (D.C.Cir.1990); United States v. Maragh, 894 F.2d 415 (D.C.Cir.1990); United States v. Winston, 892 F.2d 112, 115 (D.C.Cir.1989); United States v. Carrasquillo, 877 F.2d 73 (D.C.Cir.1989); United States v. Lloyd, 868 F.2d 447 (D.C.Cir.1989). That Huffman “escorted” or “followed” Jones (on Jones’s account) might have been significant if Jones had told Huffman that he had nothing to say and had begun walking away; we leave this question to a case that raises it. But here Jones told Huffman that he was going to the bus to retrieve his student ID and his ticket; having said that, he should not have been surprised or intimidated when Huffman came with him, if only for the purpose of saving Jones the trouble of reentering the station after retrieving his papers. See Brady, 842 F.2d at 1314-15 (defendant took “crucial initiatives [such as] suggesting that the ticket would have to be retrieved from his room” in the train). Thus, no seizure at all occurred until Jones’s flight triggered pursuit culminating in his capture.
Jones argues that the capture by Brennan and Oxendine constituted an arrest without probable cause. We need not address the government’s argument that the *931entire transaction — from the moment of capture to the moment, ten to thirty minutes later, when the officers opened Jones’s tote bag back at the bus station— was a Terry stop, justified by the suspiciousness of Jones’s flight. See Terry v. Ohio, 392 U.S. 1, 88 S.Ct. 1868, 20 L.Ed.2d 889 (1968). Seen as separate phases, first a Terry stop and then an arrest, the police conduct was legitimate.
The initial phase — the stopping of Jones, his being ordered to the ground and handcuffed, and his transfer to the car— was both consistent enough with the investigative and security purposes of a Terry stop, and short enough, to qualify as such. A Terry stop does not turn into a full arrest merely because the officers use handcuffs and force the suspect to lie down to prevent flight, so long as the police conduct is reasonable. United States v. Laing, 889 F.2d 281, 285 (D.C.Cir.1989). The stop may also include forced relocation of the suspect. United States v. White, 648 F.2d 29, 34-40 (D.C.Cir.1981). Jones’s flight and failure to obey Brennan’s order to stop supplied the necessary articulable suspicion. A suspect is “free to leave” a non-seizure interview, but when he does so by abruptly bolting after having consented to talk, the officers are free to draw the natural conclusions. See United States v. Nurse, 916 F.2d 20, 24 (D.C.Cir.1990); United States v. Haye, 825 F.2d 32, 34 (4th Cir.1987). Once Jones told the officers that he had possessed and thrown away a gun, almost immediately after the initial capture, they had probable cause to arrest him.
Jones argues that the officers violated his rights by searching his tote bag. The trial court ruled that Jones had no reasonable expectation of privacy in the bag, as he had abandoned it, either by fleeing from the station or by disavowing ownership. The latter suffices, as we have held that “voluntary denial of ownership demonstrates sufficient intent of disassociation to prove abandonment.” United States v. Lewis, 921 F.2d 1294, 1302 (D.C.Cir.1990). Nothing in the record indicates that Jones’s denial of ownership was involuntary or defective for any other reason.
. * * * * * *
Jones argues that the trial court violated his sixth amendment rights by failing to compel the government to produce the name and address of a potential witness until the day of the trial. See Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). The police had interviewed a passenger on the bus but the government did not call her as a witness. The government said that she would not provide exculpatory evidence and that it could not now reach her, and refused to provide her name and address to the defense. The court ordered the government to release this information and it did so, on the day of the trial. Jones argues that the passenger might have testified that the government had opened his bag before obtaining his consent; moreover, by contradicting the testimony of the officers as to this issue, the passenger might have impeached their testimony as to others.
There are two separate elements to this complaint, both defective. First, there is a suggestion of the existence of the undisclosed Brady material. But the suggestion is fanciful. There is not a shadow of a hint that Jones’s fellow traveller had told the government anything remotely exculpatory. Second, there is an argument that, assuming the name and address of the traveller to be in themselves exculpatory (or sufficiently promising as leads to exculpatory evidence as to be on that account disclosable under Brady — itself a stretch), the government’s delay in turning it over might have prejudiced the defendant. United States v. Tarantino, 846 F.2d 1384, 1416-18 (D.C.Cir.1988), holds that a defendant’s delay claim can prevail if he “establishes] that had the [material] been disclosed earlier, there is a probability sufficient to undermine our confidence in the actual outcome [so] that [we believe that] the jury would- have acquitted.” Id. at 1417. Defense counsel was in a position to produce the raw material for such a claim, by finding and interviewing the passenger. As he neither found her, nor even offered *932evidence that she had become less accessible between some hypothetical date of proper disclosure and the trial itself, another “if” is added to the pile — we have no way of knowing whether the delay even affected the evidence available to the defense, much less whether the incremental evidence (if any) might have influenced the jury. The likelihood of these ifs working out.so as to alter the outcome is effectively nil. Brady provides no basis for reversal.
sit * * * * *
Jones argues that the trial court erred by allowing the prosecutor to ask a leading question and by allowing a witness to give hearsay testimony. Both allegations of error attack Detective Beard’s testimony that a passenger identified Jones as the person who owned the tote bag he found on the bus. The leading question, however, elicited testimony connecting Jones and the tote bag by so obscure a link that we doubt that any jurors saw it; the other hearsay more clearly described the passenger’s perception of the Jones/bag link, but was in fact a follow-up on redirect to defense questions that brought out similar, but misleading, hearsay. In any event, any conceivable error was harmless, as the admissible evidence linking Jones to the tote bag was overwhelming. When Beard entered the bus after learning of Jones’s flight and had the passengers claim their luggage, the tote bag was the only leftover item and Jones the only missing passenger. Jones had said that he was a student at Norfolk State University (although apparently he was not) and the tote bag contained a t-shirt bearing the words “Norfolk State Express” and a sweatshirt bearing the words “Norfolk State University”. The tote bag contained a plane ticket in Jones’s name purchased in Norfolk for a flight from Norfolk to New York and a bus ticket for the trip from New York to Norfolk. Because we are sure that a jury would have found Jones guilty even in the absence of the disputed but murky testimony, we need not finally resolve its admissibility. See Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967); United States v. Stock, 948 F.2d 1299, 1302 (D.C.Cir.1991).
* * * * * *
Finally, Jones contends that the trial court imposed an unlawful sentence. The sentencing guidelines set his offense level at 34 which, given a criminal history score of 1, established a sentencing range of 151 to 188 months. U.S.S.G. § 2D1.1(a)(3). A two-point reduction for acceptance of responsibility yielded a sentencing range of 121 to 151 months. Id. at § 3El.l(a). The government had refused to plea bargain with Jones, although that refusal, of course, did not prevent Jones from pleading guilty and relying on the judge’s independent response. In giving Jones a sentence of 127 months the judge said:
I do think that there is some premium that should be recognized for pleading guilty in advance of trial rather than taking a case to trial in which the defendant knows that he is guilty and he is properly charged and there is no defense to it.
I do, however, think that in the circumstances, I intend to give Mr. Jones a major portion of the benefit that he derives from his acceptance of responsibility-
I would, had this case been disposed of with a plea in advance of trial, have sentenced him at the very bottom of the guidelines and imposed the minimum sentence that I could possibly have imposed.
Because, however, the case did go to trial, I am going to add an additional six months to the guideline sentence that I intend to impose, and will impose a sentence of 127 months.
I am articulating this so that anybody that wishes to take it to the sentencing commission and/or the court of appeals may do so. I would like to have some thought given to the considerations I’ve articulated here today. And I would like to know whether or not there is some constitutional error I commit by recognizing that the case was taken to trial, albeit a matter of constitutional right to take the case to trial, rather than ac*933knowledging in advance the guilt that was obviously supported by the proof.
Jones argues that imposition of what the trial judge called the “additional” six months “punishes” him for exercising his right to trial in violation of the U.S. Constitution. Jones’s Brief at 36. See United States v. Watt, 910 F.2d 587, 592 (9th Cir.1990) (“a sentencing court cannot consider against a defendant any constitutionally protected conduct”) (emphasis in original). We find no constitutional obstacle to the extra time.
A triad of decisions now nearly 20 years old might seem to draw in question any sentencing practice that disfavored the decision to go to trial. In United States v. Jackson, 390 U.S. 570, 88 S.Ct. 1209, 20 L.Ed.2d 138 (1968), the Court struck down a statute that authorized only a jury to impose a death sentence, leaving life imprisonment the ceiling for those who waived jury trial. The Court reasoned that such a rule (a) had the effect of deterring the exercise of the rights to demand a jury trial and to plead not guilty, and (b) was not necessary to achieve the stated goal of mitigating the severity of capital punishment. Id. at 581-82, 88 S.Ct. at 1216-17.
One might dismiss Jackson as driven by the especially excruciating character of choices risking the death penalty, or by a perceived disproportion of the consequences for exercise of the right to trial (and to jury trial), as weighed against the benefits of encouraging ap plea, or by the extreme implausibility of the government’s defense (that the rule mitigated the severity of the punishment), but North Carolina v. Pearce, 395 U.S. 711, 89 S.Ct. 2072, 23 L.Ed.2d 656 (1969), speaks more broadly. There the Court held that a trial court violated due process by imposing on remand a sentence longer than the one the defendant successfully challenged on appeal. Reasoning that “vindictiveness” against a successfully appealing defendant could not legally play a part in his resen-tencing, and that fear of such vindictiveness “may unconstitutionally deter a defendant’s exercise of the right to appeal”, id. at 725, 89 S.Ct. at 2080, the Court established a prophylactic rule: a longer sentence imposed after a prior sentence was vacated would be presumed the result of unconstitutional vindictiveness or retaliation, but the trial judge could rebut the presumption if at resentencing he identified intervening conduct by defendant that justified a longer sentence. Id. at 726, 89 S.Ct. at 2081.
Finally, in Blackledge v. Perry, 417 U.S. 21, 94 S.Ct. 2098, 40 L.Ed.2d 628 (1974), the Court held that a defendant who had been convicted of a misdemeanor in a bench trial, and who sought to exercise his right to a jury trial de novo, could not be repro-secuted on a more serious felony charge. The Court explained this outcome, and reconciled it to some intervening cases that had come out the other way,1 by saying that the due process clause was not offended by “all possibilities of increased punishment upon retrial after appeal, but only by those that pose a realistic likelihood of ‘vindictiveness.’ ” Id. at 27, 94 S.Ct. at 2102. While not defining “vindictiveness”, the Court seemed to equate it with a desire to discourage exercise of the right to trial or appeal. Id. at 27-28, 94 S.Ct. at 2102-03.
Other cases, typically later ones, establish that in many contexts the government (state or federal) may impose sentencing differentials that favor defendants who plead guilty over those who go to trial. The most obvious cases are those in which the Supreme Court has affirmed the constitutionality of plea bargaining — the practice by which a criminal defendant gives up his right to a trial in exchange for what he believes to be a better prospect in sentencing than would have flowed from convic*934tion after trial. Thus in Brady v. United States, 397 U.S. 742, 90 S.Ct. 1463, 25 L.Ed.2d 747 (1970), the Court found a guilty plea voluntary and valid even though the defendant risked the death penalty if he went to trial but was assured no worse than life imprisonment if he pleaded guilty. (The plea was under 18 U.S.C. § 1201(a), the very provision invalidated in United States v. Jackson, but was entered before the decision in Jackson.) And in Borden-kircher v. Hayes, 434 U.S. 357, 98 S.Ct. 663, 54 L.Ed.2d 604 (1978), where the prosecutor responded to a defendant’s refusal to plead to charges carrying a range of two-to-ten years by reindicting him under a habitual offender statute carrying a mandatory life sentence, the Court found no constitutional violation in the prosecutor’s conduct. Recognizing that the prosecutor sought to induce a guilty plea by threatening the defendant with the recidivist indictment, and by carrying out the threat, the Court denied that conduct so motivated showed “vindictiveness” of the sort forbidden by cases such as Pearce. See 434 U.S. at 360-64, 98 S.Ct. at 666-69. In both Brady and Bordenkircher the Court identified the interests of preserving both prosecutorial and judicial resources as justifying the provision of inducements to plead guilty. See Brady, 397 U.S. at 752, 90 S.Ct. at 1471.2
It is implicit in Brady and explicit in the later decision of Corbitt v. New Jersey, 439 U.S. 212, 99 S.Ct. 492, 58 L.Ed.2d 466 (1978), that the state may create a sentencing scheme that systemically favors pleas of guilty. In Corbitt the Court found no violation of a defendant’s right to trial in a statute that required a life sentence for first degree murder when defendant went to trial, but permitted the judge to impose a lower sentence if he pleaded non vult or nolo contendere. The majority regarded the. decision as a logical application of guilty plea cases such as Brady and Bordenkircher. Id. at 218-23, 99 S.Ct. at 497-500. Justice Stewart concurred, rejecting the majority's equation of the case with plea bargaining on the ground that settlements by a prosecutor with his adversary stood on a different basis from general state legislation. But, noting that even a pleading defendant could be sentenced to the maximum, whereas one going to trial might be convicted of a lesser-included offense or even acquitted, he argued that it was “impossible to state with any confidence that the New Jersey statute does in fact penalize a defendant’s decision to plead not guilty.” Id. at 226-27, 99 S.Ct. at 501-02. As the defendant who goes to trial always enjoys the possibility of an acquittal, the uncertainty that reconciled Justice Stewart to the New Jersey statute would appear to be present no matter how clear-cut the sentencing differential.
While Jackson, Pearce and Blackledge may seem to suggest that all practices tending to deter the exercise of a right to trial or appeal by attaching a risk of heavier punishment constitute forbidden “vindictiveness” (or “realistic likelihood of ‘vindictiveness’ ”, see Blackledge, above), any such broad reading is inconsistent with Bordenkircher and Corbitt. The Supreme Court has made two fairly recent efforts to reconcile the cases. In one, Alabama v. Smith, 490 U.S. 794, 109 S.Ct. 2201, 104 L.Ed.2d 865 (1989), it refused to apply the Pearce presumption to a sentence imposed on a defendant after trial that was longer than what he had received after a guilty plea that he had successfully challenged. The Court, overruling a case decided as a companion to Pearce, explained that where the initial sentence was imposed after a guilty plea, it was proper to infer that any longer sentence was due to the fuller information developed at trial. Id. at 801-02, 109 S.Ct. at 2205-06. But this cannot be read as suggesting that only the possibility of additional information can justify a harsher sentence. In Corbitt, for example, the statute mandated life for those who went to trial, allowing the court discretion only for those who pleaded guilty. Not only did the statute extinguish discretion for judges who presumably had access to *935the broadest range of information, but the Court explicitly said that “a State may encourage a guilty plea by offering substantial benefits in return for the plea.” 439 U.S. at 219, 99 S.Ct. at 497 (footnote omitted).
United States v. Goodwin, 457 U.S. 368, 384, 102 S.Ct. 2485, 2494, 73 L.Ed.2d 74 (1982), suggests another basis of reconciliation. There the Court found no undue risk of “vindictiveness” in a prosecutor’s charging the defendant with a felony after he had refused to plead guilty to a misdemeanor charge and demanded a jury trial. The defendant sought to distinguish Bor-denkircher on the theory that the insistence on a jury trial was so burdensome on the prosecution as to pose excessive risks of vindictiveness. The Court rejected that theory. Contrasting Blackledge, it stressed that a higher charge after trial would entail “duplicative expenditures of prosecutorial resources”, id. at 383, 102 S.Ct. at 2494, and similarly that in Pearce the appeal required the trial court “ ‘to do over what it thought it had already done correctly’ ”, id. (quoting Pearce). If, then, we set aside Jackson on any of the special grounds suggested above, such as the dis-proportionality of means to ends, perhaps the answer is that the inference of improper vindictiveness normally arises only when defendant’s exercise of a right subjects judge or prosecutor to what may seem a pointless and irritating repetition, as in Pearce and Blackledge. “Vindictiveness”, then, may refer to pique aroused by the prospect of personal inconvenience or even loss of face (Pearce), rather than to a legitimate concern over waste of prosecuto-rial or judicial resources. Of course the two overlap, as repetition of a trial also consumes the prosecutorial and judicial resources — resources that courts, under such doctrines as res judicata, law of the case, and double jeopardy, may legitimately husband. Compare Goodwin, 457 U.S. at 376-77, 102 S.Ct. at 2490-91 (expressing concern that the values behind those doctrines might subconsciously motivate a vindictive response to defense efforts' to secure retrial). Thus the Court has clearly recognized the possible mingling of legitimate and illegitimate purposes, and has acted only against those practices with a relatively high risk of illegitimate ones predominating.
We do not mean to suggest that the repetition principle is exclusive. First, some sentencing differentials might well be disproportionate to any possible justification. As we have said, that might explain Jackson. (If it does, then Brady needs to be explained. A possible answer is that invalidation of gre-Jackson pleas might have released a host of kidnappers who could have escaped ultimate conviction because of the staleness of evidence. Cf. Teague v. Lane, 489 U.S. 288, 309-10, 109 S.Ct. 1060, 1074-75, 103 L.Ed.2d 334 (1989) (plurality) (discussing the interest in finality); Linkletter v. Walker, 381 U.S. 618, 637-38, 85 S.Ct. 1731, 1742-43, 14 L.Ed.2d 601 (1965).) Second, some schemes might seem arbitrary even if not disproportionate. This may explain Justice Stewart’s suggestion in his Corbitt concurrence that a statute imposing an automatic 50% discount for guilty pleas might be unconstitutional. See 439 U.S. at 227, 99 S.Ct. at 501. Third, judges who encourage guilty pleas by aggressively flourishing sentencing differentials may lose their appearance of impartiality. See United States v. Crocker, 788 F.2d 802, 809 n. 3 (1st Cir.1986); see also note 3 below. But the Supreme Court precedents plainly do not invalidate sentencing differentials aimed at encouraging guilty pleas.
Corbitt, incidentally, refutes the dissent’s effort to dismiss the cases permitting differential sentencing as limited to negotiations between defense and prosecutor, parties of “relatively equal bargaining power”. Dissent at 941 (quoting Borden-kircher, 434 U.S. at 362, 98 S.Ct. at 667). A state legislature laying down a flat rule such as that upheld in' Corbitt is not bargaining at all; it is decreeing. While the dissent purports to reject any idea that “a defendant must pay for exercising his constitutional right to trial”, Dissent at 939, its real objection is to the identity of the purchaser: prosecutors may buy the right, legislatures may, but judges may not.
Relevant precedent in this circuit predates the more important Supreme Court *936cases on differential sentencing (especially, Corbitt, Goodwin, and Smith), and so can lay no claim to our loyalty. United States v. Hopkins, 464 F.2d 816, 822 (D.C.Cir.1972), held that the trial judge erred by increasing the sentence because the defendant refused to express remorse and to plead guilty. Scott v. United States, 419 F.2d 264, 274 (D.C.Cir.1969), held that although a prosecutor may engage in plea bargaining, a trial judge may neither involve himself in the process nor even express a policy of differential sentencing. Although clearly the prosecutor’s proper role is quite different from the judge’s, Scott’s apparent ban on any judicial consideration of the defendant’s choice between plea and trial is hard to reconcile either with Blackledge, which suggested that a prosecutor is at least as likely a source of vindictiveness as a judge, 417 U.S. at 27, 94 S.Ct. at 2102, or with Roberts v. United States, 446 U.S. 552, 100 S.Ct. 1358, 63 L.Ed.2d 622 (1980), which upheld judicial consideration of the defendant’s failure to cooperate with the government, a failure that is typically (though not invariably) linked with the decision to stand trial.
Many of the other circuits apparently believe that to give longer sentences to defendants who plead guilty than to those who do not impermissibly burdens the right to trial. We can ignore those cases that predate the critical Supreme Court developments. See, e.g., Fielding v. LeFevre, 548 F.2d 1102, 1106 (2d Cir.1977) (trial judge’s statement to defendant that he would get a more severe sentence if he did not plead guilty, if true, would violate right to trial); United States v. Capriola, 537 F.2d 319 (9th Cir.1976) (disparity of sentence length between codefendant who pleads guilty and codefendant who goes to trial signifies, in the absence of formal reasons, unconstitutional burdening of right to trial); Hess v. United States, 496 F.2d 936, 938 (8th Cir. 1974) (sentence may not be related to defendant’s decision whether to plead guilty); Baker v. United States, 412 F.2d 1069, 1073 (5th Cir.1969) (same). And we can ignore those cases that, while more recent, rely on outdated precedents and do not consider the relevant Supreme Court developments. See, e.g., United States v. Monroe, 943 F.2d 1007, 1017-18 (9th Cir.1991) (while suggesting that Capriola has been obviated by Sentencing Guidelines, still apparently endorsing its conclusion that the decision to plead may not bear on the sentence); United States v. Frost, 914 F.2d 756, 774 (6th Cir.1990) (decision to plead may not bear on sentence, though mere disparity of sentences among codefendants who pled guilty and those who did not is not sufficient evidence of such a practice); Watt, 910 F.2d at 592 (trial judge erred by enhancing sentence because defendant did not plead guilty). That leaves, as far as we can find, just United States v. Mazzaferro, 865 F.2d 450, 457-60 (1st Cir.1989), where the court vacated a 20-year sentence and fine imposed on a defendant who stood trial, when more culpable codefendants with far worse criminal records, who pled guilty, received only ten. While Mazzafer-ro purports to rely on the whole string of Supreme Court plea bargaining cases, it does not even acknowledge the conflict between the Pearce and Brady lines, much less apply the reconciliations offered in Smith and Goodwin. See Mazzaferro, 865 F.2d at 458. Other than its own precedents, see Crocker, supra, Longval v. Mea-chum, 693 F.2d 236 (1st Cir.1982),3 the Mazzaferro court could find support only in circuit court cases decided before 1978. 865 F.2d at 459-60. Moreover, the Mazzaf-erro outcome might well be justifiable on one of the principles suggested above— that the differential should not be disproportionate to the state interests advanced.
One line of circuit court decisions lends affirmative support to our position — the *937line upholding § 3El.l(a) of the Guidelines, which calls for a two-point reduction in offense level for “acceptance of personal responsibility”, together with the Commentary’s clear indication that a defendant who stands trial has an uphill fight to win the reduction (Application Note 2, “Conviction by ... trial does not automatically preclude a defendant from consideration for such a reduction”; Note 3: “Entry of a plea of guilty prior to the commencement of trial combined with truthful admission of involvement in the offense and related conduct will constitute significant evidence of acceptance of responsibility”):4 United States v. Gonzalez, 897 F.2d 1018, 1021 (9th Cir.1990); United States v. Henry, 883 F.2d 1010, 1011 (11th Cir.1989); United States v. White, 869 F.2d 822, 826 (5th Cir.1989). See also United States v. Parker, 903 F.2d 91, 105 (2d Cir.1990) (upholding § 3El.l(a), but with no focus on link between acceptance of responsibility and plea).
The cases, to be sure, invoke a somewhat obscure formula. White, for example, though explicitly assuming “that the sole purpose of this guideline is to encourage guilty pleas,” 869 F.2d at 826, goes on to say: “The fact that a more lenient sentence is imposed upon a contrite defendant does not establish a corollary that those who elect to stand trial are penalized.” Id. We do not read this as invoking a metaphysical distinction between subtracting time for pleading guilty and adding time for going to trial. Rather, we understand it to accord with our reading of the Supreme Court’s decisions — that sentencing differentials based on guilty pleas are not unconstitutional burdens on the right to stand trial. See also Roberts, 445 U.S. at 557 n. 4, 100 S.Ct. at 1362 n. 4 (“We doubt that a principled distinction may be drawn between ‘enhancing’ the punishment imposed upon the petitioner and denying him the ‘leniency’ he claims would be appropriate if he had cooperated.”); United States v. Bell, 905 F.2d 458, 462 (D.C.Cir.1990) (for judge to give a defendant who goes to trial a higher sentence than a coconspirator who pleads guilty and cooperates is not to “punish” him, citing Roberts).
In light of the circuit court decisions on § 3El.l(a), the dissent’s effort to distinguish Corbitt mystifies. The dissent argues that a legislature may reward a plea by means of bright line rules, while a judge may not, because the legislature’s rules (1) are less likely to be vindictive and (2) give a clear warning. Dissent at 943-44. That analysis does not work even for Corbitt itself, for there the legislature simply set a floor under post-trial sentences, giving judges discretion over the existence and size of any discount for a plea. There is nothing in Corbitt about the relative risks of vindictiveness in legislatures and judges, nor about the virtues of clear warning. On the contrary, the opinion emphasizes the importance of discretion, and, in finding “no difference of constitutional significance between Bordenkircher and this case”, treats judges and prosecutors the same. See Corbitt, 439 U.S. at 221-22, 99 S.Ct. at 498-99. Similarly, the Guidelines’ invitation to individual judges to consider guilty pleas in assessing “acceptance of responsibility” clearly vests the district judges with discretion over the matter.
We recognize, of course, that the § 3El.l(a) cases do not endorse the proposition that a judge may justify a refusal to reduce the sentence on the theory that he may encourage guilty pleas — only that a judge may use a refusal to plead *938guilty as evidence that the defendant has not accepted responsibility. For example, in United States v. Rodriguez, 959 F.2d 193, 197 (11th Cir.1992), the court set aside a sentence where the judge’s inquiries suggested that he regarded an intent to appeal as inconsistent with remorse. Cf. North Carolina v. Pearce. The court understood § 3E1.1 to allow the judge to use the absence of a guilty plea as evidence of lack of remorse, but not as such conclusive evidence as to overcome strong evidence the other way, evidently regarding that reading as constitutionally necessary. Even under this view, the trial judge here could evidently have awarded the two-level reduction for acceptance of responsibility (lowering the minimum sentence by 30 months, to 121 months), and then added six months, noting that because of defendant’s failure to plead guilty, his acceptance of responsibility was only borderline. We believe, however, that the precedents give the trial court a slightly broader authority — to consider the institutional value of guilty pleas as an explicit, independent basis of sentence adjustment.
The Supreme Court has held that plea bargaining may be justified by the state’s interest in conserving prosecutorial and judicial resources. See pp. 934-35, above {Brady, Corbitt, Goodwin). This interest can be pursued only if trial judges engage in differential sentencing. If we put aside such factors as risk aversion and attorney costs (which will not affect the defendant if he is indigent),5 a guilty plea will be attractive to a defendant only if (a) the expected value of the sentence after the plea (an average of each possible sentence, weighted in accord with probability) is less than (b) the expected value of the sentence after trial (an average, taking into account both likely sentence lengths (weighted for likelihood) and the chances of acquittal). For example, if all convicted of a given crime were sentenced to 10 years, and a quarter of those who went to trial were acquitted, the expected value of the sentence for those going to trial would be 7.5 years. Defendants would not plead guilty (except to the extent they were moved by genuine remorse) if they thought such a plea would yield a sentence longer than 7.5 years (expected value).6
Thus the interest in preserving prosecu-torial and judicial resources through lesser sentences requires that judges engage in differential sentencing — either by deferring to prosecutorial recommendations (as is common in state systems) or to legislative directives (as in Corbitt), or independently. It would be highly formalistic to forbid only the latter. To deny judges the power to consider the institutional concern explicitly would lead to some combination of two dubious results. First, it might limit use of such considerations to specialized formats (e.g., where lesser offenses can be substituted7 or where judges play a rather submissive role in sentencing), leaving judges as participants in somewhat rigid types of plea bargaining, but as automa-ta. Second, as there is no mechanism at the disposal of appellate courts for identifying sentence discounts for guilty pleas, it might shove judicial reliance on institutional concerns underground (or slightly underground, as § 3El.l(a) seems to do). In light of the first alternative, reversal here would draw a line between implicit judicial endorsements of a trade-off between standing trial and sentence increments, which the Supreme Court accepts, and explicit endorsements. Underground differentials would simply compound the artificiality by adding hypocrisy.
Accordingly, we affirm Jones’s conviction and sentence.
So Ordered.
. See Chaffin v. Stynchcombe, 412 U.S. 17, 93 S.Ct. 1977, 36 L.Ed.2d 714 (1973) (that jury gave longer sentence on remand than prior jury had given defendant before successful appeal does not give rise to Pearce presumption); Colten v. Kentucky, 407 U.S. 104, 92 S.Ct. 1953, 32 L.Ed.2d 584 (1972) (that judge after de novo trial gave defendant greater punishment than he had received in a prior, expedited non-jury trial does not give rise to the Pearce presumption of vindictiveness); see also Ludwig v. Massachusetts, 427 U.S. 618, 96 S.Ct. 2781, 49 L.Ed.2d 732 (1976) (possibility of higher sentence after appealing from conviction after expedited bench trial to de novo jury trial does not give rise to the Pearce presumption).
. The Court reaffirmed the constitutional validity of these interests in Corbitt v. New Jersey, 439 U.S. 212, 220 n. 9, 99 S.Ct. 492, 498 n. 9, 58 L.Ed.2d 466 (1978), and United States v. Goodwin, 457 U.S. 368, 378-79 & n. 10, 102 S.Ct. 2485, 2491-92, & n. 10, 73 L.Ed.2d 74 (1982).
. In Crocker and Longval the court was apparently concerned that trial judges, by giving enthusiastic pre-trial or mid-trial warnings of the consequences of going to trial (or persisting in trial), lost their appearance of impartiality. See especially Crocker, 788 F.2d at 809 n. 3. We view such conduct as quite distinct from an , implicit policy or a pattern of differential sen-fences, though recognizing that the latter have a signaling effect. See Scott, 419 F.2d at 274. Further, judicial endorsement of plea bargaining, either by following prosecutors’ leniency recommendations or by accepting plea-bargained dismissal of extra counts, carries a' similar signal, no matter how little it may be intended.
. On April 30, 1992 the Sentencing Commission submitted a Guideline amendment to Congress to permit judges to increase the reduction to three levels if the offense level is 16 or greater, and the defendant provides the government with complete information on his offense or “timely notifies] authorities of his intention to enter a plea of guilty, thereby permitting the government to avoid preparing for trial and permitting the court to allocate its resources efficiently”. See United States Sentencing Commission, Amendments to the Sentencing Guidelines for the United States Courts, 57 Fed.Reg. 20148, 20156 (1992). This amendment may have been a response to observations such as the one made by this court in United States v. Dukes, 936 F.2d 1281, 1282 (D.C.Cir.1991), suggesting that the Sentencing Guidelines were designed to permit the practice of plea bargaining "without explicitly telling defendants they will receive a higher sentence if they go to trial."
. Another complicating factor may be discounts for distant years — i.e., the last ten years of a 20-year sentence may seem less weighty than the ten years of a ten-year sentence.
. The record suggests that judges in fact establish differentials, and some studies suggest that they correspond to ones this analysis suggests are appropriate. See Douglas A. Smith, The Plea Bargaining Controversy, 77 J.Crim.L. & Criminology 949 (1986); see especially id. at 959-60 for explanation of the effect of likelihood of acquittal.
.Because the sentencing range for most crimes is broad, this is often an unwieldy device. The Guidelines’ provision for consideration of "relevant conduct” makes it even less practicable. See U.S.S.G. § 1B1.3; United States v. Salmon, 948 F.2d 776 (D.C.Cir.1991); Dukes, supra.