The defendant, a high- ranking official of the Gangster Disciples, a large Chicago street gang whose activities are described in our recent opinion in United States v. Jackson, 207 F.3d 910 (7th Cir. 2000), was convicted of having ordered the murder of (1) a person who was assisting in a federal criminal investigation and (2) that person and one other in furtherance of a continuing criminal enterprise, and was sentenced to death. 18 U.S.C. sec.sec. 924(i), 1121(a); 21 U.S.C. sec. 848(e). The victims were gang members who had been arrested but who had then been released pending their trials and who the defendant feared were working with the government to catch him. He does not deny having committed the two murders; his appeal primarily challenges the conduct of the sentencing hearing. He does, however, raise one point about the conduct of the trial itself--that his right to represent himself was infringed--and we begin there.

Two weeks before the trial began, the defendant filed a pro se motion captioned "Defendants Motion to Proceed Pro-Se." In it he argued that his lawyers' representation of him was so deficient that it violated his right to effective assistance of counsel. But rather than asking for the appointment of new counsel, the motion concluded: "Petitioner knows absolutely nothing about the law. But petitioner feels strongly that as his life is on the line, he can do more for his defense than his attorney's have so far." It is unlikely that his statement about knowing nothing about the law is false modesty. The defendant is not an educated person, and his IQ is only 74.

The motion was never ruled on. Apparently it had gotten lost in the shuffle, United States v. Taglia, 922 F.2d 413, 416 (7th Cir. 1991), for the judge stated at a post-trial hearing that she did not remember having seen it and that she would have remembered it if she had seen it, given the gravity and novelty of the case--this was only the second federal death penalty trial in the Northern District of Illinois since the reinstatement of the federal death penalty, and the first to result in a death sentence. The defendant did not renew the motion. His lawyers were unaware of and never mentioned it. He made a number of pro se motions during the course of the trial and in none did he express any dissatisfaction with his lawyers or a desire to represent himself.

Although a defendant has an absolute right to defend himself against a criminal charge, however grave, unless he is mentally incompetent to decide to do so, Godinez v. Moran, 509 U.S. 389, 396-400 (1993); Davis v. Greer, 13 F.3d 1134, 1138 (7th Cir. 1994), the right can be waived either expressly or by implication. There are two types of implied waiver. One, the only one that can properly be called "waiver," is where an intention to relinquish the right, although not expressed, can be inferred. The other, properly called "forfeiture" rather than "waiver" in recognition that waiver is canonically defined as an intentional relinquishment of a right, Johnson v. Zerbst, 304 U.S. 458, 464 (1938), is where the right is taken away from its holder as a penalty for failure to assert it in a clear and timely manner. It is not always clear when a case is one of actually implied waiver or one of forfeiture. The "waiver" of the right of self-representation illustrates the problem. When as in the usual case the defendant is represented by a lawyer, the fact of representation is taken to be the defendant's waiver of his right to represent himself, since "representation by counsel and self-representation are mutually exclusive entitlements," Cain v. Peters, 972 F.2d 748, 750 (7th Cir. 1992), so that "assertion of one constitutes a de facto waiver of the other." United States v. Singleton, 107 F.3d 1091, 1096 (4th Cir. 1997). "Defendants forfeit self- representation by remaining silent at critical junctures before or during trial." Cain v. Peters, supra, 972 F.2d at 750. See also Munkus v. Furlong, 170 F.3d 980, 983-84 (10th Cir. 1999). Failure to assert the right of self- representation waives it without regard to the intentions of the defendant in not asserting it.

Among the grounds (catalogued in id. at 984) for forfeiture of the right is delay in asserting it. United States v. Oakey, 853 F.2d 551, 553 (7th Cir. 1988). This case was more than a year old, and on the verge of trial, when the defendant, who until then had been represented by counsel, filed his motion. But because a motion for self- representation is timely if made before the jury is empaneled, United States v. Akers, 215 F.3d 1089, 1097 (10th Cir. 2000); United States v. Walker, 142 F.3d 103, 108 (2d Cir. 1998), unless made for the purpose of delaying or disrupting the trial, Moore v. Calderon, 108 F.3d 261, 264 (9th Cir. 1997), which is not argued, we set the question of timeliness to one side and with it the issue of forfeiture of the defendant's right to represent himself. What this is is a case of implicit waiver in the strict, the intentional sense. The defendant did not want to represent himself, though he didn't say so in so many words. The purpose of the motion, it is apparent, was to express in the most dramatic possible fashion his current dissatisfaction with his lawyers. No even marginally rational person who knew absolutely nothing about the law would want to defend himself against a capital charge without a lawyer's assistance. The defendant's fit of dissatisfaction with his lawyers soon passed. He neither moved to have them replaced nor renewed his motion to be permitted to represent himself. His appellate counsel, a specialist in defending death-penalty cases, has not pointed to any conduct by the defendant's trial lawyers that might have impelled the defendant to think himself better able than they to defend the case. The only plausible inference from the defendant's conduct is that he acquiesced in the denial by judicial inaction of his motion and thereby deliberately relinquished his right of self-representation. Cain v. Peters, supra, 972 F.2d at 750; Wilson v. Walker, 204 F.3d 33, 37-39 (2d Cir. 2000) (per curiam); Brown v. Wainwright, 665 F.2d 607, 610-11 (5th Cir. 1982) (en banc); United States v. Montgomery, 529 F.2d 1404, 1406 (10th Cir. 1976).

We add that as he has made no representation that if we order a new trial he will persist in his desire to represent himself, his claim that his right of self-representation was infringed may be moot, as well as having no merit for the reasons just indicated. For if as we expect he would be represented by lawyers at any new trial, he would not have vindicated the right of self- representation upon which he premises his appeal from the denial of that right. The point is not that at a subsequent trial he would be estopped to invoke his right to counsel, an argument rejected in the only cases to have considered the issue. United States v. McKinley, 58 F.3d 1475, 1483 (10th Cir. 1995); Johnstone v. Kelly, 812 F.2d 821 (2d Cir. 1987) (per curiam). The point is rather that if he wants on remand exactly what he had in his first trial, namely representation by competent lawyers, it is difficult to understand what he lost by the denial of his motion he had at the first trial what he wants at the second.

We turn to the sentencing issues. One of the jurors who participated in the deliberations that resulted in the defendant's being found guilty failed to show up for the sentencing hearing and was immediately replaced by one of the alternates, who had sat through the trial but had not participated in the jury deliberations. The judge made no effort to find out why the juror who was replaced had not shown up, but it is a sound practice immediately to replace a no-show juror, as authorized by Fed. R. Crim P. 24(c); see United States v. Peters, 617 F.2d 503, 505 (7th Cir. 1980) (per curiam); United States v. Gay, 967 F.2d 322, 324-25 (9th Cir. 1992); United States v. Rodriguez, 573 F.2d 330, 332 (5th Cir. 1978); United States v. Domenech, 476 F.2d 1229, 1232 (2d Cir. 1973), out of consideration for the remaining jurors and in order to remind them of the seriousness of their duty.

The defendant's lawyer made no objection at trial to dropping the tardy juror, or to the fact that under the then Fed. R. Crim. P. 24(c) (1998), the alternate juror should have been discharged when the jury retired to consider the defendant's guilt, United States v. Patterson, 23 F.3d 1239, 1252 (7th Cir. 1994); United States v. Josefik, 753 F.2d 585, 587 (7th Cir. 1985); United States v. Webster, 162 F.3d 308, 347 (5th Cir. 1998), or--the argument he now presses most vigorously on us--that the participation of the alternate in the sentencing deliberations violated 18 U.S.C. sec. 3593(b)(1). That section provides that a federal death-penalty sentencing hearing is to be conducted "before the jury that determined the defendant's guilt," with certain exceptions, as where the defendant waived his right to a jury trial or the jury that determined his guilt has already been discharged. sec. 3593(b)(2). The defendant argues that the rights conferred by this provision and by Rule 24(c) are ones his lawyer could not waive for him, and so the failure to object at trial to their denial should be disregarded.

Certain rights are personal to the criminal defendant. See, e.g., Fed. R. Crim. P. 11(c)(6) ("Before accepting a plea of guilty . . . the court must address the defendant personally in open court and inform the defendant of, and determine that the defendant understands . . . the terms of any provision in a plea agreement waiving the right to appeal or to collaterally attack the sentence"); Jones v. Barnes, 463 U.S. 745, 751 (1983) ("the accused has the ultimate authority to make certain fundamental decisions regarding the case, as to whether to plead guilty, waive a jury, testify in his or her own behalf, or take an appeal"); United States v. Shukri, 207 F.3d 412, 416 (7th Cir. 2000); Sexton v. French, 163 F.3d 874, 885 (4th Cir. 1998). But rights that have a mainly tactical significance and require legal training to appreciate and weigh are for the defendant's lawyer to assert or to waive as the lawyer sees fit, e.g., United States v. Boyd, 86 F.3d 719, 723 (7th Cir. 1996); United States v. Washington, 198 F.3d 721 (8th Cir. 1999); United States v. Plitman, 194 F.3d 59, 63 (2d Cir. 1999); see generally Wayne R. LaFave, Jerold H. Israel & Nancy J. King, Criminal Procedure sec. 11.6(a), pp. 598-603 (2d ed. 1999), though subject of course to eventual judicial second-guessing should the defendant claim that the lawyer made a tactical blunder so grave as to fall below the minimum level of professionally competent representation.

A highly pertinent example is jury selection. United States v. Boyd, supra. For what we have here is an alleged right to shape the composition of the jury. And so if there is a right to insist that should one of the jurors who deliberated on guilt drop out before the sentencing hearing an entire new jury be impaneled for sentencing, it is a right that falls within the domain of tactics rather than that of basic rights. This conclusion is compelled by common sense as well as by the decision in Boyd. Bearing in mind that it takes only one juror to nix a death sentence, a lawyer who senses that one or more of the jurors who found his client guilty nevertheless seems sufficiently simpatico not to vote for death might prefer to retain the original panel with the addition of one of the alternates than to take his chances with a completely new panel. The judge is not required to voir dire the defendant to make sure the latter agrees with the lawyer's tactical decision. That would actually handicap most criminal defendants by inviting them to make tactical decisions that most of them are incompetent to make.

In any event we do not think that the procedure that was employed violates the statute. The statute makes no provision for the situation that occurred here, leaving it to the good sense of the judges to deal with. We find guidance to the proper resolution in the 1999 amendment to Rule 24 of the Federal Rules of Criminal Procedure, which, altering the previous practice (to which we made reference earlier), allows the trial judge to replace a regular juror with an alternate during deliberations, which must then recommence. Fed. R. Crim. P. 24(c)(3). In other words, the fact that the alternate missed some of the deliberations is no longer regarded as a fatal objection, or indeed as any objection, to his participating in the jury's decision. Compare United States v. Josefik, supra, 753 F.2d at 587. The analogy to the procedure employed here is close. The deliberations that eventuated in the sentence of death were in two stages, a guilt stage and a sentencing stage. The alternate missed the first stage but participated in the second. True, the entire deliberations did not recommence; but the issues of guilt and of punishment are sufficiently distinct that the alternate didn't have to hear the deliberations on the former issue in order to be able to participate meaningfully in the deliberations on the latter issue. He had sat through the entire trial, which is the important thing.

The defendant argues that the alternate might fear he had "missed" something in the part of the deliberations that he had not been party to. But that is equally true in a Rule 24(c)(3) situation, since recommencing the deliberations would not wipe out the original jurors' memories of whatever discussion had preceded the alternate's joining them. As someone not committed to the defendant's guilt, never having voted on the question of guilt, the alternate added at the sentencing stage might actually have a greater inclination to lenity than the jurors whom he was joining.

The next issue is whether the defendant was unfairly surprised by false testimony allegedly presented by the government on rebuttal at the sentencing hearing. To decide to impose (technically, "recommend," but the recommendation is binding, 28 U.S.C. sec. 3594) the death penalty, the jury must decide that the aggravating factors, that is, factors beyond the murder itself that warrant imposing the ultimate penalty on this defendant, sufficiently outweigh any mitigating factors to justify the imposition. sec. 3593(e). There is a list of aggravating factors in section 3592(c) (for murder) but the list is stated to be nonexhaustive. Evidence of an aggravating or mitigating factor need not be admissible under the rules of evidence. sec. 3593(c).

The government alleged as a nonstatutory aggravating factor in this case "future dangerousness." The government pointed out that gang leaders have been known (the boss of the Gangster Disciples, Larry Hoover, notoriously) to continue to direct the gang's affairs from prison either by telephone or through visitors, and that the defendant himself, when in jail awaiting trial, had threatened his codefendant. In response the defendant called to the stand a psychologist who had toured the federal prison in Florence, Colorado, which is the maximum-security prison in the federal system, the successor to Alcatraz and Marion. Department of Justice, "Release Preparation Program," 61 Fed. Reg. 38042-02, 38043 (Jul. 22, 1996); Mark Johnson, "Colorado Facility is Pacesetter of Newest 'Supermax' Prisons," Houston Chronicle, June 20, 1999, p. 8. He testified about the security arrangements in the prison's control unit, where prisoners are kept essentially in solitary confinement. He assured the jury that if it recommended that the defendant be sentenced to life in prison rather than to death, the defendant would no longer be a threat to anyone, since he could be sentenced to spend the rest of his life in the control unit at Florence.

In rebuttal the government put on a federal prison warden, formerly an assistant warden at Florence, to testify about Florence and about the policies of the federal Bureau of Prisons. To the extent that these policies are prescribed or codified in statutes or regulations, as distinct from being informal policies, this testimony was improper, though the defendant makes no issue of its admissibility and so any objection is waived. Witnesses testify about fact, not law. When a legal proposition is relevant to the jury's consideration, the proper procedure is for the judge to instruct the jury on the proposition. E.g., Bammerlin v. Navistar Int'l Transportation Corp., 30 F.3d 898, 900 (7th Cir. 1994); Harbor Ins. Co. v. Continental Bank Corp., 922 F.2d 357, 366 (7th Cir. 1990); Nieves-Villanueva v. Soto- Rivera, 133 F.3d 92, 99 (1st Cir. 1997); Molecular Technology Corp. v. Valentine, 925 F.2d 910, 919 (6th Cir. 1991). If the Bureau of Prisons is forbidden by law to confine a prisoner in a control unit for his entire life on the basis of evidence presented at his trial, that is something for the judge to tell the jury, not for a witness to testify to.

Much of what the warden testified to was factual in character, however, for example that Florence has a limited capacity and its control unit can accommodate only 68 inmates, that most murderers and most gang leaders in federal prisons are not at Florence, let alone in Florence's control unit, that most prisoners in Florence came there after misbehaving at other facilities rather than directly from being sentenced, and that even prisoners in the control unit can have visitors and can make one 15-minute telephone call a month. He also testified that "it is not permissible, by the Bureau of Prisons policy, to keep an inmate in that status [i.e., in a control unit] indefinitely."

It is unclear whether he meant a purely informal policy or was testifying to his understanding of some legal restriction. But in fact there is a legal restriction. Regulations by the Bureau of Prisons, to which the Attorney General has delegated her plenary authority over the management of federal prisons, see 18 U.S.C. sec.sec. 4001(b), 4042, require federal prison wardens to consider six factors in deciding whether to place an inmate in a control unit. All six relate to the inmate's behavior in prison, such as possession of a weapon in the prison, 28 C.F.R. sec. 541.41(b)(3), or causing injury to other people in the prison. sec. 541.41(b)(1). The warden may consider a seventh factor, the nature of the inmate's offense of conviction, but only "in combination with other factor(s) as described in paragraph (b)." sec. 541.41(b)(7). "An inmate may not be considered [for confinement in a control unit] solely on [the basis of] the nature of the crime which resulted in that inmate's incarceration." Id.

If the Attorney General or the head of a federal law enforcement or intelligence agency determines "that there is a substantial risk that a prisoner's communications or contacts with persons could result in death or serious bodily injury to persons," the warden can be authorized to take "special administrative measures" that "may include housing the inmate in administrative detention and/or limiting certain privileges, including, but not limited to, correspondence, visiting, interviews with representatives of the news media, and use of the telephone, as is reasonably necessary to protect persons against the risk of acts of violence or terrorism." sec. 501.3(a). These restrictions may, however, be imposed only in 120-day increments; and each time, before they can be reimposed, the warden must conduct the risk determination afresh. sec. 501.3(c).

The limitations in these regulations imply that the Bureau of Prisons could not assign a prisoner directly upon his admission to the federal prison system to spend the rest of his life in the control unit without the possibility of reconsideration. The regulation requiring the bureau to review an inmate's control unit status "at least once every 60 to 90 days . . . to determine the inmate's readiness for release from the [Control] Unit," 28 C.F.R. sec. 541.49(d), points in the same direction.

The facts to which the warden testified and the law that he may or may not have been alluding to point to the same conclusion, that there can be no assurance that if the defendant were sentenced to life in prison he could not commit further serious crimes. The defendant calls the warden's testimony "false" and argues that since it came in on rebuttal he didn't have a chance to meet it and so was unfairly surprised. If true (and if the evidence was deemed prejudicial), this would presumably entitle him to a new sentencing hearing, just as, had the falsity neither been discovered nor discoverable until after the hearing had ended, he could have sought a new hearing on the basis of newly discovered evidence. See, e.g., United States v. Austin, 103 F.3d 606, 609 (7th Cir. 1997); United States v. Reed, 986 F.2d 191, 192-93 (7th Cir. 1993); United States v. Sinclair, 109 F.3d 1527, 1531-32 (10th Cir. 1997); United States v. Moore, 54 F.3d 92, 99 (2d Cir. 1995). But the warden's testimony, though it did not track the regulations exactly, was not false. The impression that he conveyed of practice and legal policy was correct.

We know from cases in this court involving murders by prisoners in the control units of federal prisons, see United States v. Fountain, 768 F.2d 790, modified, 777 F.2d 345 (7th Cir. 1985) (per curiam); United States v. Silverstein, 732 F.2d 1338(7th Cir. 1984); United States v. Fountain, 642 F.2d 1083 (7th Cir. 1981); cf. Shoats v. Horn, 213 F.3d 140, 141 (3d Cir. 2000); Echols v. Sullivan, 521 F.2d 206 (5th Cir. 1975) (per curiam), that such units cannot be made totally secure. And we know that anyone who has access to a telephone or is permitted to receive visitors may be able to transmit a lethal message in code. We also know that nothing in federal law authorizes a judge to sentence a prisoner to life in the control unit. Quite apart from the fact that "a sentencing court has no authority to order that a convicted defendant be confined in a particular facility, much less placed in a particular treatment program," United States v. Williams, 65 F.3d 301, 307 (2d Cir. 1995), a prison control unit is an internal disciplinary mechanism that is not intended or designed for lifetime commitment. The Bureau of Prisons could not, therefore, commit a prisoner to the control unit for life, refusing to consider circumstances that might render his joining the open population of the prison harmless, such as extreme old age or the dissolution of the gang with which he had been affiliated. "Of course, administrative segregation may not be used as a pretext for indefinite confinement of an inmate" to a segregation unit (i.e., a control unit). Hewitt v. Helms, 459 U.S. 460, 477 n. 9 (1983); cf. Bono v. Saxbe, 620 F.2d 609, 614 (7th Cir. 1980); In re Long Term Administrative Segregation of Inmates Designated as Five Percenters, 174 F.3d 464, 472 (4th Cir. 1999); United States v. Felipe, 148 F.3d 101, 111 (2d Cir. 1998).

A prison's control unit is not intended as a punishment for the crime that got the prisoner into prison in the first place, like a sentence of imprisonment at hard labor. Its purpose rather is to deter and prevent violations of prison disciplinary rules and to protect prisoners, guards, and in some cases people outside the prison, or the society at large, against dangerous conduct by the prisoner. See 28 C.F.R. sec. 541.40(a) ("in an effort to maintain a safe and orderly environment within its institutions, the Bureau of Prisons operates control unit programs intended to place into a separate unit those inmates who are unable to function in a less restrictive environment without being a threat to others or to the orderly operation of the institution"). If, no matter what the nature of the activities that got him into prison in the first place, the prisoner poses no danger to anyone that would justify putting him into or keeping him in a control unit, such confinement is, as the regulations we cited earlier make clear, unauthorized.

Against this the defendant cites recent cases in which, he claims, federal judges have sentenced criminal defendants to be confined in a control unit. United States v. Felipe, No. S16 94 CR. 395 JSM, 1997 WL 220302 (S.D.N.Y. Apr. 29, 1997), aff'd in relevant respect, 148 F.3d 101, 109-11 (2d Cir. 1998); United States v. Yousef, S12 93 CR. 180 (S.D.N.Y. Jan. 8, 1998) (sentencing transcript); United States v. Jones, No. 96-458, 97-0355 (D. Md. 1998). Only Felipe involved a judicial order, however, and the order didn't purport to sentence the defendant to the control unit for life. The judge reserved the right "hopefully to adjust the conditions, if a change of circumstances occurs with the passage of time"--which he soon did. Benjamin Weiser, "Judge Gives OK For New Member of Prison Rec Club," Plain Dealer (Cleveland), March 13, 1999, p. A1. (To our astonishment, the modification includes, if the newspaper is to be believed, express authorization for Felipe to consort with two of the most notorious inmates in American prisons today, Timothy McVeigh and Theodore Kaczynski.)

The judge in the Felipe case found his authority for sending the defendant to the control unit in 18 U.S.C. sec. 3582(d). That statute authorizes the sentencing judge to "include as a part of the sentence" of imprisonment for certain crimes, including some of the noncapital drug-related offenses of which our defendant was convicted and that we have not discussed because there is no challenge to the judgment regarding them, "an order that requires that the defendant not associate or communicate with a specified person, other than his attorney, upon a showing of probable cause to believe that association or communication with such person is for the purpose of enabling the defendant to control, manage, direct, finance, or otherwise participate in an illegal enterprise." That is a far cry from a sentence to life imprisonment in a control unit, as is the parallel Bureau of Prisons regulation that we cited earlier. The statute requires the government to be able to name a specific person that it has probable cause to believe will associate or communicate with the defendant for illegal purposes--though the Second Circuit in Felipe stretched the statute to permit the sentencing judge to cut off the defendant from contact with everyone in the world except designated family members, 148 F.3d at 110. And if the government does make the required showing, the only consequence is that the defendant will not be permitted to associate or communicate with that person. It is doubtful whether the statute authorizes indefinite confinement in the control unit, as distinct from a limitation on visits, phone calls, or association with specified other inmates, though this we need not decide. All other objections to the defendant's effort to rely on Felipe to one side, he did not argue in the district court that he could be sentenced under section 3582(d).

The warden also testified that a member of the Aryan Brotherhood, the most notorious of U.S. prison gangs, had while imprisoned at Florence managed to convey an order to members of the gang at another prison to kill two inmates of that prison, and that the order had been carried out. The defendant objects, and the government concedes, that this was hearsay evidence; but as we noted earlier, the federal death-penalty statute does not require that evidence of aggravating and mitigating factors be admissible under the rules of evidence, only that its probative value not be outweighed by the danger of its confusing, misleading, or prejudicing the jury. 18 U.S.C. sec. 3593(c). The balancing is committed to the discretion of the district judge, United States v. Hall, 152 F.3d 381, 397 (5th Cir. 1998), not here abused. The incident involving the Aryan Brotherhood murders, which is consistent with previous conduct by that infamous gang, see, e.g., United States v. McKinney, 954 F.2d 471, 472-74, 478-79 (7th Cir. 1992); Gometz v. Henman, 807 F.2d 113 (7th Cir. 1986); United States v. Fountain, supra; United States v. Mills, 704 F.2d 1553, 1555 (11th Cir. 1983), came to the warden's attention through official channels and concerned a matter of transcendent professional interest to a prison official. He was in a good position to form a judgment of the reliability of this information, which went directly to the issue of the ability of the federal prison system to defang the murderers in its custody.

The defendant next objects under the Brady rule to the government's failure to turn over "evidence" that inmates in the prison in which the Aryan Brotherhood hits occurred believed that the deaths of the inmates in question had other causes. This evidence, limited as it was to rumors that sweep prisons when any untoward event occurs, was so valueless that its exclusion from the trial could not be considered material. United States v. Villarreal, 963 F.2d 725, 730 (5th Cir. 1992); Hopkinson v. Shillinger, 866 F.2d 1185, 1213, reheard en banc on other grounds, 888 F.2d 1286 (10th Cir. 1989).

We add as a detail that while the defendant was of course entitled to counter the government's evidence that he would be a continued menace to society while in prison, that being evidence offered to establish an aggravating factor, 18 U.S.C. sec. 3593(c); cf. Gardner v. Florida, 430 U.S. 349, 362 (1977), he should not have been permitted to present to the jury, as he was, evidence of the existence of maximum-security federal prisons decked out with control units, in order to establish a mitigating factor. A mitigating factor is a factor arguing against sentencing this defendant to death; it is not an argument against the death penalty in general. See Penry v. Lynaugh, 492 U.S. 302, 328 (1989); Eddings v. Oklahoma, 455 U.S. 104, 110 (1982); Lockett v. Ohio, 438 U.S. 586, 604 n. 12 (1978) (plurality opinion). The argument that life in prison without parole, especially if it is spent in the prison's control unit and thus in an approximation to solitary confinement, sufficiently achieves the objectives aimed at by the death penalty to make the latter otiose is an argument addressed to legislatures, not to a jury. This is apparent from the fact that the list of mitigating factors in the federal death- penalty statute does not include the harshness or ugliness or (some would say) the immorality of the death penalty, but only factors specific to the defendant. See 18 U.S.C. sec. 3592(a), and in particular subsection (a)(8) ("other factors in the defendant's background, record, or character or any other circumstances of the offense that mitigate against imposition of the death sentence") (emphasis added).

The argument is also illogical, as it amounts to saying that because this defendant is so dangerous, he does not deserve to be sentenced to death, since his dangerousness will assure his secure confinement. And its illogic shows that it is really an argument against the death penalty, period, since if this defendant should be spared because he is unusually dangerous, surely less dangerous murderers should not be executed either, even though, because they are less dangerous, they are less likely to be confined securely.

The last issue involves the verdict form. The statute requires the jury to include as part of the verdict special findings with respect to any aggravating factors that it determines to exist. But it also provides that "any member of the jury who finds the existence of a mitigating factor may consider such factor established." 18 U.S.C. sec. 3593(d). Whether the juror must also be permitted to include such a finding (that is, his finding that a mitigating factor exists) in the verdict, or in some separate written statement of his own, may be doubted--has in fact been rejected by the only two courts to consider the issue, United States v. Paul, 217 F.3d 989, 999 n.6 (8th Cir. June 27, 2000); United States v. Hall, supra, 152 F.3d at 413-- but is conceded by the government on this appeal.

The jury was given two separate verdict forms, one for each of the two murders that the defendant had committed. Each form lists more than 20 possible mitigating factors, most of which are equally applicable to both murders. For example, both forms ask whether any jurors think that "as a young child, Darryl Lamont Johnson witnessed his father's constant and serious physical abuse of his mother." On one of the forms, zero jurors thought this; on another, two did. There are a number of similar discrepancies. The defendant either did or did not witness his father's abuse of his mother; the answer has nothing to do with which murder the jury was considering.

The general rule, however, the wisdom of which this case substantiates, is that inconsistent findings in a jury verdict do not invalidate the verdict. E.g., United States v. Powell, 469 U.S. 57, 64-69 (1984); United States v. Akram, 152 F.3d 698, 701 (7th Cir. 1998); United States v. Sims, 144 F.3d 1082 (7th Cir. 1998). A jury that inconsistently convicts the defendant of one offense and acquits him of another is as likely to have erred in acquitting him of the one as in convicting him of the other. "Inconsistent verdicts therefore present a situation where 'error,' in the sense that the jury has not followed the court's instructions, most certainly has occurred, but it is unclear whose ox has been gored. Given this uncertainty, and the fact that the Government is precluded from challenging the acquittal, it is hardly satisfactory to allow the defendant to receive a new trial on the conviction as a matter of course." United States v. Powell, supra, 469 U.S. at 65. The principle is applicable here, since the government cannot appeal the jury's refusal to impose the death penalty. 18 U.S.C. sec. 3595(a).

Of course, if the inconsistencies were such as to indicate that the verdict was a product of irrationality, it would have to be set aside. The death penalty statute is explicit about this. "Whenever the court of appeals finds that the sentence of death was imposed under the influence of passion, prejudice, or any other arbitrary factor . . . the court shall remand the case for reconsideration under section 3593 or imposition of a sentence other than death." 18 U.S.C. sec. 3595(c)(2)(A) (emphasis added); United States v. Paul, supra, 217 F.3d at 1004-05; United States v. Webster, supra, 162 F.3d at 354. But that is not a necessary or even the most likely inference in this case. It is more likely that several jurors were of two minds about, say, the relation between the defendant's father and mother, and their irresolution is reflected in their inconsistent answers. But of course jurors disagree among themselves; that is nothing new. They are required to agree about their verdict, not about every fact. What is important here is that every juror who found more mitigating factors present with respect to one of the murders than with respect to the other nevertheless voted that the aggravating factors taken as a whole outweighed the mitigating factors as a whole with regard to each murder. In other words, there is no reason to suspect that any juror was in doubt about the bottom line, though he may have wavered with respect to just how many mitigating factors were present. To put it differently: if a juror couldn't make up his mind whether the defendant had proved 2 mitigating factors or 10 mitigating factors, but was clear in his mind that even in the latter event the aggravating factors outweighed them, there would be no basis for thinking that he had voted irrationally. Wainwright v. Lockhart, 80 F.3d 1226, 1231-32 (8th Cir. 1996).

The verdict form was, however, confusing. It invited inconsistent findings, by listing the mitigating factors twice. For future reference in a multiple-murder capital case, we suggest that the jury be asked about the presence of any mitigating factors that are common to both murders on a separate form rather than on each verdict form. This is assuming that it is proper to make such inquiry of the jury, an issue we leave open because the government waived any objection to the procedure.

Affirmed.