The validity of sentencing is the issue on this appeal.

Appellant was convicted by a jury on three counts of an indictment, all based upon a single armed robbery of a bank. Following denial of appellant's post-trial motions, he was sentenced on May 9, 1967 as follows:

Count I, under 18 U.S.C. § 2113(a), charging entering a bank with intent to commit a felony — 10 years;

Count II, also under 18 U.S.C. § 2113 (a), charging robbery of the bank — 5 years, consecutive with the preceding sentence;

Count III, under 18 U.S.C. § 2113(d), charging bank robbery attended by jeopardizing life with a dangerous weapon — 5 years probation, to be served at the expiration of the sentence of Count II.

An appeal was taken, and this court affirmed the conviction in a per curiam opinion. United States v. Corson, 389 F.2d 563 (1968).

On August 31, 1969, while imprisoned, appellant sought to invoke the power of the district court under Rule 35, Federal Rules of Criminal Procedure, "to correct an illegal sentence at any time" by moving for vacation of the sentences imposed under Counts I and II. The court responded, however, by vacating the sentences under Counts II and III on October 21, 1969 and letting stand the heavier sentence under Count I.

The sole question presented by this appeal is whether the district court's Order of October 21, 1969 constituted a proper and lawful implementation of the Supreme Court's holding in Prince v. United States, 352 U.S. 322, 77 S. Ct. 403, 1 L. Ed. 2d 370 (1957). In Prince, the Court held that it was not Congress' intention, in establishing a series of greater and lesser offenses under the bank robbery statute, to pyramid the penalties therefor.1 Accordingly, the Court held that it was error for the trial court to have sentenced Prince to consecutive terms of twenty years for robbery and fifteen years for unlawful entering, since the "mental element (intent to steal) merges into the completed crime if the robbery is consummated." 352 U.S. at 328, 77 S.Ct. at 407. The case was therefore remanded to the district court for resentencing "in accordance with this opinion." 352 U.S. at 329, 77 S. Ct. 403.

Even before Prince, it was generally agreed among the circuits that a defendant could not be given multiple sentences under the bank robbery statute as punishment for a single bank robbery.2 The circuits have disagreed both before and after Prince, however, as to the theory upon which the correction of an erroneously cumulative sentence should rest. Basically two approaches have been taken — merger and the intention of the sentencing judge.3

1. The Merger theory. The above-quoted merger language in Prince has been read by several circuits, including our own, as requiring a holding that sentence may validly be imposed only upon that count which charges the most aggravated form of the offense since, in theory, the less aggravated offenses have "merged" into it and ceased to exist as separate, punishable offenses.4 This is the theory urged by appellant, who, if successful in urging a mechanical application of this rule, would be entitled to be released from prison and merely placed on probation for five years.

2. The court's intention. Other circuits, most notably the Eighth, have adopted a more flexible approach — that the sentence retained should be that which most clearly approximates the intention of the district judge at the time of the original sentencing. This theory assumes that the sentencing judge clearly intended to impose on the defendant at least as severe a penalty as that indicated for any single count. They have therefore permitted to stand the heaviest penalty, irrespective of whether it was imposed upon the most aggravated of the counts. They have therefore vacated all sentences but the longest.5 The Government urges this approach, which in this case would result in our affirming the lower court's retention of the ten-year prison term.

A review of the cases using these two approaches reveals the striking fact that, regardless of the approach used, the result has almost invariably been to uphold the heaviest sentence and vacate the lighter sentences.6 One is also struck by the effect the rather convoluted penalty provisions of § 2113 have had upon the various courts' choice of approach.7

The most inclusive offense in § 2113 is aggravated robbery, which offense also carries the heaviest maximum penalty. In those cases in which the defendant has been convicted on aggravated robbery as well as the lesser included offenses, the "merger" theory has generally worked well to correct an erroneous pyramiding of sentences. Since the heaviest sentence has usually been imposed on the aggravated count, it has done little violence to the intention of the sentencing judge to uphold that sentence on the ground that the lesser offenses merged into the most inclusive offenses and no longer formed a proper basis for sentencing.8

Courts have had great difficulty with the merger theory, however, in cases in which the more severe sentence was not imposed on the most inclusive offense. Typically, such cases have involved two counts — one for entering with the intent to commit a felony and one for larceny.9 The problem arises because Congress has authorized a much heavier penalty for unlawful entering, the less inclusive of these two crimes, and therefore sentencing judges have often imposed correspondingly more severe sentences on this less inclusive offense. Were the merger theory to be applied to this sort of case, a defendant (1) would be penalized for having failed, after entering with the requisite intent (punishable by twenty years in prison and a fine), to have consummated the larceny, (2) would be rewarded for having actually committed the larency (punishable by no more than 10 years in prison and a fine), and (3) would be even more handsomely rewarded if, despite having committed the larceny, he was unsuccessful in garnering loot in excess of $100 (in which case, he could be sentenced to only one year in prison and a fine). Because of this anomaly, and because the merger theory, applied in this situation, would do great violence to the intention of the sentencing court, courts have been understandably reluctant to apply the merger theory to this sort of case.

There has been a common variation on this situation, arising (as in the instant case) when, despite a conviction on an aggravated count punishable by 25 years in prison and a fine, the trial judge has nonetheless imposed the heaviest sentence on one of the lesser counts. The courts have similarly refused to apply the merger theory to this sort of case.10

This court has had no reason, prior to this case, to question the wisdom of the merger theory, since Welty, Conway, McKenzie, and Chester all involved situations in which the application of that theory did not lead to the mischievous results outlined above. This case, however, reveals plainly the difficulties inherent in that approach. A defendant found guilty of aggravated bank robbery and upon whom the district judge intended to impose a sentence of no less than 15 years plus 5 years probation would incur a sanction of only 5 years probation because the district judge has erred. Therefore, we have been compelled to reexamine the underpinnings of the merger theory. Upon consideration of all the relevant authorities, we have concluded that the merger theory is inconsistent with sound judicial policy and ought not hereafter be followed.

The merger theory stems from the Supreme Court's decision in Prince v. United States, supra. There, the Supreme Court was faced with a situation in which the district court had sentenced a defendant convicted of a single bank robbery to 20 years for robbery and 15 years for entering to be served consecutively. Interpreting the statutory scheme to resolve an ambiguity in favor of leniency the Supreme Court held that when it established lesser included offenses in the Bank Robbery Act, Congress did not intend to permit the pyramiding of penalties imposed for a single robbery. In 1937 Congress had amended the Bank Robbery Act of 1934 to plug a glaring loophole by making unlawful entry and larceny from a bank federal crimes. Prior to the amendment it was possible to enter a bank and steal money from the till during the distraction of the teller without violating federal law. It was in this context of interpreting Congressional intent to create lesser included offenses without increasing the maximum penalty for the most inclusive offense that the Court stated:

"[T]he unlawful entry provision was inserted to cover the situation where a person enters a bank for the purpose of committing a crime but is frustrated for some reason before completing the crime. The gravamen of the offense is not in the act of entering. * * * Rather the heart of the crime is the intent to steal. This mental element merges into the completed crime if the robbery is consummated. To go beyond this reasoning would compel us to find that Congress intended, by the 1937 amendment, to make drastic changes in authorized punishments. This we cannot do." 352 U.S. at 328, 77 S.Ct. at 406 (Italics added).

We do not read this language, however, either as reflecting the common law doctrine of merger of offenses,11 or as having formulated some new, and analogous, "merger" doctrine. Since the evil at which Prince was directed was the pyramiding of sentences, we see no reason to ascribe to the above "merger" language the wholly gratuitous effect of limiting a sentence to the most inclusive count. Such an effect is neither necessary to effectuate, nor required by, the holding in the case.

If the merger theory is to be rejected, how then should sentence be imposed when there are multiple convictions arising out of a single bank robbery?

The most obvious alternative to the merger theory is the approach implicitly mandated in the "court's intention" cases, which would allow the district court to impose sentence on any one of the counts, but not on more than one.12 We must reject this approach as well, however, in part because the approach actually does not remedy the pyramiding problem at which Prince was directed.

Since the convictions on the various counts would not "merge" under this approach, the defendant would presumably be entitled to appeal his conviction on all the counts. To enable him to do so, judgment would have to be entered on each of them. The entry of a valid appealable judgment of conviction on each count required that sentence be imposed thereon. (For purposes of determining appealability, the term "sentence" includes probation with suspension of the imposition of any other penal sanctions. Korematsu v. United States, 319 U.S. 432, 63 S. Ct. 1124, 87 L. Ed. 1497 (1943).) "Final judgment in a criminal case means sentence. The sentence is the judgment." Berman v. United States, 302 U.S. 211, 212, 58 S. Ct. 164, 166, 82 L.Ed 204 (1937). See Federal Rules Criminal Procedure 32(b). Thus, in addition to imposing a sentence on one count, the court would also have to impose, at the very least, probation on the remaining counts, since "judgment which suspends sentence without imposing probation is improper and is a nullity." United States v. Fried, 436 F.2d 784, 787 (6th Cir. 1971); see Ex Parte United States, 242 U.S. 27, 37 S. Ct. 72, 61 L. Ed. 129 (1916); United States v. Ellenbogen, 390 F.2d 537 (2nd Cir.), cert. den. 393 U.S. 918, 89 S. Ct. 241, 21 L. Ed. 2d 206 (1968); United States v. Sams, 340 F.2d 1014 (3rd Cir.), cert. den. 380 U.S. 974, 85 S. Ct. 1336, 14 L. Ed. 2d 270 (1965); United States v. Graham, 325 F.2d 922 (6th Cir. 1963); 18 U.S.C. § 3651; cf. Miller v. Aderhold, 288 U.S. 206, 53 S. Ct. 325, 77 L. Ed. 702 (1933). Unfortunately, a sentence on one count plus suspension of sentence with probation on the remaining counts, while creating appealable judgments on every conviction, at the same time violates the prohibition against pyramided sentences established in Prince.

We must also reject the single-sentence approach because there are situations in which it would lead to results as anomalous as those which would be produced by the merger approach in the instant case. For example, were this court on an appeal to reverse only the conviction supporting the one sentence upon which a prison term had been imposed, the district court would be unable subsequently to salvage its original sentence by transferring the prison sentence to one of the other counts on which only probation had been imposed. Rule 35 of the Federal Rules of Criminal Procedure provides no remedy, since it is expressly limited to the correction of "illegal" sentences, and the law is clear that a valid sentence may not be altered once it has been affirmed on appeal and become final. See Castle v. United States, 399 F.2d 642, 647-648 (5th Cir. 1968); Pugliese v. United States, 353 F.2d 514 (1st Cir. 1965); United States v. Tuffanelli, 138 F.2d 981, 983 (7th Cir. 1943). Even more fundamentally, the trial court's resentencing the defendant on one of the remaining counts would violate the double jeopardy clause of the Fifth Amendment. Ex parte Lange, 85 U.S. (18 Wall.) 163, 21 L. Ed. 872 (1874); United States v. Benz, 282 U.S. 304, 307, 51 S. Ct. 113, 75 L. Ed. 354 (1930); United States v. Welty, supra; Ekberg v. United States, 167 F.2d 380 (1st Cir. 1948).13 Inherent in the single-count sentencing procedure, therefore, is the very real possibility that the reversal of one count of a multi-count bank robbery indictment could effectively immunize a defendant from the district court's intended sentence of imprisonment and "unavoidably result in an overall miscarriage of justice." McMillen v. United States, 386 F.2d 29, 37 (1st Cir. 1967) cert. den. 390 U.S. 1031, 88 S. Ct. 1424, 20 L. Ed. 2d 288 (1968).

For these reasons, we think that the only practicable way of implementing Prince is to impose a general sentence on all counts for a term not exceeding the maximum permissible sentence on that count which carries the greatest maximum sentence. Should there be a reversal of one or more counts, there would be no bar to the district court's imposition of a new sentence so long as it did not exceed the maximum authorized by any of the remaining counts. Moreover, there would be no double jeopardy problem so long as the new sentence was the same as or less than that originally imposed.14

We are aware that this Court has, for good reason, expressed a dissatisfaction with general sentences and has declared it "highly desirable that the trial judge in imposing sentence on an indictment containing more than one count deal separately with each count." United States v. Rose, 215 F.2d 617, 630 (3rd Cir. 1954). We believe, however, that Rose is necessarily inapplicable to bank robbery cases, where multiple sentences are forbidden despite a conviction on more than one count. Even the Fifth Circuit, which has declared general sentences not only "undesirable" but illegal altogether, has recognized such an exception. Hall v. United States, 356 F.2d 424 (5th Cir. 1966). The imposition of one general sentence seems particularly suited to the mandates of Prince. In the imposition of sentences hereafter, we see no other means of insuring in all cases that sentence may validly be imposed on a defendant who stands convicted after appeal of violating one or more subsections of the Bank Robbery Act.15

In this case, the trial court erroneously imposed separate sentences on all three robbery counts. It is impossible to say that certain of these sentences rather than others were "illegal" under Rule 35. Rather, it was the cumulation of sentences, the sentencing in its entirety which was "illegal." It was therefore error for the district court merely to have vacated the sentences on counts II and III and to have left standing the sentence on count I. Thus, all the sentences originally imposed were invalid and ought to have been vacated in their entirety, so that the appellant could then be resentenced.16

The case will therefore be remanded to the district court, with directions (1) to vacate its Order of October 21, 1969, (2) to vacate the multiple sentences originally imposed, and (3) in a proceeding in conformity with Rule 43, to impose a general sentence on all three counts not to exceed the ten year sentence originally imposed on count I.17

HASTIE, Circuit Judge (dissenting).

I dissent because I believe the law does not permit the solution of the dilemma of this case that this court is now sanctioning.

The sentence of 5 years probation that the district court imposed on Count 3 for the most aggravated of three crimes charged in the present indictment was beyond doubt a lawful sentence, as I believe all members of the court recognize. After this court affirmed the conviction, 1968, 389 F.2d 563, the case was returned to the district court and that sentence became final. A year later, the district court undertook to set it aside in order to validate a sentence on another count that could not lawfully coexist with it.

It is my understanding of the law that a sentencing judge's power to modify a valid sentence that he has imposed is of limited duration. At common law and in our federal system before the adoption of the present Rules of Criminal Procedure, such power expired with the relevant term of court. United States v. Mayer, 1914, 235 U.S. 55, 35 S. Ct. 16, 59 L. Ed. 129; United States ex rel. Quinn v. Hunter, 7th Cir. 1947, 162 F.2d 644. Now, a district court's power in this regard is defined and limited by the provision of Rule 35 restricting judicial power to reduce a valid sentence to a time "within 120 days after sentence is imposed, or within 120 days after receipt by the court of a mandate issued upon affirmance of the judgment or dismissal of the appeal. * * *"

Thus, 120 days after it received our mandate issued pursuant to our 1968 affirmance of McCullough's conviction, the district court's power to modify the sentence on Count 3 expired. Yet, more than a year later in October 1969, the district court vacated that sentence and this court now sanctions that action.

The dilemma of this case arises because so long as the valid sentence on Count 3 shall remain in force, the sentences on Counts 1 and 2, charging lesser crimes based upon parts of the total transaction covered by the charge in Count 3, are invalid. Heflin v. United States, 1959, 358 U.S. 415, 79 S. Ct. 451, 3 L. Ed. 2d 407. When such a mistake is discovered within 120 days after sentence or within 120 days after the district court has received a mandate of a reviewing court that has sustained the conviction, a comprehensive resentencing is still within judicial control and power of correction. Thus, the district court would have power to vacate the valid but not yet final sentence, as permitted by Rule 35, and then let the otherwise invalid sentence stand.

The same is true when the mistake is discovered on direct appeal. The reviewing court can remand the case for correction of sentence since under Rule 35 judicial power to correct the sentence or sentences has not yet expired. See, for example, United States v. Chester, 3rd Cir. 1969, 407 F.2d 53. But that is not the situation that now confronts us. For, as pointed out earlier in this dissent, the period within which McCullough's valid sentence on Count 3 could be vacated or otherwise changed has long since run.

The error, as I see it, in the majority opinion lies in treating all three sentences as voidable, even after the originally valid sentence, for the offense that included the charged lesser offenses, had become final. I have considered the possibility of avoiding this conclusion by viewing the punishment originally imposed upon McCullough as a single defective sentence, correctible as an entirety in a way that will achieve the result originally intended. But here the indictment charged three distinct crimes in three separate counts and three consecutive sentences have been imposed for these three crimes. To deny that procedural actuality would be to reach for a desirable result through the instrumentality of a palpable fiction. Therefore, I am unable to escape the conclusion that we are confronted with three sentences, one valid and now unchangeable, the other two invalid and insalvageable.

The prisoner is, in my view, entitled to precisely the remedy he is seeking, relief from the two invalid sentences. I do not object to the method of sentencing proposed by the court for the avoidance of the problems of this case in the future.