Defendants, James L. Sarvis and Irie E. Leonard, were tried jointly with Francis A. Salters on a five-count indictment charging them with felony murder,1 first degree murder,2 armed robbery,3 robbery,4 and second degree burglary.5 The jury found Sarvis guilty of armed robbery and second degree burglary. It convicted Leonard of first degree murder, armed robbery, and second degree burglary. Salters was acquitted.

Sarvis and Leonard both appeal. They advance a number of contentions as to why their convictions should be reversed. We conclude that there were three errors of sufficient magnitude to warrant reversal and a new trial. They were: (1) the failure of the district court to instruct the jury, as requested by defendants, to scrutinize carefully the testimony of two important government witnesses to whom immunity had been granted; (2) the district court's limitation of the cross-examination of one of the important government witnesses with reference to two unrelated felony charges pending against him; and (3) the district court's failure to give an immediate cautioning instruction when the testimony of an arresting officer was admitted to impeach Leonard's testimony in his own defense. Although we do not find them to be grounds for reversal, we are constrained to discuss some of the other issues raised by defendants' remaining contentions because of the likelihood that they will arise again on retrial.

While we will state additional facts in the discussion of the contentions to which they are especially pertinent, we begin with a general outline of the case. Benjamin Rudd was brutually robbed and murdered in his basement apartment at 854 21st St., N.E., Washington, D.C. during the evening of September 15-16, 1970. Rudd arrived home in a taxicab about 11:00 p.m. When he left the cab, a group of young men standing across the street called to him. Rudd joined the group for a minute and then crossed the street to his apartment, accompanied by two of them. Six witnesses identified one of the men accompanying Rudd as the defendant Leonard. Three witnesses identified the other as the defendant Sarvis. Rudd apparently felt threatened and sought refuge in the apartment of his neighbor, Virginia Gaskins. Eventually, Rudd was coaxed out into the apartment hallway.

At this point, the two men identified as Sarvis and Leonard assaulted Rudd; Leonard knocked him down, took his keys and wallet, and dragged him into the apartment. Then, about six members of the group began to plunder Rudd's apartment. Among these were Albert Jones and Curtis Hughes, both of whom testified for the government under a grant of immunity, and Francis Salters, the third defendant. Other members of the group, including Larry Mauldin and Robert King, watched from outside the apartment building but did not enter. Both Mauldin and King testified for the government, as did Donald Montgomery, another member of the group who had departed earlier.

While these men rifled Rudd's apartment, Rudd was beaten, tortured, and eventually killed. Although none of the witnesses observed the lethal blow, their testimony pointed to the culpability of Leonard or Sarvis, or both. Jones testified that Leonard told him 'he (Leonard) was going to kill Ben . . .. There was nothing else to do . . . because he had seen . . . his (Leonard's) face and Jimmie's (Sarvis') face.' Hughes testified that he observed Sarvis and Leonard torturing the deceased. Hughes further testified that Leonard told him he was going to kill Rudd, and that he observed Leonard take a butcher knife from the kitchen to the bedroom where Sarvis was harassing the victim with a sculpture knife. Hughes also testified that as Salters left the apartment, Salters said that 'they were going to kill him.'

Sarvis and Leonard based their defense on separate alibis. Leonard testified; Sarvis did not. Salters in effect admitted his presence, but contended that he did not participate in the crimes. He did not testify.

In addition to the highly incriminating testimony of Jones and Hughes, both of whom were obviously accomplices, a substantial portion of the government's evidence came from witnesses who were involved, either as onlookers or participants, with the group of young men whose members allegedly killed the victim and ransacked his apartment.6 During the police investigation, Jones and Hughes first denied knowledge of the crime, but later admitted their complicity and agreed to cooperate. The government granted both immunity pursuant to 18 U.S.C. 6002 (Supp.1973).7 Thus, Jones and Hughes testified both as accomplices and under a grant of immunity.

Neither Sarvis nor Leonard specifically requested the district court to instruct the jury that the testimony of Jones and Hughes should be received with caution because they were accomplices; rather, they requested instructions admonishing the jury to receive this evidence with caution because Jones and Hughes had been granted immunity.8 The district court denied the requested immunity instruction. While it offered to instruct that the testimony of Jones and Hughes, who testified under a grant of immunity, 'is to be scrutinized and evaluated in the same manner as the testimony of all other witnesses who testified during the case,' in fact, the district court gave only the standard instruction as to witnesses' testimony and failed to mention the immune status of Jones and Hughes. Sarvis and Leonard claim error in the failure to give either an accomplice or an immunity instruction.

A. There can be no doubt but that an accomplice instruction would have been proper, because, in Crawford v. United States, 212 U.S. 183, 203-204, 29 S. Ct. 260, 268, 53 L. Ed. 465 (1909), the Supreme Court held that the testimony of a witness who turned 'state's evidence'

is not to be taken as that of an ordinary witness of good character in a case, whose testimony is generally and prima facie supposed to be correct. On the contrary, the evidence of such a witness ought to be received with suspicion, and with the greatest care and caution, and ought not to be passed upon by the jury under the same rules governing other and apparently credible witnesses.

The Court has characterized accomplice testimony as 'inevitably suspect' and unreliable. Bruton v. United States, 391 U.S. 123, 136, 88 S. Ct. 1620, 20 L. Ed. 2d 476 (1968). See On Lee v. United States, 343 U.S. 747, 757, 72 S. Ct. 967, 96 L. Ed. 1270 (1952).

We have no doubt that it is better practice for a district court sua sponte to caution the jury concerning an accomplice's testimony, and that it is reversible error to refuse to give a requested accomplice instruction. Freed v. United States, 49 U.S.App.D.C. 392, 266 F. 1012 (1920). However, neither defendant requested an accomplice instruction; nor did either object to its omission. Rule 30, F.R.Cr.P., bars the claim on appeal unless the district court's failure to give an accomplice instruction was 'plain error' in the circumstances of this case. Rule 52(b), F.R.Cr.P.

That the failure to give an accomplice instruction may, in certain exceptional circumstances, constitute plain error has been recognized among the crcuits. Tillery v. United States, 411 F.2d 644 (5 Cir. 1969); McMillen v. United States, 386 F.2d 29, 35-36 (1 Cir. 1967), cert. denied, 390 U.S. 1031, 88 S. Ct. 1424, 20 L. Ed. 2d 288 (1968).9 But such recognition has not been extended to a case like the instant one where nonaccomplice testimony corroborated the accomplice testimony and was sufficient in itself to serve as a basis for conviction. Thus in Tillery, where the court found plain error, virtually all of the government's case rested upon the accomplice's testimony, the accomplice changed his testimony several times prior to trial to the extent that the court characterized it as 'extremely unreliable, if not incredible and insubstantial' (Id., 411 F.2d at 648), and the case was factually close.10 Similarly, in McMillen, plain error was noticed where the district court instructed as to the now discredited presumption of truthfulness, 'the only evidence connecting . . . (defendant) with the . . . (crime) came from the mouths of others involved in that escapade,' and 'these were not accomplices who had paid their price to society and were free from carrot or stick.' 386 F.2d at 36. See also United States v. Evans, 398 F.2d 159, 163-164 (3 Cir. 1968).

By contrast, in the instant case, nonaccomplice testimony corroborated the accomplice testimony to a significant extent against Leonard, and to a lesser extent against Sarvis. As to Leonard, four other witnesses substantially corroborated the testimony of Jones and Hughes. While three of these witnesses were at least onlookers, they did not participate in the robbery or murder and were not accomplices. Gaskins, the fourth witness, was the victim's neighbor and was in no way connected with the miscreant group. One other witness, King, identified Sarvis. Although King did 'hang around' with the group to watch from the outside of the apartment building, King did not participate in looting Rudd's apartment, so that he, too, was not an accomplice.

Of course, Jones and Hughes were under carrot and stick to some extent and they changed their stories before trial, thereby casting some doubt on their credibility. They did so once from claiming innocence to admitting guilt; but, unlike the accomplice in Tillery who changed his story four times, their behavior was not extraordinary and does not evince the same disposition to prevarication which prompted the court in Tillery to discount heavily the accomplice's testimony. Moreover, the jury was not instructed that they were presumed truthful. Both with regard to Leonard and Sarvis, we conclude that the omission of the accomplice instruction was not plainly erroneous.

B. Defendants' contention that there was error in the refusal to grant a special instruction with regard to the testimony of witnesses to whom the government has granted immunity, stands on a surer footing because the requests were made and refused. Defendants have not cited, nor have we found, any case which holds that it is reversible error to refuse the requested immunity instruction. However, we find persuasively analogous the rule requiring the district court, upon request, to caution the jury that the testimony of paid informers should be received with circumspection. Fletcher v. United States, 81 U.S.App.D.C. 306, 158 F.2d 321 (1946); United States v. Masino, 275 F.2d 129, 133 (2 Cir. 1960). See United States v. Kinnard, 150 U.S.App.D.C. 386, 465 F.2d 566 (1972).

In On Lee, 343 U.S. at 757, 72 S.Ct. at 973, the Court articulated the rationale of the rule that a trial court upon request should instruct caution with regard to the testimony of paid informers:

The use of informers, accessories, accomplices, false friends, or any of the other betrayals which are 'dirty business' may raise serious questions of credibility. To the extent that they do, a defendant is entitled to broad latitude to probe credibility by cross-examination and to have the issues submitted to the jury with careful instructions.

Like the paid informer who has a pecuniary interest in testifying, or an accomplice who has a penal interest in shifting guilt to his cohort, a witness who has been granted immunity has an interest in incriminating another to save himself. Cf. McMillen, supra. Typically, as in this case, the government offers immunity to a lesser or minor co-conspirator, or accomplice, who stands in jeopardy of being convicted, as an incentive to testify against others more significantly involved. A potential witness, so confronted, realizes that he can procure his own freedom by incriminating another, and therein lies his motivation to falsify.11 We hold, therefore, that the district court committed reversible error in refusing to instruct the jury when requested, to consider with caution the testimony of Jones and Hughes, important government witnesses who had been granted immunity from prosecution. We add only that we are not to be understood as saying that, on retrial, if Sarvis and Leonard request an instruction with regard to Jones and Hughes that their testimony is to be viewed with caution both because they were accomplices and because they have been granted immunity from prosecution, they will be entitled to double instructions. Rather, the district court should merge both requests into a single instruction that the testimony of Jones and Hughes is to be scrutinized with caution because they are accomplices and because they have been granted immunity.

IMPEACHMENT-- PENDING FELONY CHARGES.

We find reversible error also in the district court's limitation of the efforts of Sarvis and Leonard to cross-examine Jones concerning two unrelated felony charges which were pending against him.12 Prior to Jones' voir dire testimony, Mr. Nesbitt, the Assistant United States Attorney prosecuting the case, advised the court and defense counsel that a narcotics charge and a robbery charge were currently pending against Jones in the district. Mr. Nesbitt then represented, no doubt in absolute good faith, that:

I have not been in touch with either counsel at any stage of the proceeding (against Jones), your Honor, and have made no representation myself, with regard to disposition of these cases, and to my knowledge, Mr. Kleiboemer (the Assistant United States Attorney prosecuting the other charges) never made any representations regarding dispositions either to the Defendant or to counsel.

In fact, the previous day, another Assistant United States Attorney had advised Jones' counsel that the government would be willing to accept a plea of guilty to one misdemeanor in either case and would ask the court to dismiss the other charges. Jones' counsel immediately advised Jones of the government's offer while Jones was in the witness room waiting to testify for the government in this case.13 Jones testified for the government on voir dire the next day and at trial later in the week. He had previously testified for the government before the grand jury.

At trial, defense counsel attempted to cross-examine Jones by asking whether the government had made any disposition of the unrelated charges pending against him. The district court sustained the government's objection, ruling that the defense had no factual basis for this line of questions. It further instructed the jury to disregard defense counsel's questions because the defense had no proof that any disposition had been made. The district court's ruling effectively foreclosed the defense from unearthing and establishing any factual basis for bias or self-interest arising out of the pending charges. Undoubtedly, the district court's ruling was based on the government's innocent misrepresentation.

It has generally been held that a witness' self-interest or bias 'may be shown . . . in a criminal case when the witness testifies for the state and it is shown that an indictment is pending against him, or that he is an accomplice or co-indictee in the crime on trial.' McCormick, Evidence 40 (Cleary ed. 1972). In Masino, the court reversed where the district court had barred defense cross-examination of two prosecution witnesses concerning the recent dismissal of narcotics charges against them. The court's reasoning warrants quotation, at some length:

When a witness in a criminal case is being questioned as to his possible motives for testifying falsely wide latitude should be allowed in cross-examination . . .. It was highly relevant and material to bring out that the . . . (state criminal) charge . . . had been quashed upon the intercession of the Assistant United States Attorney . . .. This is the kind of situation where the widest possible cross-examination should be permitted. The appellant was entitled to have the jury know what had happened with respect to the charge, including any part which representatives of the government had played, so that the jury could draw its own conclusions with respect to possible motives for (the witness') testimony. The indictment and its disposition was a matter so intimately related to (the witness') possible motives to falsify and his relationship to the government which had called him as its witness that the trial court should have allowed full exploration of these matters on cross-examination.

The permissible scope of exploration on cross-examination is not curtailed by the absence of explicit government promises of leniency, for the defense may attempt to show government 'conduct which might have led a witness to believe that his prospects for lenient treatment by the government depended on the degree of his cooperation.' United States v. Campbell, 426 F.2d 547, 549 (2 Cir. 1970). Jones might reasonably have surmised that there was some connection between his testimony in this case and the government's offer, arriving as it did while he was waiting to testify in this case.15

The defendants were entitled to be given broad latitude in their effort to impeach the credibility of government witnesses on cross-examination. Aford v. United States, 282 U.S. 687, 51 S. Ct. 218, 75 L. Ed. 624 (1931). The district court's ruling and instruction severely circumscribed defense counsel in his regard and therefore was reversible error. Since we reach this conclusion, we need not independently consider the prejudicial effect of the government's erroneous but good faith representations to the court and defense counsel. Compare Giglio v. United States, 405 U.S. 150, 92 S. Ct. 763, 31 L. Ed. 2d 104 (1972) with United States v. Hykel, 461 F.2d 721 (3 Cir. 1972); United States ex rel. Dale v. Williams, 459 F.2d 763, 767 (3 Cir. 1972). See also Brady v. Maryland, 373 U.S. 83, 83 S. Ct. 1194, 10 L. Ed. 2d 215 (1963); Napue v. Illinois, 360 U.S. 264, 79 S. Ct. 1173, 3 L. Ed. 2d 1217 (1959).

IMPEACHMENT-- IMMEDIATE LIMITING INSTRUCTION.

The final error requiring reversal and a new trial arose during Leonard's testimony in his own behalf. On cross-examination, the prosecutor asked Leonard whether he had told the arresting officer that the 'other man'-- Sarvis-- lived across the street. Leonard unequivocally denied making the statement.16 On rebuttal, the government called the arresting officer who was permitted to testify, over Sarvis' objection, that Leonard had made the alleged statements to him. Again over Sarvis' objection, the prosecutor then engaged the arresting officer in a long colloquy detailing the events of Sarvis' arrest.

Sarvis contends that the district court erred in admitting the officer's testimony about Leonard's alleged statement, and that even if the testimony was admissible the district court erred in failing immediately to instruct the jury that it was admissible solely for the purpose of impeaching Leonard's credibility. Sarvis further contends that the remainder of the officer's testimony was improper and irrelevant.17

Defendants' first contention is substantially foreclosed by Nelson v. O'Neil, 402 U.S. 622, 629-630, 91 S. Ct. 1723, 1727, 29 L. Ed. 2d 222 (1971), which held that

where a co-defendant takes the stand in his own defense, denies making an alleged out-of-court statement implicating the defendant, and proceeds to testify favorably to the defendant concerning the underlying facts, the defendant has been denied no rights protected by the Sixth and Fourteenth Amendments.

Since Sarvis could have cross-examined Leonard, the admission of Leonard's and the officer's testimony concerning Leonard's alleged extra-judicial statement does not transgress the confrontation clause. O'Neil, supra. United States v. Hawk Wing, 459 F.2d 428 (8 Cir. 1972). Cf. Bruton, supra.

The officer's rebuttal testimony contradicting Leonard's denial was admissible, however, solely as impeachment evidence. The jury could not consider it as substantive evidence against Leonard, or in any respect whatsoever against Sarvis. Sarvis did not request a limiting instruction immediately after the officer's testimony and the district court did not give one. In its charge, the district court did instruct the jury in general terms as to the limited effect to be given impeachment evidence. After the district court's charge, Sarvis requested an additional instruction advising the jury to apply '(a) voluntary confession or admission' only against the defendant making it, but did not otherwise object. The district court denied the request.

In this circuit, it has been firmly established that when evidence is admitted for the limited purpose of impeaching a witness, it is plain error, in the absence of manifest waiver, for the district court to omit an immediate cautioning instruction. Jones v. United States, 128 U.S.App.D.C. 36, 385 F.2d 296, 300 (1967) (Per Curiam); United States v. Thompson, 150 U.S.App.D.C. 403, 465 F.2d 583, 585 n. 10 (1972). See United States v. McClain, 142 U.S.App.D.C. 213, 440 F.2d 241, 245-246 (1971). See also United States v. Bobbitt, 146 U.S.App.D.C. 224, 450 F.2d 685, 691 (1971). There is no indication in the record that Sarvis or Leonard purposefully waived the instruction. The district court's general instruction on impeachment evidence, buried as it was in a very long charge and coming at the end of a three-week trial, did not adequately protect the interests which gave rise to this rule. The purpose served by a limiting instruction is particularly salutary in a joint trial where, as here, the jury may not consider the evidence, even as restricted, as against one defendant. In Nelson, the court noted that the trial judge had cautioned the jury not to consider as against one defendant the evidence of his co-defendant's prior extrajudicial statement. 402 U.S. at 624, 91 S. Ct. 1723, 29 L. Ed. 2d 222. Hawk Wing was careful to note that after the government's rebuttal witness testified to one defendant's prior inconsistent statement, the trial judge cautioned the jury concerning the impeachment character of the testimony. 459 F.2d at 430.

We therefore conclude that admission of the testimony concerning Leonard's alleged prior inconsistent statement without an immediate instruction limiting its admissibility to impeachment purposes requires reversal of both Sarvis' and Leonard's convictions.

SEVERANCE.

We reject the claim of reversible error in the district court's refusal to grant defendants' repeated motion for severance. Each defendant advances multiple arguments to bolster this contention. After setting out some general considerations, we will discuss them seriatim in the factual context in which each arose.

A. Rule 8(b) of the F.R.Cr.P. authorizes joinder where 'two . . . defendants . . . are alleged to have participated in the same act or transaction . . . constituting an offense . . ..' Joinder of the three defendants was therefore proper. Rule 14 provides the district court with discretion to grant a severance 'if it appears that a defendant . . . is prejudiced by a joinder . . ..' Pursuant to the mandate of Rule 14, 'the trial judge has a continuing duty at all stages of the trial to grant a severance if prejudice does appear.' Schaffer v. United States, 362 U.S. 511, 516, 80 S. Ct. 945, 948, 4 L. Ed. 2d 921 (1960); United States v. Wilson, 140 U.S.App.D.C. 220, 434 F.2d 494, 499 (1970). Since the district court possessed 'great discretion' with respect to the grant or denial of a severance, its decision will be reversed only upon an affirmative showing that the district court clearly abused this discretion and that a defendant was prejudiced thereby. United States v. Gambrill, 146 U.S.App.D.C. 72, 449 F.2d 1148, 1159 (1971); Wilson, 434 F.2d at 500. In reviewing the exercise of the district court's discretion, this court has approvingly concluded that 'the balance has been struck in favor of joint trials . . .,' United States v. Hines, 147 U.S.App.D.C. 249, 455 F.2d 1317, 1334, cert. denied, 406 U.S. 975, 92 S. Ct. 2427, 32 L. Ed. 2d 675 (1972), quoting United States v. Krechevsky, 291 F. Supp. 290, 294 (D.C.1967). This reflects the strong federal policy favoring joinder because it

expedites the administration of justice, reduces the congestion of the trial dockets, conserves judicial time, lessens the burden upon citizens who must sacrifice both time and money to serve upon juries, and avoids the necessity of recalling witnesses who would otherwise be called upon to testify only once.

Hines, 455 F.2d at 1334, quoting Parker v. United States, 404 F.2d 1193, 1196 (9 Cir.1968), cert. denied, 394 U.S. 1004, 89 S. Ct. 1602, 22 L. Ed. 2d 782 (1969). See Rakes v. United States, 169 F.2d 739, 744 (4 Cir.), cert. denied, 335 U.S. 826, 69 S. Ct. 51, 93 L. Ed. 380 (1948).

B. Sarvis contends that he was prejudiced, and severance was required, because the government's case against his co-defendants was more damaging that the case against him. Six witnesses testified that Leonard was one of the men who entered the victim's apartment, while only three-- Jones, Hughes, and King-- similarly placed Sarvis. Although there was some confusion in the testimony of these and other witnesses concerning Sarvis' clothes, all three witnesses firmly and unequivocally identified Sarvis at trial. Sarvis hypothesizes that due to this disparity in the weight of the evidence, the jurors may have permitted the evidence against his co-defendants to could their consideration of his guilt. This court has recognized that 'a disparate quantum of evidence against each of two co-defendants may conceivably require a severance under some circumstances,' but only if the evidence against one is "far more damaging." Gambrill, 449 F.2d at 1159; McHale v. United States, 130 U.S.App.D.C. 163, 398 F.2d 757, cert. denied, 393 U.S. 985, 89 S. Ct. 462, 21 L. Ed. 2d 447 (1968).

Although the evidence placing Leonard at the crime was both quantitatively and qualitatively greater than the evidence against Sarvis, the disparity does not meet the 'far more damaging' test. The evidence against Sarvis was both substantial and compelling. It was in no sense an inconsequential part of the trial, swallowed and tainted by a relatively overwhelming mass of evidence against Leonard. Cf. United States v. Kelly, 349 F.2d 720, 777 (2 Cir.1965), cert. denied, 384 U.S. 947, 86 S. Ct. 1467, 16 L. Ed. 2d 544 (1966). On this record, the jury could 'reasonably be expected to compartmentalize the evidence as it relates to separate defendants.' United States v. DeLarosa, 450 F.2d 1057 (3 Cir.1971), cert. denied sub nom., Jones v. United States, 405 U.S. 957, 92 S. Ct. 1188, 31 L. Ed. 2d 235 (1972). And the district court specifically instructed the jury to determine the guilt or innocence of each defendant by considering only his own conduct and the evidence which applied to him. Finally, the verdicts demonstrate that the jury understood and fulfilled its obligation under the charge.

C. Both Sarvis and Leonard contend that prejudice requiring severance resulted from the apparent conflict between their defenses on the one hand, and Salters' on the other hand, and also between their defenses. Sarvis and Leonard relied on separate alibi defenses. Salters admitted his presence, but denied participation in the killing. Salters did not testify, but prosecution witnesses Hughes and Mauldin were permitted to testify that Salters said at the scene of the crime that Sarvis and Leonard had knocked out the victim, were tearing his apartment apart, and were going to kill him. In his closing argument, counsel for Salters emphasized that Salters did not participate in the killing by often referring to the perpetrators as 'they,' with the implication, according to Sarvis and Leonard, that the pronoun 'they' pointed Salters' accusing finger at them. Finally, Leonard urges that Sarvis' efforts to discredit the testimony which placed him with Leonard had the effect of undermining Leonard's alibi defense.

Defendants rely heavily on Rhone v. United States, 125 U.S.App.D.C. 47, 365 F.2d 980, 981 (1966), where this court recognized that:

Prejudice from joinder of defendants may arise . . . where the defendants present conflicting and irreconcilable defenses and there is a danger that the jury will unjustifiably infer that this conflict alone demonstrates that both are guilty.

See DeLuna v. United States, 308 F.2d 140 (5 Cir.1962). Subsequent cases stress the word 'alone,' and conclude that where independent evidence of each defendant's guilt supports the jury's verdict, the conflict does not necessarily prejudice the defendants to the extent that a district court's refusal to sever must be considered a clear abuse of discretion. United States v. Hurt, 155 U.S.App.D.C. 217, 476 F.2d 1164 (1973); Wilson, 434 F.2d at 500; United States v. Robinson, 139 U.S.App.D.C. 286, 432 F.2d 1348 (1970). See Hines, 455 F.2d at 1333-1335. But cf. ABA Project on Minimum Standards for Criminal Justice, Joinder and Severance, 2.3(b) at 41 (1967). The spectacle of conflicting stories and veiled accusations emanating from the defense table no doubt adversely affected defendants' defenses. However, the alibis of Sarvis and Leonard were not contradicted solely by Salters' defense. The government presented substantial independent evidence incriminating Sarvis and Leonard. Moreover, much of the government's evidence came from other members of the group who, like Salters, were present during the commission of the crimes. Hughes, Jones, Mauldin and King all testified that they were present and observed some of the events. The testimony of these witnesses diminished the impact of Salters' defense which derived significance mainly from the implication that he was present and observed Sarvis and Leonard. Mauldin and Hughes did testify to statements by Salters which incriminated Sarvis and Leonard. And for obvious reasons, counsel for Salters did not join the vigorous objection of Sarvis and Leonard to his hearsay testimony. But, as will be shown, this hearsay testimony would have been admissible even if Sarvis and Leonard had been tried separately. See Hurt, supra; Wilson, 434 F.2d at 502.

Neither Sarvis nor Leonard objected to the closing argument by counsel for Salters. He did not refer to the other defendants by name. With the evidence establishing that at least six men were in the group which ransacked the victim's apartment we cannot say that Salters' counsel's unvariegated emphasis of the word 'they' in the closing argument constituted plainly prejudicial error. Rule 52(b), F.R.Cr.P. Thus, it cannot be said that the conflict between the defendants' alibi defense and Salters' defense created a serious danger that the jury would seize upon it alone as the dispositive indicia of guilt.

The danger that Salters' conflicting defense might have seriously prejudiced Sarvis and Leonard was, however, exacerbated by Salters' failure to testify and the consequent inability of Sarvis and Leonard to cross-examine him. In Hurt and Wilson, the court based its reasoning to some extent on the fact that the defendants whose defenses conflicted all testified and were subject to cross-examination.

In Gambrill, one defendant in a joint trial proffered two witnesses whose testimony provided alibis for both the defendant and his co-defendant. The co-defendant, wanting to disassociate himself from what he considered to be a patently incredible alibi, sought and obtained an instruction from the court which limited the alibi testimony to the defendant who proffered the witnesses. This instruction had the effect of implicitly impugning the proffering defendant's alibi, an effect he could not dispel because the co-defendant's failure to testify precluded cross-examination. Although the court reversed on another ground, it suggested that if a similar problem arose on remand, tne district court should grant a severance. 449 F.2d at 1161-1162.

Even though a district court might properly exercise its discretion in these circumstances to grant a severance, the refusal to do so here was not a clear abuse of discretion. The damaging implications of Salters' defense were corroborated by substantial testimony from several similarly situated witnesses. The inability of Sarvis and Leonard to cross-examine Salters thus did not create such prejudice that the district court's refusal to grant their motions for a severance constituted a clear abuse of discretion. 'That a defendant might have a better chance of acquittal if tried separately does not establish his right to severance.' Wilson, 434 F.2d at 501.18

D. Witnesses Hughes and Mauldin were permitted to testify over the objections of Sarvis and Leonard to statements made by Salters during the commission of the crimes which incriminated Sarvis and Leonard.19 The district court apparently admitted Mauldin's testimony only for the purpose of proving that Salters made the statement and not to prove the truth of Salters' statement, although its ruling is not without confusion and it gave no explicit instructions on this point. The district court apparently admitted Hughes' testimony under the rather nebulous exception termed 'res gestae.'

The exception to the hearsay rule for contemporaneous declarations which partake of the event is often subsumed under the rubric of res gestae, and is an increasingly recognized exception. McCormick, supra, 298. See Model Code of Evidence Rule 512(a); Proposed Federal Rule of Evidence 803(2). Similarly, res gestae often is used to refer to another generally recognized exception to the hearsay rule for spontaneous declarations or excited utterances. Murphy Auto Parts Co. v. Ball, 101 U.S.App.D.C. 416, 249 F.2d 508 (1957), cert. denied, 355 U.S. 932, 78 S. Ct. 413, 2 L. Ed. 2d 415 (1958). See McCormick, supra, 297; 6 Wigmore, supra, 1749-1757, 1768. The testimony of Mauldin and Hughes may properly have been admitted under either of these exceptions.

Sarvis and Leonard contend that since they courd not cross-examine their co-defendant Salters, admission of hearsay testimony of his statements violated their sixth amendment right of confrontation and required a severance. They rely primarily on Bruton v. United States, supra, which held that in a joint trial, where the confession of a defendant was admitted only against him and he did not testify, his co-defendant's constitutional rights were not sufficiently protected by a cautionary instruction. The Court reasoned:

because of the substantial risk that the jury, despite instructions to the contrary, looked to the incriminating extrajudicial statements in determining petitioner's guilt, admission of . . . (the) confession in this joint trial violated petitioner's right of cross-examination secured by the Confrontation Clause of the Sixth Amendment.

391 U.S. at 126, 88 S. Ct. at 1622. The Court focused on the dangers raised

where the powerfully incriminating extrajudicial statements of a codefendant, who stands accused side-by-side with the defendant, are deliberately spread before the jury in a joint trial. Not only are the incriminations devastating to the defendant but their credibility is inevitably suspect, . . . The unreliability of such evidence is intolerably compounded when the alleged accomplice, as here, does not testify and cannot be tested by cross-examination.

Id., at 135-136, 88 S.Ct. at 1628.

Salters' statements were incriminating, and he availed himself of his fifth amendment right not to testify. But, unlike the confession in Bruton, Salters' statements were admissible against all of the defendants under a generally recognized exception to the hearsay rule. Significantly, Bruton expressly reserved this issue:

There is not before us, therefore, any recognized exception to the hearsay rule insofar as petitioner is concerned and we intimate no view whatever that such exceptions necessarily raise questions under the Confrontation Clause.

Id., at 128 n. 3, 88 S. Ct. at 1624. Thus, Bruton does not alone resolve this issue.

In California v. Green, 399 U.S. 149, 90 S. Ct. 1930, 26 L. Ed. 2d 489 (1970), the Court addressed the constitutional issue presented by the admission of hearsay in a criminal trial, and sustained a state statute which permitted the introduction of prior inconsistent extrajudicial statements as substantive evidence in a criminal trial. The Court explained:

While it may readily be conceded that hearsay rules and the Confrontation Clause are generally designed to protect similar values, it is quite a different thing to suggest that the overlap is complete and that the Confrontation Clause is nothing more or less than a codification of the rules of hearsay . . .. We have more than once found a violation of confrontation values even though the statements in issue were admitted under an arguably recognized hearsay exception . . .. The converse is equally true: Merely because evidence is admitted in violation of a long-established hearsay rule does not lead to the automatic conclusion that confrontation rights have been denied.

Id., at 155-156, 90 S.Ct. at 1933.

The Supreme Court applied this emerging doctrine in Dutton v. Evans, 400 U.S. 74, 91 S. Ct. 210, 27 L. Ed. 2d 213 (1970) (plurality opinion). In Dutton, hearsay evidence of the statement of a co-conspirator which incriminated the defendant was admitted and the co-conspirator did not testify. The testimony was admitted pursuant to a state hearsay exception for statements by co-conspirators made after the termination of a conspiracy. Dutton concluded that the application of this exception 'in the circumstances of this case did not violate the Constitution.' Id., at 88, 91 S.Ct. at 219. The Court distinguished bruton on three grounds: (1) Bruton involved a joint trial while the defendant in Dutton was tried alone; (2) Bruton involved the ability of jurors to consider the hearsay only against the confessing defendant; and (3) Bruton involved a confession wholly inadmissible against one defendant while the hearsay in Dutton was admissible as to the defendant. The Court discussed a plethora of factors which impelled the conclusion that admission of the accomplice's hearsay statement did not violate the confrontation clause: (1) the evidence was not 'crucial' or 'devastating,' but was peripheral; (2) the defendant was not deprived of the right of confrontation on the issue of whether the co-conspirator actually made the statement; (3) the statement contained no express assertion about past fact; (4) the co-conspirator had personal knowledge of the identity and role of the other participants; (5) the possibility that the statement was founded on faulty recollection was remote; (6) the statement was spontaneous; and (7) it was against penal interest. The opinion concluded that:

These are indicia of reliability which have been widely viewed as determinative of whether a statement may be placed before the jury though there is no confrontation of the declarant.

Although many of the 'indicia of reliability' in Dutton and the instant case are similar, it is readily apparent that the cases differ in three significant aspects: the instand case was a joint trial, the contested evidence was important, and the hearsay declarant was an accomplice. We conclude, however, that for several reasons these factors do not sufficiently aggravate the circumstances to compel the conclusion that the confrontation clause required severance in this case: The disputed testimony was admissible under a generally recognized hearsay exception as against all of the defendants, and would have been admissible against each of them had they been tried separately. Even if the defendants had been tried separately, Salters most probably would have invoked the fifth amendment. See Gorin v. United States, 313 F.2d 641, 645-646 (1 Cir.), cert. denied, 374 U.S. 829, 83 S. Ct. 1870, 10 L. Ed. 2d 1052 (1963). Therefore, little can be made of the theoretical opportunity to cross-examine him if the case had been severed. Thus, the setting of a joint trial may have heightened the impact of the statements, but it in no way affected the 'indicia of reliability' which Dutton considered most relevant.

Although the statements undoubtedly had dramatic impact, they were not 'crucial' in the sense that they dispositively bolstered an othewise weak case. The prosecution presented substantial eyewitness testimony which incriminated Sarvis and Leonard, which the hearsay statements supplemented.

Since Salters was an accomplice, the reliability of his hearsay declaration was 'inevitably suspect.' Bruton, 391 U.S. at 135-136, 88 S. Ct. 1620. Thus, these hearsay declarations were surrounded by less compelling indicia of reliability than those admitted in Dutton. Had these declarations been a singular or critical element in sustaining the government's burden of proof, their admission in a trial where the defendant could not cross-examine the declarant would raise a serious confrontation clause issue. But in this case, their reliability was bolstered and their importance diminished by the substantial and corroborative eyewitness testimony presented. In the circumstances of this joint trial, admission of Salters' hearsay declarations did not violate the confrontation clause. We therefore conclude that under Dutton, the confrontation clause did not require a severance in this case, even though a severance may not have been improper.

IMPEACHMENT-- DRUG USE.

Sarvis and Leonard further contend that the district court erred in restricting their cross-examination of Curtis Hughes concerning his use of heroin, in forbidding them to rebut Hughes' testimony on cross-examination with direct extrinsic evidence from his father, and in refusing to allow expert medical testimony regarding the general effects of habitual drug use on perception. We disagree.

On voir dire, Hughes admitted that he took heroin about once or twice a week, and that this use had continued until the beginning of trial. When asked whether he had taken heroin on the day of the crimes, he replied, 'I can't recall.' At trial defense counsel cross-examined Hughes concerning his drug habit. Hughes' testimony differed somewhat from his voir dire testimony. He testified that he had not taken drugs on the day he observed the crime. He further testified that he had not taken drugs for several months. When defense counsel then attempted to question Hughes concerning the effects of his drug use, the district court sustained the government's objection, ruling that defense counsel could not pursue that line of questioning unless he first established that Hughes had used drugs on the day he observed the crimes. The further rulings referred to above ensued.

The authorities are in disarray concerning the propriety of impeaching a witness' credibility by showing drug use:

as to drug addiction to which more social odium has been attached, many decisions allow it to be shown to impeach, even without evidence that it did in the particular case affect truth-telling, although apparently more courts, absent a particular showing of effect on the witness's veracity, would exclude ti . . .. The excluding courts seem to have the better of the arguments. It can scarcely be contended that there is enough scientific agreement to warrant judicial notice that addiction in and of itself usually affects credibility. Certainly it is pregnant with prejudice. On the other hand, there is an increasing recognition among non-legal authorities that addiction may in various instances be linked with personality and other defects which do bear upon credibility.

McCormick, supra, 44. See Kelly v. Maryland Casualty Co., 45 F.2d 782 (W.D.Va.1929), aff'd, 45 F.2d 788 (4 Cir. 1930); 3 A Wigmore, supra, 934.

This circuit has recognized the propriety of developing the matter of drug addiction for the purpose of attacking, not the addict's general credibility,21 but his ability and capacity to observe the events in question. United States v. Kearney, 136 U.S.App.D.C. 328, 420 F.2d 170, 173-174 (1969); United States v. Fowler, 151 U.S.App.D.C. 79, 465 F.2d 664 (1972). See Wilson v. United States, 232 U.S. 563, 34 S. Ct. 347, 58 L. Ed. 728 (1914). In Kearney, the court affirmed where the district court had deferred cross-examination on this issue pending establishment of a foundation. The court concluded:

A judge may not fairly block all probing of an issue like narcotics use by the prosecution's sole eyewitness to a murder. On the other hadn, . . . the matter of drug addiction, which involves social transgression and the possibility of illegal conduct, is properly approached with awarencess of the potential for prejudice of the jury. In some cases the desirable procedure may lie in permitting defense counsel to establish his foundation by questioning the witness out of the hearing of the jury . . ..

420 F.2d 174. In Fowler, the court reversed a conviction because the district judge had blocked all inquiry into the witness's drug habit. The court declared that the defendant has a 'right to cross-examine . . . as to whether he (the witness) was using narcotics at the time he observed appellant commit the alleged offense.' 465 F.2d 665. Implicit in this holding is the conclusion that this finding is a sine qua non to further cross-examination.

The district court's ruling on the scope of Hughes' cross-examination is consistent with the thrust of these cases. It did not prohibit all cross-examination concerning Hughes' drug habit, but allowed defense counsel the opportunity to establish a foundation consisting of evidence that Hughes had used narcotics on the day he observed the events. Defense counsel was unable to establish this foundation on cross-examination and therefore had no evidentiary ground to sustain the contention that drugs interfered with Hughes' capacity to observe the events.

Perhaps Hughes' responses concerning his onw drug habit were suspect,22 and contradicted to some extent his voir dire testimony. Nevertheless, Hughes' drug habit was plainly a collateral issue. We would hesitate to establish a rule which would encourage sub-trials on the highly sensitive issue of narcotics use before the jury.23 Therefore, we conclude that the trial judge's ruling barring defense counsel's proffer of extrinsic evidence to contradict Hughes' testimony on cross-examination was correct. See McCormick, supra, 47; 3A Wigmore, supra, 1000-1007.

Finally, we recognize the desirability of increasing judicial appreciation of the general effects of drug use on perception, memory, and credibility and the knowledge that expert testimony may bring to this troublesome issue. However, we cannot conclude that the district court was in error when it excluded the proffer of such evidence, given the state of the record.

IMMUNITY.

At the government's request, the district court granted immunity to witnesses Jones and Hughes pursuant to 18 U.S.C. 6002 (Supp.1973).24 This statute authorized the court to grant so-called 'use-restriction' immunity.25 Because of an error in the order prepared for the district court's signature, it erroneously purported to grant transactional immunity26 to Jones and Hughes. The government concedes that the district court inadvertently exceeded its statutory authority. Neither defendant objected at trial, Rule 52(b), F.R.Cr.P., but each claims reversible error in what occurred.

Sarvis and Leonard contend that they were prejudicially injured by this excessive grant of immunity, since, they submit, Jones and Hughes might not have testified if they had known that they could legally have the benefit of only use-restriction immunity. For this proposition and for the proposition that they have standing to raise the issue, defendants rely on Ellis v. United States, 135 U.S.App.D.C. 35, 416 F.2d 791 (1969).

We think that defendants' reliance on Ellis is misplaced. In Ellis, the district court exceeded its judicial powers by granting immunity where it could under law granting immunity where it could under stances, it was possible that the witness may never have chosen to testify. But here, Jones and Hughes would have been compelled to testify fully, even if the distrivy court had properly restricted its grant to use-restriction immunity. Their only alternative was a contempt citation. Jones and Hughes at no time had an option not to testify.

Defendants further aruge that since Kastigar v. United States, 406 U.S. 441, 92 S. Ct. 1653, 32 L. Ed. 2d 212 (1972) had not been decided at the time of trial, Jones and Hughes might have thought that mere use immunity was unconstitutional, and would therefore not have testified if they had been granted only use immunity. First, no evidence in the record supports this imaginative supposition. Secondly, we cannot discern a grave potential for prejudice or injustice in the proceedings as they transpired. The district court's erroneous grant of excessive immunity was not plain error which prejudicially impinged upon substantial rights. Rule 52(b), F.R.Cr.P. Before retrial, we assume that the district court will correct its order granting immunity so as to obviate a repetition of this contention.

Reversed; new trial granted.

BAZELON, Chief Judge (concurring in part and dissenting in part):

Defendants Leonard and Sarvis were charged with having participated in a particularly brutal crime-- the robbery, torture and murder of Benjamin Rudd. At the conclusion of their joint trial, the jury found both defendants guilty of armed robbery and second-degree burglary. Sarvis was acquitted of the murder counts; Leonard was convicted of first-degree murder. A third defendant, one Francis Salters, tried together with Leonard and Sarvis on all counts except that charging first-degree murder, was acquitted.

I join the Court's decision to reverse the defendants' convictions. I agree that the trial court erred in failing to instruct the jury to receive with caution the testimony of witnesses to whom the Government had granted immunity, and in refusing to allow defense counsel to cross-examine an important Government witness concerning felony charges then pending against him. In addition, I concur in the Court's conclusion that, in the circumstances here, the trial court did not err in its several rulings on narcotics addiction impeachment.1

On two points, however, I part company with the majority. First, I cannot agree with what I see as a misapplication of plain error and harmless error standards in the Court's treatment of the accomplice instruction issue. I assume that, on retrial, the defense will request and the trial court will give the hybrid accomplice-- immunity instruction suggested by the majority.2 I deal, therefore, with the accomplice instruction issue, not because it promises to affect the retrial of these defendants, but because of the import of the plain error-harmless error dichotomy both to the proper exercise of our appellate function and to the definition of the trial court's duty to further the fairness and accuracy of the guilt-determining process.

My second concern here in the complex of problems raised by the trial court's repeated denial of defendant Sarvis' motions to sever his trial from that of defendant Leonard, and of motions by both Leonard and Sarvis for the separate trial of their co-defendant Salters. The denial of these motions implicates not only the scope of the trial court's discretion under Rule 14,3 but also both the rules governing admission of hearsay evidence and constitutional rights of confrontation and, arguably, compulsory process. I have grave doubt that the majority has correctly applied to the situation here the body of doctrine revealed, however obscurely, in the Supreme Court's Confrontation Clause cases. But even if I am wrong on this point, I think the majority mistaken in supposing, as it seems to do, that its Confrontation Clause Discussion concludes the issues raised by these appellants under the rubric of severance. I had not conceived that a trial court's actions under Rule 14, and in admitting evidence must rise to the dimension of constitutional errors before they may require reversal.

Moreover, the majority's analysis of the severance issues seems to rest in major part on the standard notion that joinder advances the efficiency of the criminal process. Whatever one may think of the proposition that procedural fairness in criminal trials may or should be sacrificed, even at the margin, to judicial efficiency, the notion that joinder does in fact promote efficiency is, I am convinced, at least questionable in the general run of cases and demonstrably wrong in the cases before us. As explained in Section II, infra, I would hold that the trial court here abused its discretion in refusing either to grant the defendants' motions for severance or, in the alternative, to exclude from evidence the hearsay statements of co-defendants Leonard and Salters.

A. In Section II-A of its opinion, the majority states that 'it is better practice for a district court sua sponte to caution the jury concerning an accomplice's testimony . . ..'4 The majority further states that Hughes and Jones, critical witnesses against the appellants, 'were obviously accomplices . . ..'5 Nonetheless, it goes on to hold that the trial court's failure to give, sua sponte, an accomplice instriction with respect to these witnesses' testimony-- testimony that the majority terms 'highly incriminating'-- was, as to both Leonard and Sarvis, not plainly erroneous.6 In Section II-B, however, the majority holds that the trial court's refusal to give a cautionary instruction concerning the testimony of these same witnesses, based on the fact that each had been granted immunity to elicit his testimony, is reversible error requiring a new trial.

The divergence between the majority's reaction to the accomplice instruction issue and its reaction to the use-immunity instruction issue is not, of course, the product of the substance of the issues themselves. On the contrary, the two issues are, in many cases, two sides of the same coin, since use immunity will most often be granted, as it was here, to accomplices or co-conspirators. In such cases, both instructions are aimed at substantially the same problem: the 'serious questions of credibility' raised by testimony of 'informers, accessories, accomplices, false friends.'7 Accomplices, whether testifying under a grant of use immunity or not, may have every reason to sanitize their own roles in the events at issue and to exaggerate the culpability of the persons being tried, in part perhaps to depict themselves as penitent and small fish, unworthy of the prosecutor's attention, in part to win the prosecutor's favor and, perhaps, to secure his tacit or overt agreement to forego their prosecution.8 Moreover, since his version of events is consistent with the Government's view of the case and often only the defendant can give it the lie, the testifying accomplice may reasonably suppose that he need little fear the consequences of perjury.

These grave dangers inherent in the testimony of accomplices do not render them incompetent as witnesses. Any such rule would, as a practical matter, make many prosecutions impossible. Instead, in an effort to prevent these dangers from affecting the accuracy of the jury's verdict, we rely on instructions to the jury segregating accomplices' testimony from that of witnesses of 'good character . . . whose testimony is generally and prima facie supposed to be correct;'9 accomplices' testimony, in contrast, 'ought to be received with suspicion, and with the very greatest care and caution . . ..'10

The majority recognizes the substantial identity between the accomplice and immunity instruction issues in this case in its statement that, on retrial, the two instructions should be merged into one. Its divergent treatment of the two issues rests, therefore, not on substance but on the fact that an immunity instruction was requested by defense counsel; an accomplice instruction was not. The Court's analysis is premised on the harmless error-plain error distinction set forth in F.R.Crim.P. 52. If an objection has been made and erroneously overruled, then the error is reversible and requires a new trial, unless, under the harmless error rule, the reviewing court can say with some certainty that the error cannot have prejudiced the defendant.11 If a rule of evidence or some other rule was violated but no objection was made, there was error; but the error does not require reversal unless it was 'plain error'-- unless, as the usual formulation runs, the error was so prejudicial that reversal is required to avoid 'a manifest miscarriage of justice.'12 Adhering to these principles in this case, the majority concludes that the testimony of Hughes and Jones was not so lacking in weight as to make failure to give an immunity instruction harmless; but at the same time, it does not deem this testimony so significant that failure to give an accomplice instruction was plain error.

I do not contend that the majority miscomprehends the standard principles of harmless and plain error, nor that these principles are inconsistent with the language of Rule 52. I do contend, however, that as usually construed and applied, Rule 52's dichotomous standards blink the realities of the contemporary criminal justice system and tolerate an excessively great margin of error in criminal convictions.

The purpose of the dual standards of review is, as has often been said, to enforce the general requirement that defense counsel timely object to errors, thus affording the trial court an opportunity to avoid or correct them and preventing the costs, monetary and otherwise, occasioned by reversal.13 The problem, of course, is that the prejudice caused a defendant by error does not somehow evaporate or diminish simply because his counsel has failed to object. If, for example, the premises of the accomplice instruction rule are correct-- if it is true that without the instruction, the chances are significantly greater that the jury will give undue credence to accomplice testimony and therefore return what may be an erroneous verdict-- then these dangers must exist in precisely equal measure whether defendant's counsel has requested the instruction or not. In short, the defendant's guilt may be decided, not alone by the facts and law of his case, but, in significant measure, by the errors and omissions of his counsel.

There may once have been some justification, although I think it was never strong, for compelling defendants to bear the consequences of their counsel's mistakes. But in a system in which the great majority of defendants are represented by counsel selected and funded, not by the defendants themselves, but through the courts,14 this justification no longer exists. On the contrary, I think the harmless error-plain error distinction, as normally applied, is a classic example of a misdirected deterrent.

To a considerable extent perhaps, the unfairmess of holding the defendant responsible for the omissions and inadvertences of his counsel can be remedied by establishing standards of adequate assistance and enforcing the standards by reversal.15 But such standards alone are not enough. There are too many cases, like this case, in which omissions of counsel are not so great as to justify a finding of incompetency under any meaningful standard, and yet are sufficiently significant to have affected the jury's verdict. Nor can the problem be met by a greater willingness to find plain error when defendants have been represented by appointed counsel, rather than counsel of their own choice.16 Given the 'inescapable remoteness of appellate review,'17 it is, in my view, impossible to estimate with any precision the quantum of prejudice suffered by a defendant, at least so long as this quantum is above the harmless error floor; in any event, our estimate cannot logically depend on whether, or how much, the defendant has paid for his counsel.

I do not mean to denigrate in any way the vital importance of the adversary system to the guilt-determining process. I am convinced, however, that the time has come formally to abandon the game theory of the criminal process-- the notion, in Mr. Justice Brennan's terms, that criminal trials are 'sporting events' matching the wits of prosecution and defense, rather than quests for truth concerning the guilt of the accused.18 Responsibility for the fairness of the process, for the reasonable accuracy of jury verdicts, rests, and should rest, not on the defense alone, but also on the trial court and, indeed, the Government.19 I think the plain error rule should be invoked, not merely to avoid gross miscarriages of justice (assuming that we can recognize such occasions when we them), but as a means of defining and, if need be, enforcing the district court's independent duty to take steps necessary to advance the fairness and accuracy of the trials that take place before it.

In some measure, of course, this is already the rule's province. It has almost universally been held that the trial court's failure to instruct the jury as to all necessary elements of the charged offense is reversible error, despite failure of counsel to object.20 So also this Court has held that 'whenever evidence is admitted only for a limited purpose, it is plain error, in the absence of manifest waiver, to omit an immediate cautioning instruction.'21 Most recently, in United States v. Kinnard, I expressed my view that it is plain error for the trial court to fail to instruct the jury 'to weigh with extreme caution the testimony of an addict-informer . . . because of the possibility of the addict's special interest and motive to fabricate.'22

In this case, I would hold that it was plain error-- error 'affecting substantial rights'-- for the trial court to fail to instruct the jury concerning the dangers of accomplice testimony. The factors that lead me to include the accomplice instruction within the independent duty of the trial court are the same as those present in McClain and Kinnard. First, it must remain the responsibility of defense counsel to establish the factual predicate making such an instruction appropriate. But once that factual predicate becomes clear, whether through defense efforts or otherwise, 'the judge is put on notice that an instruction on credibility should be submitted to counsel as part of the routine set of instructions which the trial court offers.'23 Second, the appropriateness of an accomplice instruction, like that of the addict-informer instruction of Kinnard and the limited-purpose evidence instruction of McClain, does not generally relate to a defendant's particular theory of a case, nor does it depend on any novel proposition of law. When accomplice testimony is presented, 'the trial court should be aware, without prompting of counsel, of the special dangers posed to the fairness of the trial,'24 dangers that, as noted above, have long been recognized. If, because of the peculiarities of the case, defense counsel discerns prejudice from the instruction, he can, of course, request that it not be given.

It may be objected that acceptance of the analysis suggested above would not only undermine Rule 30's requirement that instructions be requested, but also would give positive encouragement to counsel to court errors, since errors, although not objected to, may well lead to reversal. But I doubt that this would be the case. On the contrary, I think that this use of the plain error rule would lead to fewer rather than more errors noted on appeal. Rather than rely on the defense interest alone, it gives the trial court and, perhaps more importantly, the Government a distinct interest in abiding rules designed to advance the accuracy of the criminal process. And to the extent that this is true, it forwards not only the accuracy of the process, but its efficiency as well.

This does not mean, however, that I think failure to give an accomplice instruction, even when fully warranted by the facts of the case, must inevitably lead to reversal. There remains room for application of the harmless error rule, if the error goes only to evidence relevant to a point peripheral to the central determination of guilt or evidence that is corroborated in every material respect.25 This rule, however, is itself open to misuse26 and must be cautiously applied in criminal cases.27 In any event, the rule has no proper application here. Whatever else can be said of the testimony of Hughes and Jones, I agree with the majority that errors affecting the credence given it by the jury cannot be deemed harmless.

B. I admit that my analysis of the plain error rule does not comport with its standard interpretation; indeed, in advancing it, I may be in a minority of one, and that, as Judge L. Hand once remarked, 'is rather too slim a party to carry a banner.'28 But even under the standards the majority has applied, I disagree with the result it has reached on the accomplice instruction issue insofar as that issue applies to defendant Sarvis.

In Tillery v. United States, the Fifth Circuit held that failure to give an accomplice instruction may be plain error when accomplice testimony is 'so critical that the traditional caveat as to its evaluation and use was an indispensable part of the court's charge.'29 The majority distinguishes this case from Tillery on the ground, inter alia, that there, 'virtually all of the government's case rested upon the accomplice's testimony,' while here 'nonaccomplice testimony corroborated the accomplice testimony to a significant extent against Leonard, and to a lesser extent against Sarvis.'30

I think this conclusion correct as to defendant Leonard. Four witnesses other than the accomplices Hughes and Jones placed Leonard at the scene on the night of the murder. Two of these witnesses, King and Mauldin, testified that they saw Leonard accompany Rudd to his apartment building.31 Witness Virginia Gaskins testified that Leonard was waiting in the hallway of the building when Rudd left her apartment.32 Witness King stated that, upon looking through the window of the front door of the building, he observed Leonard 'holding Mr. Rudd up,'33 an occurrence corroborated by the testimony of witness Mauldin,34 and then enter Mr. Rudd's apartment.35 Witnesses King and Mauldin further testified that, subsequently, Leonard emerged from the apartment,36 at which time, according to King, he in effect announced to the assemblage his intent to kill Mr. Rudd in order to prevent him from identifying Leonard and the others to the police.37 Given this evidence, I agree with the majority that the Hughes and Jones testimony was not so critical to a finding of Leonard's guilt as to justify a holding of plain error under the Tillery standard.

But I do not see how this conclusion can rightly be extended to defendant Sarvis. The only admissible evidence, other than the testimony of Hughes and Jones, that even arguably implicated Sarvis in any aspect of the crime was the testimony of witness King. Sarvis appears, however, only in three fleeting references in the course of King's direct testimony: King identified Sarvis as having been among the group of young men standing across the street from the apartment building when Rudd first arrived;38 King stated that he saw Sarvis come out of the apartment;39 finally King observed Sarvis standing at the bottom of the steps in the hallway of the building and subsequently re-enter Rudd's apartment.40 On redirect, however, King said that 'some of my thoughts were getting mixed up,' and that Sarvis, not Salters, was in the hallway when Leonard came to the apartment door.41 These wisps of evidence can be taken as indicating, at most, Sarvis' presence in the building and, perhaps, in the victim's apartment on the night of the crime. Had the Government's case against Sarvis rested on this evidence alone, I have grave doubt that it could properly have survived a motion for judgment of acquittal, either on the armed robbery count or on the counts charging murder.

In contrast, the testimony of Hughes and Jones directly implicated Sarvis in the crimes against Mr. Rudd. Jones testified that he saw Sarvis, together with Leonard, rob Rudd in the hallway of the apartment house.42 Both Hughes and Jones stated that they later saw Sarvis 'standing over Rudd' in the bedroom of Rudd's apartment;43 Hughes observed Sarvis 'cutting' Rudd with a knife.44

I cannot agree that this highly incriminating testimony was, in any meaningful sense of the term, 'substantially corroborated' by the testimony of witness King. On the contrary, I am convinced by the record that the jury's verdict against Sarvis on the armed robbery count rested-- and, in law, can only have rested45 -- on the testimony of the accomplices, with only marginal support, if any, from that of King. The remainder of the witnesses in the case were either uncertain that any second man participated directly with Leonard or were entirely unable to say who that second man might have been. Moreover, Sarvis presented an alibi and mistaken identity defense, supported by three witnesses, that, although impeached to some extent, was not inherently unbelievable. Therefore, contrary to the majority, I conclude that as to Sarvis the Tillery conditions for finding plain error are clearly present.

A. I have no doubt that defendant Sarvis was prejudiced by joinder-- prejudiced to a degree sufficient to compel the District Court to exercise its power and responsibility under Rule 1446 to sever Sarvis' trial from that of Leonard. This prejudice did not derive simply from the disparity in the quality and quantity of evidence adduced against Leonard, on the one hand, and Sarvis, on the other. As the majority acknowledges and as I have attempted to demonstrate above, such a disparity certainly existed in this case. But I agree with the majority that, by itself, the disparity would not have been so great as to warrant a conclusion that the jury may have confounded the evidence and therefore may have drawn support for its verdict against Sarvis from evidence relevant to Leonard alone.47

The problem here is of a different but no less serious variety. It has long been recognized that, to a greater or lesser degree, there inheres in every joint trial a danger that the jury will disregard the bounds of admissibility, however thoroughly and emphatically those bounds are explained, and will transfer the effect of evidence admissible and admitted only as to one codefendant into the case against another.48 Here I think the conclusion inescapable that the Government, whether intentionally or not, invited the jury to engage in exactly this sort of impermissible transference, encouraging it to conclude that if Leonard was guilty, then Sarvis must be guilty as well.

When Leonard testified in his own behalf, the Government posed a series of questions on cross-examination to establish that, at the time of his arrest, Sarvis lived across the street from Leonard. Following these questions, which were unrelated to anything to which Leonard had testified on direct, the Government asked whether, when he was arrested, Leonard had said to the police: 'Aren't you going to get the other man . . . the man across the street?' Leonard replied that he had not. The Government repeated the question, to the same response.49 In rebuttal, the Government called one of the arresting officiers, who testified that Leonard had indeed made the statement attributed to him on cross-examination. The officer when on to testify, in response to questions beyond the permissible scope of rebuttal, that following Leonard's arrest, the police had crossed the street to arrest 'the other man'-- Sarvis.50 All of this occurred over the strenuous and repeated objections of Sarvis' counsel.

To the extent that this evidence was admissible at all-- and some of it clearly was not-- it was admissible only for the purpose of impeaching, by means of a prior inconsistent statement, Leonard's testimony in his own behalf. It was, as the majority grants, entirely inadmissible in the case against Sarvis and, had he been tried separately, could not properly have come before the jury.

The rules governing admissibility of evidence have, of course, some purpose. Premised as they are on a general calculus of probative value, on the one hand, and undue prejudice to the defendant, on the other, they are directed to the pragmatic concern that the jury's verdict rest, in major part at least, on reasonably reliable evidence. Here the probative value of Leonard's highly ambiguous out-of-court statement was, as to Sarvis, negligible. At the same time, its possible prejudice to Sarvis' case is manifest. As I have noted above, Sarvis' defense rested on mistaken identity and alibi; and although the Government presented substantial, competent evidence identifying Sarvis as having participated with Leonard in the crime, the evidence consisted almost exclusively of the somewhat confused testimony of admitted accomplices and was by no means unassailable. In these circumstances, I cannot but conclude that Leonard's extra-judicial statement lent possibly critical support to the Government's case, since it may well have been taken as confirmation by Leonard, against whom the case was far stronger, that Sarvis was indeed his partner in crime. I think there is no question that, had Sarvis been tried alone, admission of this highly prejudicial evidence would have required reversal.

The question thus becomes whether a different result is appropriate here, since the evidence, although inadmissible as to Sarvis, could properly have been considered by the jury in determining the credibility of Leonard. The majority concludes that the evidence was, for this reason, properly admitted, and in reversing holds only that the trial court erred in failing to give an immediate instruction limiting the permissible use of the evidence.51

I cannot agree with the supposition, apparently underlying the majority's conclusion, that, had an immediate instruction been given, the prejudice worked to Sarvis would somehow have been reduced to a tolerable level. The efficacy of an immediate instruction to meet the kind of problem presented here has long been discounted. As the Supreme Court recognized in Bruton v. United States,52 evidence admitted as to one codefendant is effectively before the jury, and quite likely will be used by the jury, as to all; and at least when the evidence is significant, limiting instructions, however immediate and emphatic, cannot be supposed capable of preventing this misuse.53

Nor do I agree with the amjority that the Supreme Court's decision in Nelson v. O'Neil54 compels either the defendant or this Court to rely on such a gossamer barrier to prejudice from inadmissible evidence, at least in the circumstances of this case. In O'Neil, the Court held that 'where a codefendant takes the stand in his own defense, denies making an alleged out-of-court statement implicating the defendant, and proceeds to testify favorably to the defendant concerning the underlying facts, the defendant has been denied no rights protected by the Sixth and Fourteenth Amendments.'55 There are, however, critical differences between the O'Neil case and this. Unlike O'Neil's codefendant, Leonard did not 'proceed to testify favorably' to Sarvis. The two presented entirely separate alibi defenses, rather than a common defense as in O'Neil, and the credibility of Sarvis' alibi could in no way have been supported by the strength and veracity of Leonard's testimony. In O'Neil, the Court said that the codefendant's denial and subsequent testimony concerning the matters contained in his alleged out-of-court statement were, as a practical matter, 'more favorable to the respondent than any that cross-examination by counsel could possibly have produced, had (the codefendant) 'affirmed the statement as his."56 But that is not the case here. Given the evidence of Leonard's guilt, the jury may well have disbelieved his denial. Yet because of his denial, Leonard was, in effect, unavailable to explain the meaning of his out-of-court statement. The ambiguity of the statement, and the consequent prejudice to Sarvis, remained. In these circumstances, I am not at all clear that O'Neil forecloses Sarvis' Confrontation Clause claim.57

But even if O'Neil mandated rejection of that claim, I think the majority wrong in supposing that that ends the matter. In O'Neil, the Court had before it a state criminal conviction, and explicitly dealt only with a Confrontation Clause challenge. The Court has, however, repeatedly stated that the Clause cannot be equated with the hearsay rule and its exceptions.58 Thus, in Dutton v. Evans,59 the Court refused to find that the Clause requires the states to comply with the federal rule excluding co-conspirator statements made during the 'concealment phase' of the conspiracy; but this holding, as Justice Harlan's concurring opinion makes clear, does not call into question the sense and vitality of the federal rule, nor does it affect the continuing necessity of its application in federal trials.60 By the same token, even if O'Neil removes the constitutional premise of Sarvis' claim, there remains the question whether, under rules of evidence that have been or that we think ought to be established, the District Court here failed in its duty to insure that the jury's verdict against Sarvis would rest on reasonably reliable, and hence admissible, evidence. Furthermore, there remains the question whether the District Court met its responsibility under Rule 14 to 'grant a severance of defendants or provide whatever other relief justice requires' when prejudice results from joinder. While the Supreme Court's opinion in O'Neil can and should inform our response to these questions of federal law, its holding in that case does not control our decision.

I could understand, although I could not agree, with a conclusion here that the calculus of prejudice and probative value I have set out above is incorrect-- that because Leonard was present and denied making the extra-judicial statement, its prejudicial effect on Sarvis was sufficiently relieved to render it admissible against him. Such a conclusion would at least deal forthrightly with the evidentiary problem, and would, of course, mean that the statement would be equally admissible were Sarvis tried alone. But once it has been determined that because of its ambiguity-- an ambiguity unrelieved by Leonard's denial-- the statement was inadmissible as to Sarvis, I see no justification for allowing the statement to come before the jury in Sarivs' case. I can see no justification for erecting what amounts, in practical effect, to one rule of evidence for defendants tried singly and quite another for defendants joined for trial under Rule 8.61 Just as the purpose of admissibility rules is the same in both classes of cases, so also the rules and their enforcement must, in my view be the same as well.

I would hold that, as a matter of the law of evidence in this Circuit, the district court erred in failing to exclude evidence of Leonard's alleged out-of-court statement from this trial. I do not mean by this that the Government should necessarily have been forced to forego use of this evidence against Leonard. If the Government believed the statement indispensable to its case against Leonard, an alternative course was open to it: Sarvis' severance and separate trial. In short, I agree with the position advanced in the American Bar Association's Project on Standards for Criminal Justice, Joinder and Severance. When, in a joint trial, the Government proposes to introduce the out-of-court statement of one codefendant that implicates but is not admissible as to another, I would hold that, under Rule 14, the Government must be put to a choice: either (1) the statement must be excluded, (2) all references implicating co-defendants must, if practicable, be entirely deleted, or (3) the defendants must be tried separately.62 Moreover, to prevent unnecessary waste of judicial resources, this choice should be put to the Government in advance of trial.

I am convinced that, as a general matter, only a prophylactic rule of this sort can meet Rule 14's mandate to prevent undue prejudice from joinder, and forestall the kind of abuse of joinder that occurred here. The fact is that leonard's out-of-court statement, standing alone, would have had, perhaps, only a slight prejudicial effect on Sarvis' case. But it did not stand alone. The Government's assiduous questioning concerning Sarvis' residence and the circumstances of his arrest were entirely irrelevant to the impeachment of Leonard. Their necessary and probable consequence, however, was to point to Sarvis as 'the other men' to whom Leonard's statement referred. Indeed, the strong emphasis that the Government placed on this evidence and on Leonard's alleged statement in its closing remarks to the jury strongly suggest that it intended the evidence to have precisely this effect,63 and that it employed joinder to produce to the jury evidence of Sarvis' guilt that it could not have introduced were Sarvis tried alone. Although I think the better rule would be to exclude Leonard's statement and all subsidiary evidence altogether, at the least the trial court should have limited the evidence strictly to that necessary for purposes of impeachment. In my view, its failure to do so here was itself reversible error, error that the majority apparently disregards and that, I presume, its opinion would allow to recur at Sarvis' retrial.

B. In Part V-D of its opinion, the majority holds that the trial court's admission of testimony, by witnesses Hughes and Mauldin, repeating statements purportedly made by codefendant Salters64 did not require, at the instance of Leonard and Sarvis, Salters' severance and separate trial. The problem presented by the admission of Salters' statements is not, of course, primarily one of severance. Unlike Leonard's out-of-court statement, Salters' statements, made, as they allegedly were, immediately after he had witnessed the events they describe, fall within a recognized exception to the hearsay rule, that for 'present sense impressions,' or contemporaneous declarations.65 As such, under the law of hearsay the statements were admissible as to all three codefendants, and would have been equally admissible as to Leonard and Sarvis had Salters been tried separately. Because of this, I think the majority correct in concluding that neither the Bruton rule,66 nor any analogue to that rule premised on Rule 14 and even handed application of the federal law of evidence,67 compelled Salters' severance or, in the alternative, exclusion of his extra-judicial statements from this joint trial.

As the majority recognizes, however, the inapplicability of Bruton does not end the matter. The Supreme Court has said that a violation of Confrontation Clause values may exist 'even though the statements in issue were admitted under an arguably recognized hearsay exception.'68 The central question before us is, therefore, whether Salters' statements were such that the Confrontation Clause would allow their admission in the case against Leonard and Sarvis only if the declarant Salters was available for cross-examination as to the content of the statements and, perhaps equally important here, as to whether the statements had in fact been made.

In its response to this Confrontation Clause problem, the majority relies on a framework of analysis derived from the plurality opinion of the Supreme Court in Dutton v. Evans.69 That opinion rested in major part on a determination that the alleged out-of-court statement of the defendant's co-conspirator could have contributed only in the most marginal way, if at all, to the prosecution's case.70 Given this, the plurality opinion concluded that the statement presented indicia of reliability sufficient to overcome any supposition that cross-examination of the declarant might in any way have affected the jury's verdict, and held, therefore, that the values protected by the Confrontation Clause were satisfied. At the same time, however, the plurality in Dutton carefully distinguished cases in which the evidence contained in out-of-court statements can be deemed 'crucial or devastating,'71 implying that in such cases, the courts' estimates of reliability cannot, consistently with the Clause, be considered an adequate substitute for cross-examination as a means of testing the value of evidence and thus making it more likely that the jury's verdict will be correct.72

Applying the Dutton analysis to the facts of this case, the majority concludes that the information contained in Salters' alleged statements, although significant, was not critical to the jury's verdict against Leonard and Sarvis, and that, therefore, no confrontation values were transgressed despite lack of opportunity to cross-examine the declarant. In reaching this conclusion, however, the majority has, I think, neglected a factor that has played a central role in the Supreme Court's Confrontation Clause decisions. In Barber v. Page, the Court held that the Clause prohibits use against a defendant of testimony given at a prior judicial proceeding by a witness absent from the jurisdiction, unless the prosecution has made a good faith effort to secure the witness's presence at trial; and this was so even though the witness had been subject to cross-examination by the same defendant at the prior proceeding.73 In California v. Green, the Court, citing Barber, said that if the out-of-court declarant 'had died or was otherwise unavailable, the Confrontation Clause would not have been violated by admitting his testimony given at the preliminary hearing-- the right of cross-examination then afforded provides substantial compliance with the purposes behind the confrontation requirement, as long as the declarant's inability to give live testimony is in no way the fault of the State.'74 In short, the Supreme Court's evaluation of Confrontation Clause claims has depended, not alone on the probable significance of extra-judicial statements to the jury's verdict and analysis of their reliability, but also on the necessity of their use-- the reasons why the out-of-court declarant was unavailable for testimony and cross-examination at trial.

It is true that in Dutton v. Evans, the out-of-court declarant was available for call by the State, since, under Georgia law, he was separately tried for his part in the alleged conspiracy. Dutton thus seems to indicate that when the effect of an out-of-court statement is apparently de minimis and when its content and the circumstances of its utterance evidence its reliability, the defendant has the burden of establishing that the declarant's live testimony at trial would in any way have benefitted his case. Indeed, as the plurality opinion in Dutton notes, the defendant could himself have called the declarant; his failure to make any effort to do so militated against his claim that cross-examination might have nullified whatever force the declarant's out-of-court statement may have had.75 This does not mean, however, that the same rule can appropriately be applied when the effect of an extra-judicial statement cannot be deemed harmless. On the contrary, following Barber and Green, I think, as a general matter, the Confrontation Clause compels exclusion of extra-judicial statements that contribute in any substantial way to the prosecution's case unless the Government can affirmatively demonstrate that the declarant is unavailable and, hence, cannot be produced for cross-examination at trial; and the required strength of this showing should depend, in turn, on the extent to which the circumstances and content of the statements indicate their reliability.76 In other words, the Clause's presumption in favor of cross-examination should be overcome, and reliability analysis substituted, only upon a showing that it is necessary and fair to do so.

No such showing was made, nor could have been made, by the Government in the circumstances of this case. Here Salters was subject to call as an involuntary witness neither by the Government nor by Leonard and Sarvis,77 and this was so precisely because, at the instance of the Government, he had been joined with Leonard and Sarvis for trial. The fact is that Salters could have been made available to testify at Leonard's and Sarvis' trial had his own trial been severed and had he, like the other accomplices in the case, been granted use immunity for his testimony.78

If Salters' statements could be supposed to have had no substantial effect on the jury's verdict, I would hold, following Dutton, that his severance and separate trial were required only if Leonard and Sarvis had demonstrated that his testimony might have afforded substantial benefit to their case.79 If Salters' statements could be deemed the central, 'crucial' support of the prosecution's case against Leonard and Sarvis, I would hold, following the implication in Dutton, and because of Salters availability through severance, that his version of the events at issue could properly be presented to the jury only through his own live testimony, and that his our-of-court statements should have been excluded, except for impeachment purposes if Salters had taken the stand.

I agree with the majority, however, that neither of these two polar rules meets the case before us. While admission of Salters' statements cannot be deemed harmless, neither can the information they put before the jury be thought 'crucial' or 'devastating.' The majority grants, however, that the statements may well have been significant to the jury's verdict, since they provided substantiation both of Sarvis' participation in the crime and of Leonard's intent to murder the victim. At the same time, as the majority admits, the statements by no means presented compelling indicia of reliability. Even though the statements, on their face, are relatively unambiguous, they were repeated at trial by persons who may well have had good reason to obscure their own roles in the matters with which the statements dealt. Yet the defendants were afforded no opportunity to question the alleged declarant as to the accuracy of his statements, nor to elicit his affirmance or denial that the statements were indeed his.

In these circumstances I cannot agree with the majority that Confrontation Clause values should be sacrificed to the purported efficiency of joinder. I do not think the Clause necessarily requires that Salters' statements should have been excluded from the prosecution's case. Nonetheless, because of their potential significance to the jury's verdict, I think these hearsay statements should have been admitted into the prosecution's case-in-chief only if the Government had agreed to Salters' severance and had itself called Salters to the stand.

C. I suspect that my disagreement with the majority on the two severance issues discussed above is the consequence, in major part, of a divergence in our views concerning the efficiency of joinder of defendants for trial and the extent to which that supposed efficiency should be preferred when it conflicts with other values that rules of criminal procedure and evidence are in general designed to protect. At one point in its opinion, the majority cites what has become the standard view, that because joinder of defendants effects savings in time and money, not only for the judicial system, but for witnesses and jurors as well, 'the balance has been struck in favor of joint trials . . ..'80 Even if judicial efficiency can properly be defined in these terms alone, I am convinced that the efficiency of joinder has been greatly exaggerated. As Justice Marshall said in dissent in Nelson v. O'Neil and as this case fully demonstrates, much of the savings in time and money through joinder is often lost 'through protracted litigation that results from the impingement or near impingement on' codefendants' rights.81 Literally hundreds of pages of the long transcript in this case are devoted to repeated bench conferences concerning problems resulting from joinder, problems that have come to this Court on appeal. Indeed, the multiple errors that infected the trial of Leonard and Sarvis lend strong support to a hypothesis that, as the number of defendants joined for trial increases, the possibilities of prejudice, and of reversible error, increase geometrically.

It is, of course, possible to avoid loss of these savings, in some part at least, by retrenching on the rights of criminal defendants joined with others for trial, by stretching the rules of evidence and diluting constitutional guarantees to meet the exigencies of the joint trial. Indeed, I am afraid that we have too often done exactly this. I cannot, however, accept the proposition that judicial efficiency can be defined in terms of time and money alone. Elaboration and application of the rules of evidence and the Confrontation guarantee are themselves directed to a fully pragmatic concern; they are designed to provide some reasonable assurance that defendants found guilty are guilty. So long as this assurance can be maintained, I am as enthusiastic as any about devices that reduce judicial costs. But if this assurance is diminished, I think the price too high. And when, at the margin, the choice is between savings in time and money on the one side and reasonable assurance of guilt on the other, I think that efficiency in the true sense and, what is more important, the integrity of the criminal justice system demand that any doubts be resolved in favor of the latter.

On these points I respectfully dissent from the opinion of the Court.