STEVENS, J., concurring in judgment Most important here, it is also possible that both persons lacked knowledge of the contents of the relevant container. Moreover, that hypothesis is compatible with individual defenses of lack of knowledge. There is no logical inconsistency between a version of events in which one person is ignorant, and a version in which the other is ignorant; unlikely as it may seem, it is at least theoretically possible that both versions are true, in that both persons are ignorant. In other words, dual ignorance defenses do not necessarily translate into "mutually antagonistic" defenses, as that term is used in reviewing severance motions, because acceptance of one defense does not necessarily preclude acceptance of the other and acquittal of the codefendant.1 In my view, the defenses presented in this case did not rise to the level of mutual antagonism. First, as to Garcia and Martinez, neither of whom testified, the only defense presented was that the Government had failed to carry its burden of proving guilt beyond a reasonable doubt. Nothing in the testimony presented by their codefendants, Soto and Zafiro, supplemented the Government's proof of their guilt in any way. Soto's testimony that he did not know the contents of the box he delivered with Garcia, as discussed above, could have been accepted in toto without precluding acquittal of his codefendant. Similarly, the jury could have accepted Zafiro's testimony that she did not know the contents of the suitcase found in her apartment, and also acquitted Martinez. It is true, of course, that the jury was unlikely to believe that none of the defendants knew what was in the box or suitcase. Accordingly, it must be acknowledged that if the jury had believed that Soto and Zafiro were ignorant, then it would have been more likely to believe that Garcia and Martinez were not. That, however, is not the standard for 1 See ante, at 538, citing cases. See also State v. Kinkade, 140 Ariz. 91, 93, 680 P. 2d 801, 803 (1984) (defining "mutually exclusive" defenses). STEVENS, J., concurring in judgment mutually antagonistic defenses.2 And in any event, the jury in this case obviously did not believe Soto and Zafiro, as it convicted both of them. Accordingly, there is no basis, in law or fact, for concluding that the testimony of Soto and Zafiro prejudiced their codefendants. There is even less merit to the suggestion that Soto or Zafiro was prejudiced by the denial of their severance motions. Neither Garcia nor Martinez testified at all, of course, and the District Court explicitly cautioned the jury that the arguments made by their attorneys were not to be considered as evidence. Ante, at 541. Moreover, the assertion by his counsel that Garcia did not know the contents of the box is not inconsistent with Soto's ignorance or innocence; nor is the similar assertion by counsel for Martinez inconsistent with Zafiro's possible innocence. In my opinion, the District Court correctly determined that the defenses presented in this case were not "mutually antagonistic." See App. 88-89. I would save for another day evaluation of the prejudice that may arise when the evidence or testimony offered by one defendant is truly irreconcilable with the innocence of a codefendant. Because the facts here do not present the issue squarely, I hesitate in this case to develop a rule that would govern the very different situation faced in cases like People v. Braune, 363 Ill. 551, 557, 2 N. E. 2d 839, 842 (1936), in which mutually exclusive defenses transform a trial into "more of a contest between the defendants than between the people and the defendants." Under such circumstances, joinder may well be highly prejudicial, particularly when the prosecutor's own case in chief is marginal and the decisive evidence of guilt is left to be provided by a codefendant. The burden of overcoming any individual defendant's presumption of innocence, by proving guilt beyond a reasonable doubt, rests solely on the shoulders of the prosecutor. Joinder is problematic in cases involving mutually antagonistic 2 Cf. Kinkade, 140 Ariz., at 93, 680 P. 2d, at 803 (distinguishing "competing" from mutually antagonistic defenses). 3 STEVENS, J., concurring in judgment defenses because it may operate to reduce the burden on the prosecutor, in two general ways. First, joinder may introduce what is in effect a second prosecutor into a case, by turning each codefendant into the other's most forceful adversary. Second, joinder may invite a jury confronted with two defendants, at least one of whom is almost certainly guilty, to convict the defendant who appears the more guilty of the two regardless of whether the prosecutor has proven guilt beyond a reasonable doubt as to that particular defendant. Though the Court is surely correct that this second risk may be minimized by careful instructions insisting on separate consideration of the evidence as to each codefendant, ante, at 540-541, the danger will remain relevant to the prejudice inquiry in some cases. 5 Given these concerns, I cannot share the Court's enthusiastic and unqualified "preference" for the joint trial of defendants indicted together. See ante, at 537. The Court correctly notes that a similar preference was announced a few years ago in Richardson v. Marsh, 481 U. S. 200, 209 (1987), and that the Court had sustained the permissibility 3 "Defendants who accuse each other bring the effect of a second prosecutor into the case with respect to their codefendant. In order to zealously represent his client, each codefendant's counsel must do everything possible to convict the other defendant. The existence of this extra prosecutor is particularly troublesome because the defense counsel are not always held to the limitations and standards imposed on the government prosecutor." United States v. Tootick, 952 F. 2d 1078, 1082 (CA9 1991). See also United States v. Romanello, 726 F. 2d 173, 179 (CA5 1984). 4 See State v. Vinal, 198 Conn. 644, 652, 504 A. 2d 1364, 1368 (1986) (in joint trial with mutually antagonistic defenses, "where one defendant is found not guilty, it becomes likely under these circumstances that the conviction of the losing defendant is more a result of his codefendant's success in defending himself than it is a product of the state's satisfaction of its constitutional duty to prove the accused guilty beyond a reasonable doubt"). 5 Tootick, 952 F. 2d, at 1082. See also People v. Braune, 363 Ill. 551, 556, 2 N. E. 2d 839, 842 (1936) (“[N]o judge, however learned and skillful," could have prevented risk of prejudice in particularly aggravated case). STEVENS, J., concurring in judgment of joint trials on at least two prior occasions. There will, however, almost certainly be multidefendant cases in which a series of separate trials would be not only more reliable, but also more efficient and manageable than some of the mammoth conspiracy cases which the Government often elects to prosecute. And in all cases, the Court should be mindful of the serious risks of prejudice and overreaching that are characteristic of joint trials, particularly when a conspiracy count is included in the indictment. Justice Jackson's eloquent description of these concerns in his separate opinion in Krulewitch v. United States, 336 U. S. 440, 454 (1949), explains why there is much more at stake here than administrative convenience. See also United States v. Romanello, 726 F. 2d 173 (CA5 1984). I agree with the Court that a "bright-line rule, mandating severance whenever codefendants have conflicting defenses" is unwarranted. See ante, at 538. For the reasons discussed above, however, I think district courts must retain their traditional discretion to consider severance whenever mutually antagonistic defenses are presented. Accordingly, I would refrain from announcing a preference for joint trials, or any general rule that might be construed as a limit on that discretion. Because I believe the District Court correctly decided the severance motions in this case, I concur in the Court's judgment of affirmance. 6 In neither Opper v. United States, 348 U. S. 84 (1954), nor United States v. Marchant, 12 Wheat. 480 (1827), however, did the Court express a "preference" for joint trials. Syllabus UNITED STATES v. HILL ET UX. CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR THE FEDERAL CIRCUIT No. 91-1421. Argued November 2, 1992-Decided January 25, 1993 Under § 57(a)(8) of the Internal Revenue Code of 1954, 26 U. S. C. § 57(a)(8) (1976 ed.), "the excess of the deduction for depletion . . . over the adjusted basis of" "property (as defined in [§]614)" is an "ite[m] of tax preference" on which a taxpayer must pay a "minimum tax" for the tax year in question. See §56(a). In computing the minimum taxes due on their interests in oil and gas deposits for tax years 1981 and 1982, respondents Hill calculated their depletion allowances according to the "percentage depletion" method, and included in the interests' adjusted bases the unrecovered costs of certain depreciable tangible items used in drilling and development (machinery, tools, pipes, etc.), as identified in § 1.612-4(c)(1) of the applicable Treasury Department regulations. The Commissioner of Internal Revenue disputed that inclusion, and assessed larger minimum taxes based on the exclusion of the tangible costs from the mineral interests' adjusted bases. The Hills paid the resulting deficiencies and filed a refund claim, which the Commissioner denied. The Claims Court granted summary judgment for the Hills in their ensuing refund suit, and the Court of Appeals affirmed. Held: The term "adjusted basis," as used in §57(a)(8), does not include the depreciable drilling and development costs identified in Treas. Reg. § 1.612-4(c)(1). Pp. 553-564. (a) The definitional scheme established by the Code and accompanying regulations suggests strongly that the "property" with which § 57(a)(8) is concerned excludes just those improvements that the Hills wish to include in adjusted basis. Section 614(a) defines "property" for § 57(a)(8)'s purposes as "each separate interest owned by the taxpayer in each mineral deposit." Treasury Reg. §1.611-1(d)(4) defines "mineral deposit" as "minerals in place," while Treas. Reg. §1.611-1(d)(3) defines "mineral enterprise" to include "the mineral deposit or deposits and improvements, if any, used in... the production of oil and gas." (Emphasis added.) Because these regulatory definitions were well established when Congress passed §57(a)(8), it is reasonable to assume that Congress relied on the accepted distinction between them in its reference to "mineral deposit" in §614. This conclusion is confirmed by Treas. Reg. §1.57-1(h)(3)'s incorporation into §57(a)(8) of § 1016 of the Code, 26 U. S. C. § 1016 (1976 ed. and Supp. V), which provides the rules |