dissenting.
In this case, the Court affirms defendant’s capital-murder conviction and death sentence in the face of significant errors that undermined the reliability of the determinations that resulted in that conviction and that sentence. Those errors, in both the guilt and penalty phases, went to the very heart of the truth-seeking function of our criminal-justice system. In sustaining the conviction and death sentence in spite of those errors, the Court tolerates a degree of irrationality and arbitrariness that we have spent the last decade and a half attempting to eradicate from the imposition of the ultimate punishment in New Jersey.
I
During the guilt phase, the State contended that defendant, after kidnapping and raping L.G., intentionally had killed her. *415Defendant, by contrast, while admitting that he had kidnapped, raped, and caused the death of L.G., insisted that her death had been accidental. Obviously, then, the critical guilt-phase issue— indeed, the only issue truly before the jury — was whether defendant was guilty of purposeful-or-knowing murder, thereby triggering death-eligibility, or whether he was guilty of felony murder but not purposeful-or-knowing murder, which would preclude death-eligibility.
Despite the clarity of that dispute and despite defendant’s presentation of sufficient evidence to allow reasonable jurors to conclude that the death had been accidental, the trial court effectively prevented the jury from fully understanding the consequences of the death-eligibility determination that it was called upon to make. The trial court committed three fundamental errors in that regard: it kept from the jury knowledge of the sentence for felony murder while providing it with full knowledge of the various penalties for purposeful-or-knowing murder; it failed to present the two different forms of murder in a way that would have enabled the jury to consider them in conjunction with one another; and it refused to instruct the jury that it could reach a nonunanimous, non-death-eligible murder verdict split between purposeful-or-knowing murder and felony murder. The trial court’s diminution of felony murder, defendant’s sole defense to capital murder, in conjunction with its singular presentation of the State’s theory, profoundly and irrevocably prejudiced defendant, and requires reversal of his conviction for purposeful-or-knowing murder.
A.
In response to a specific request by defendant, the trial court refused to instruct the jury that, if it found defendant guilty of felony murder (along with kidnapping and aggravated sexual assault) but acquitted him of purposeful-or-knowing murder, he would be sentenced to fifty-five years of parole ineligibility. The court reasoned that it was sufficient to inform the jury that a *416conviction of felony murder without a conviction of purposeful-or-knowing murder would not trigger a penalty phase. Defense counsel then requested that the court simply tell the jury that felony murder and purposeful-or-knowing murder carried the same noncapital sentence, namely, life imprisonment with thirty years of parole ineligibility. The court again declined the request. Finally, defense counsel, in a last-ditch effort, asked if he would be permitted to inform the jury during summation of the sentence for felony murder. The court, however, with the active support of the prosecutor, who argued that defendant was attempting to focus too much attention on felony murder, barred any comment by counsel on the issue. Thus, the jury was kept totally in the dark on the sentencing consequences that would flow from accepting the sole defense theory of the case, namely, that defendant was guilty only of felony murder.
In the trial court’s actual guilt-phase instruction, it informed the jury that if it convicted defendant of purposeful-or-knowing murder, he would receive either thirty years of parole ineligibility or death. Regarding felony murder, however, it simply stated that a felony-murder conviction without a conviction of purposeful-or-knowing murder would not lead to a penalty phase.
Thus, during its guilt-phase deliberations, the jury was fully aware that a conviction of purposeful-or-knowing murder — the State’s theory of the case — would result in either death or, at a minimum, a very long sentence of thirty years of parole ineligibility. It had no such assurance, however, that felony murder — the defense theory — would result in the exact same sentence as a noncapital purposeful-or-knowing-murder conviction, or that it even would lead to a substantial sentence that would adequately punish defendant. Indeed, the jury very well may have believed, given the court’s emphasis of the long sentence for purposeful-or-knowing murder and its failure to say anything about the sentence for felony murder, that felony murder carried a much lesser sentence. That one-sided presentation greatly weakened the defense theory of the case; it very well may have dissuaded the jury, *417or a potential holdout juror who entertained a reasonable doubt about defendant’s intent to kill, from giving serious consideration to a result that would have appeared speculative and uncertain, in that it left open the possibility that it not only would not result in a death sentence but that it would not even assure defendant’s imprisonment for a substantial period of time.
The trial court’s handling of that issue was totally at odds with the established standards that govern the prosecution of capital cases. It is firmly established as a matter of fundamental due process in our capital jurisprudence that juries must understand the “legal effect” of their decisions. It is also firmly settled that felony murder is a non-death-eligible alternative form of homicide to which juries must be allowed to give full consideration.
We consistently have required trial courts to inform capital juries of the legal consequences of their decisions so that they have a complete understanding of the full range of sentencing consequences, thus assuring that their decisions will be rational and informed. In State v. Ramseur, 106 N.J. 123, 311, 524 A.2d 188 (1987), we held that penalty-phase juries, when deciding between life and death, must be instructed about the noncapital sentence for purposeful-or-knowing murder. We stated that
[t]o hide from the jury the full range of its sentencing options, thus permitting its decision to be based on uninformed and possibly inaccurate speculation, is to mock the goals of rationality and consistency required by modem death penalty jurisprudence.
[Ibid.]
Cf. State v. Loftin, 146 N.J. 295, 370, 680 A.2d 677 (1996) (“A capital sentencing jury must be fully informed of its responsibility in determining the appropriateness of the death penalty.”).
Since Ramseur, we have explained that the mandate to inform juries of the legal effect of their decisions includes presentation to the penalty-phase jury of sentences that the defendant currently is serving, State v. Bey, 129 N.J. 557, 603, 610 A.2d 814 (1992) (Bey III), cert. denied, 513 U.S. 1164, 115 S.Ct. 1131, 130 L.Ed.2d 1093 (1995), of the sentences that the court likely will impose on the defendant for any nonmurder convictions in the present case, *418State v. Martini, 131 N.J. 176, 313, 619 A.2d 1208 (Martini I), cert. denied, — U.S. -, 116 S.Ct. 203, 133 L. Ed.2d 137 (1995), and of the “realistic likelihood” that the trial court will, in the present case, impose any noncapital sentences consecutively to the prior sentences, Loftin, supra, 146 N.J. at 372, 680 A.2d 677. Those holdings express the principle that, because the jury must determine the justness of the sentence in light of the defendant’s culpability, the penalty-phase jury must appreciate the consequences of its sentencing decision in order to assure a determination that is rational and not motivated by the apprehension that a nondeath sentence will constitute insufficient punishment. See Loftin, supra, 146 N.J. at 426-27, 680 A.2d 677 (Handler, J., dissenting) (describing juror confusion about the length of time normally served during life sentence).
We have not limited to the penalty phase our repeated holdings that capital juries must understand the legal effect of their decisions, including the sentences attendant to the homicide offenses with which capital defendants are charged. Our capital jurisprudence recognizes that a capital-murder trial, though bifurcated into distinct guilt and penalty phases, is, in actuality, a continuum in which guilt-phase decisions directly impact the imposition of the death penalty. For example, several forms of murder are non-death-eligible, including serious-bodily-injury murder (pre-constitutional amendment), felony murder, and accomplice-liability murder. Recognizing the consequences that the jury’s guilt-phase decisions have on the ultimate penalty determination, we have required trial courts to inform capital juries, during the guilt phase, of those consequences, including sentencing consequences.
In State v. Brown, 138 N.J. 481, 651 A.2d 19 (1994), we held that guilt-phase juries must be instructed that a failure to agree unanimously that a capital defendant committed murder “by his own conduct,” N.J.S.A. 2C:11-3c, would result in a noncapital sentence of between thirty years and life in prison with thirty years of parole ineligibility. In so holding, we stressed the *419relevance of the own-conduct determination to the ultimate penalty decision:
The jury’s final verdict in the penalty phase results either in the imposition of a life sentence with a thirty-year minimum term or a sentence of death. Similarly, when a jury in a capital case decides whether a defendant committed the homicide “by his own conduct,” its determination establishes whether that defendant will be eligible for the death penalty. Although the consequences of the own-conduct determination and the penalty-phase verdict are not identical, the analogy is compelling---- [A] capital-murder defendant may focus his or her efforts in the guilt phase on raising a reasonable doubt about issues that trigger the penalty phase, rather than vigorously contesting guilt or innocence on the murder charge____ We are convinced that, consistent with, their statutory obligation in the penalty phase, the jurors should have been '‘fully informed of the consequences of their votes and the penalties which could result in each eventuality."
[Id. at 517, 651 A.2d 19 (quotations omitted and emphasis added) (quoting Ramseur, supra, 106 N.J. at 309, 524 A.2d 188).]
In State v. Mejia, 141 N.J. 475, 485-86, 662 A.2d 308 (1995), we reiterated our belief, first expressed in Brown, that guilt-phase capital juries must understand the legal effect of their death-eligibility determinations and, accordingly, must be informed of the sentences that defendants would receive upon a conviction of non-death-eligible murder. We stated that “[t]he failure to inform the jury of the difference [in sentences], which could have diluted the jury’s responsibility for the imposition of the death penalty, ... constitutes reversible error.” Ibid, (citations and quotations omitted).
The purpose and logic of our holdings in Brown and Mejia apply fully to felony murder. Like accomplice-liability murder and serious-bodily-injury murder, felony murder is a noncapital form of murder that does not subject a defendant to a penalty-phase trial even though the jury has convicted the defendant of murder. We recognized as much in State v. Purnell, 126 N.J. 518, 530-34, 601 A.2d 175 (1992), in which we held that, where the evidence provides a rational basis for the jury to convict of felony murder, felony murder must be charged to provide the jury with the opportunity to convict of that noncapital alternative form of murder. See also John M. Cannel, New Jersey Criminal Code, Annotated, comment 13 to N.J.S.A. 2C:1-8(e) (1994) (“[I]n a *420capital ease, where there is support in the evidence for a non-capital murder conviction, the jury must be given every opportunity to convict of the charge not carrying the death penalty.”). We stated that “[t]o deprive a capital defendant of a lesser-included alternative murder charge, which arguably would have affected the deliberation of a death sentence, is not constitutionally permissible.” Purnell, supra, 126 N.J. at 532, 601 A.2d 175. We also concluded that “when the separate offense encompassed by [a charged] aggravating factor is, in itself, a basis for an alternative form of murder that is non-capital, a defendant is constitutionally entitled to have that alternative offered for jury deliberation in the guilt phase.” Id. at 534, 601 A.2d 175.
The status of felony murder as an alternative non-death-eligible form of murder, in combination with this Court’s repeated insistence that guilt-phase juries be fully informed of the legal effect of their decisions, including the sentences for noncapital alternative forms of homicide, requires that the jury understand the sentencing consequences of a conviction of felony murder. The failure of the trial court here either to inform the jury of the sentence for felony murder or to allow counsel to do the same constituted clear error. The trial court effectively “hid[] from the jury the full range of its sentencing options, thus permitting its decision to be based on uninformed and possibly inaccurate speculation ... [and] moek[ing] the goals of rationality and consistency required by modern death penalty jurisprudence.” Ramseur, supra, 106 N.J. at 311, 524 A.2d 188.
While implicitly recognizing that the trial court’s actions could not have been harmless because of the significance of felony murder to the accident defense in this case, the Court nevertheless attempts to validate the deliberate withholding of critical information from the jury. It does so, first, by unconvincingly arguing that the 1992 constitutional amendment that overruled our decision in State v. Gerald, 113 N.J. 40, 69-92, 549 A.2d 792 (1988), and that validated application of the death penalty to serious-bodily-injury murder “mooted the Gerald issue that formed the *421basis for the Mejia ultimate-outcome instruction.” Ante at 377, 700 A.2d at 331. The Court thus implies that the reasoning that supports the Mejia holding has been impugned and that Mejia is no longer authoritative. However, the rationale for the Mejia ultimate-outcome instruction — namely, that capital guilt-phase juries must be made aware of the legal effect of their decisions— does not depend on its particular application in that case. The doctrine is both fundamental and unique to capital prosecutions; it is essential because of the nature of the decision to sentence a defendant to death, a decision like no other in the criminal law. We have applied the doctrine in various contexts, including but not limited to serious-bodily-injury murder. E.g., Brown, supra, 138 N.J. at 517, 651 A.2d 19 (accomplice-liability murder). The vitality of the doctrine did not derive solely from the constitutionality of subjecting to the death penalty those who intended only to cause serious bodily injury. Indeed, Mejia was decided several years after the constitutional amendment, and its failure to limit its rationale to pre-constitutional-amendment cases is a convincing indication that the Court did not perceive that the broad ultimate-outcome principle was limited to that restricted context. The Court’s limitation of this aspect of Mejia seriously misperceives and greatly undermines a critical component of our capital jurisprudence.
Apparently recognizing that its logic about the “moot[ing]” of Mejia is insupportable, especially because of Brown’s application of the exact same principle of informing the jury of the legal effect of its guilt-phase decisions, the Court resorts to the only action that it can take to salvage this defective conviction and death sentence, namely, to refuse to apply the inconvenient (albeit well-reasoned) precedents that stand in the way. After dispatching Mejia, the Court converts Brown’s holding that guilt-phase juries must be informed of the noncapital sentence for accomplice-liability murder into “dicta,” which it then overrules. Ante at 377, 700 A.2d at 331.
The Court attempts to justify its shift away from informing juries of the legal effect of their guilt-phase decision by asserting *422that “[generally, juries in criminal cases are not informed of the consequences of returning guilty verdicts.” Ante at-. The concurrence further elaborates on the Court’s new-found equation of capital and noncapital criminal proceedings, stating that
[e]ven in a capital case, during the guilt phase the jury should not be distracted from its function by specific sentencing information about noncapital homicide charges that could intrude on the jury’s deliberative responsibility to return verdicts on the defendant’s guilt or innocence of the charged crimes, as well as to determine death-eligibility.
[Ante at 411-412, 700 A.2d at 349 (Stein, J., concurring).]
What both the Court and the concurrence utterly fail to grasp is that a capital jury’s guilt determinations — unlike a noncapital jury’s guilt determinations — are inherently linked to its sentencing decisions, thus making information about the relevance of guilt-phase decisions to the ultimate punishment anything but “distract[ions].”
In relying on our noncapital rule that juries should not be told about the sentencing consequences of their guilt determinations, in order to justify the same rule in the capital context, the Court ignores the fact that both the Legislature and this Court repeatedly have determined that the taking of human life by the State is so different from other forms of punishment that procedures unique to capital trials and unheard of in noncapital trials are appropriate. The Legislature, for example, has vested exclusive and mandatory appellate jurisdiction in this Court, N.J.S.A. 2C:11-3e; has determined that juries, not judges, generally should impose sentence, N.J.S.A. 2C:11-3c(1); has provided incredible leeway to defendants in the presentation of mitigating evidence, N.J.S.A. 2C:11-3c(5)(h); and has provided for nonunanimous penalty verdicts that redound to defendants’ benefit, N.J.S.A. 2C:11-3c(3)(c). None of those extraordinary procedures exist in the noncapital criminal context.
This Court, as well, has treated capital cases differently from noncapital cases. E.g., Loftin, supra, 146 N.J. at 339, 680 A.2d 677 (“While an in banc voir dire is ordinarily deemed adequate to ensure an impartial jury, we insist on an individualized voir dire *423for capital cases because of the range of discretion entrusted to a jury in a capital sentencing hearing.... Because juries have so much more discretion, there is a greater need to screen out those jurors who cannot be impartial.”) (quotations and citations omitted); State v. Biegenwald, 106 N.J. 13, 30, 524 A.2d 130 (1987) (Biegenwald II) (“[W]e note that in capital cases trial courts should be especially sensitive to permitting attorneys to conduct some voir dire.”); Ramseur, supra, 106 N.J. at 324, 524 A.2d 188 (“Because death is a uniquely harsh sanction, this Court of necessity will more readily find prejudice resulting from prosecutorial misconduct in a capital case than in other criminal matters....”). The Court’s determination today that, because noncapital juries are not told the full legal effect of their decisions, capital juries should be deprived of that information makes no sense in light of the fundamental premise undergirding modern capital jurisprudence, namely, that death is different.
Yet, however misguided the Court’s holding, it is not fatal to capital defendants who seek death-eligibility determinations by juries that understand the import of their guilt-phase death-eligibility determinations. Fortunately, the Court preserves, for future cases,1 all aspects of full jury knowledge of the legal effect of guilt-phase decisions except for the exact number of years that non-death-eligible forms of murder carry. For example, trial courts must tell juries which forms of homicide do and do not lead to a penalty phase and that all of the homicide offenses are “extremely serious” and carry “severe prison sentences.” Ante at *424378-379, 700 A.2d at 331-332. Implicit in the Court’s holding is a requirement that trial courts also inform juries that the noncapital sentences for certain forms of non-death-eligible and death-eligible murder are equivalent (accompanied, of course, by an instruction not to consider the comparative lengths of the sentences in determining guilt or lack thereof).
Obviously, just as trial courts must not tell juries the numerical sentence for noncapital forms of homicide, nor are they permitted to inform juries of the noncapital sentence for death-eligible purposeful-or-knowing murder (i.e., thirty years to life). In other words, a jury determining guilt must not be informed of the sentence for purposeful-or-knowing murder that would result from a penalty-phase determination that death is not the appropriate sentence. If the guilt-phase jury were to know the noncapital sentence for death-eligible murder without knowing the sentence for non-death-eligible homicide, an obvious inequity would result because the jury could be induced to convict of death-eligible murder, which would be a known quantity, in order to assure a long sentence. As the Court today recognizes, juries should not be left to speculate about the results of their guilt-phase determinations. Ante at 378, 700 A.2d at 332. Not informing juries about the exact length of sentences for either death-eligible or non-death-eligible murder while telling them that, whatever the form of murder of which they convict, the sentence will be severe (and, in the case of purposeful-or-knowing murder and felony murder, the noncapital sentences will be equivalent) will allow the jury to fulfill its function without being misled by knowing the sentence for some forms of homicide but not
* * *
Despite the limits that the Court places on its overruling of our well-reasoned precedent, its refusal to apply the ultimate-outcome rule in this case constitutes a sharp departure from our firm tradition and consistent practice of placing primary emphasis on the procedurally and substantively fair application of the death penalty. The Court effectively holds that a rule going directly to *425the heart of the truth-seeking process should not be applied to this ease, despite the essentiality of the rule to the defense theory and despite defense counsel’s repeated invocation of the rule.
The real issue, I submit, is not about the continued vitality of Mejia or whether the ultimate-outcome issue was briefed in Brown and Mejia, ante at 411, 700 A.2d at 349 (Stein, J., concurring). Instead, it is about a capital defendant’s right to present his defense and not to have the jury kept ignorant and clueless about the consequences of accepting that defense, especially when the jury has been fully apprised of the consequences of accepting the State’s theory. Cf. State v. Martin, 119 N.J. 2, 16-17, 573 A.2d 1359 (1990) (reversing conviction based on trial court’s failure to mold instruction to fit defense theory of case despite its tailoring of instructions to track State’s theory). If our longstanding insistence that capital juries understand the “legal effect” of their decisions means anything, it means that this defendant cannot be executed based on the verdict of a jury left entirely in the dark about a crucial aspect of the case.
B.
The Court also deflates the doctrine of noncapital alternative murder offenses, a doctrine that is fundamental to the fair and constitutional administration of our system of capital punishment. Based on the competing theories of intent and accident, defendant requested that the trial court instruct the jury that it should evaluate purposeful-or-knowing murder and felony murder simultaneously and in tandem with one another. The purpose of that request was to enable the jury to consider both forms of murder without initially tilting in favor of one over the other. The trial court, however, refused the request, instead instructing the jury that it should consider felony murder only after deliberating about the State’s theory of purposeful-or-knowing murder. It may be the case that, under all the circumstances, the trial court’s sequential presentation of the two forms of murder was harmless error because the court clearly informed the jury that it could convict of *426felony murder while acquitting of purposeful-or-knowing murder and that such a result would avoid a penalty phase. Nevertheless, I strongly disagree with the Court’s (mis)treatment of the vital due-process doctrine of non-death-eligible alternative murder offenses.
Although we traditionally have required juries to consider lesser-ineluded offenses only after they acquit of the greater offense, e.g., State v. Harris, 141 N.J. 525, 552-53, 662 A.2d 333 (1995), we have been much more skeptical of the sequential presentation of offenses where the “lesser” offense is actually an alternative form of homicide that avoids a penalty phase. Our skepticism about sequential presentation of noncapital alternative offenses has stemmed from the risk that juries may give undue attention to the first offense and may not consider the second offense until their minds are made up and they already have determined to convict of the first, death-eligible, offense. We have noted that “[o]ne problem with a sequential charge is that it may cause a jury that believes a defendant guilty of something to convict on the first and most serious charge.” Mejia, supra, 141 N.J. at 484, 662 A.2d 308. In the context of the death penalty, that risk is simply unacceptable because of the high stakes involved.
Thus, in State v. Coyle, 119 N.J. 194, 221-24, 574 A.2d 951 (1990), we held that, where a rational basis exists for the jury to convict a capital defendant of passion/provocation manslaughter, which is not death-eligible, instead of death-eligible purposeful-or-knowing murder, the two must be presented simultaneously so that the jury does not consider whether the defendant had an intent to kill without also considering whether passion or provocation actually caused the killing. We noted that sequential presentation
had the potential to foreclose jury consideration of whether passion/provocation should reduce an otherwise purposeful killing from murder to manslaughter. Thus, despite the evidence of passion/provocation in the record, the jury may have convicted defendant of murder simply by finding that “it [was] his conscious object to cause death or serious bodily injury,” without having considered the possibility of a manslaughter verdict.
[Id. at 222-23, 574 A.2d 951.]
*427In Mejia, supra, 141 N.J. at 483-85, 662 A.2d 308, we made clear that Coyle’s condemnation of sequential charges in the capital context was not limited to passion/provocation manslaughter. We noted that, like passion/provocation manslaughter,
serious-bodily-injury murder is an alternative form of homicide, not a lesser-included offense of “intent to kill” murder.... Under N.J.S.A. 2C:11-3a, murder is defined as a criminal homicide committed by an actor who knowingly or purposely intends to cause death or serious bodily injury resulting in death. Thus, one who kills with either intent is a murderer.
[Id. at 484, 662 A.2d 308.]
Mejia and Coyle stand for the proposition that where a rational basis exists for the jury to convict a capital defendant of a non-death-eligible alternative form of homicide, not only must the trial court charge that offense, State v. Koedatich, 112 N.J. 225, 340, 548 A.2d 939 (1988) (Koedatich I) (accomplice-liability murder), cert. denied, 488 U.S. 1017, 109 S.Ct. 813, 102 L. Ed.2d 803 (1989); State v. Harvey, 121 N.J. 407, 413, 581 A.2d 483 (1990) (Harvey I) (serious-bodily-injury murder), cert. denied, 499 U.S. 931, 111 S.Ct. 1336, 113 L.Ed.2d 268 (1991); State v. Grunow, 102 N.J. 133, 145, 506 A.2d 708 (1986) (passion/provocation manslaughter), but it must charge it in such a way that the jury will consider it simultaneously with its consideration of death-eligible purposeful- or-knowing murder. Because, under New Jersey law, felony murder is also a non-death-eligible alternative form of homicide, Purnell, supra, 126 N.J. at 530-34, 601 A.2d 175, it also must be presented to the jury simultaneously so that the jury will give it due consideration.
Despite the clarity of our precedents, the Court concludes that felony murder is no more than a traditional lesser-included offense of purposeful-or-knowing murder and thus holds that it may be charged sequentially, regardless of its significance in a particular case. Ante at 363-370, 700 A.2d at 324-327. That conclusion cannot be sustained. In Purnell, supra, in explaining the standard under which trial courts should determine whether to charge felony murder in the first place, we referred to lesser-included *428offenses by analogy. 126 N.J. at 581, 601 A.2d 175. We made it very clear, however, that the significance of felony murder went beyond that of a traditional lesser-ineluded offense. We stated that “when the separate offense encompassed by the aggravating factor is, in itself, a basis for an alternative form of murder that is non-capital, a defendant is constitutionally entitled to have that alternative offered for jury deliberation in the guilt phase.” Id. at 534, 601 A.2d 175. Thus, in Purnell, we specifically pointed to the non-death-eligible status of felony murder as the significant factor in our requirement that it be charged if a rational basis exists for the jury to convict of it. Given that the only difference, in terms of penal consequences, between purposeful-or-knowing murder and felony murder is that one carries the potential for a death sentence whereas the other does not (they both carry noncapital sentences of thirty years to life, N.J.S.A. 2C:11-3b), Purnell’s holding that felony murder must be presented to the jury would be superfluous and unnecessary if felony murder were viewed only as a traditional lesser-ineluded offense and not as a means for the jury to convict of a noncapital offense.2 The Court misperceives that essential basis of Purnell.
Felony murder’s status as a true alternative form of murder is confirmed by the structure.of the murder statute, which defines “murder” as killing accompanied by purpose to cause death or serious bodily injury, knowledge that death or serious bodily injury will occur, or the intent to engage in an enumerated predicate felony. N.J.S.A. 2C:11-3a. Felony murder is thus a *429coequal form of murder that should be charged as such and not relegated to the status of a lesser offense.
The Court, referring to the Coyle decision, asserts that simultaneous charging of purposeful-or-knowing murder and passion/provocation manslaughter is necessary because the mental states of the two are “interrelated” and “shade[ ] from one into the other.” Ante at 369, 700 A.2d at 327. By contrast, says the Court, “there is no connection between the required mental state for purposeful-or-knowing murder and that for felony murder, the latter being a strict-liability offense.” Ibid. Yet, as the facts here demonstrate, the mental states underlying purposeful-or-knowing murder and felony murder can be quite “interrelated.”
In this case, defendant killed either intentionally or unintentionally. Just as the jury in Coyle could not convict of purposeful-or-knowing murder without finding a lack of passion/provocation and just as the jury in Mejia could not find the existence of a specific intent to kill without concluding that the defendant did not simply possess an intent to cause serious bodily injury, the jury in this ease could not conclude that defendant had intended to kill L.G. without finding an absence of accident. As in Coyle and Mejia, the mental state for the non-death-eligible alternative offense in this case was inextricably bound up with the mental state for purposeful-or-knowing murder and should have been presented as such.
Indeed, the analogy between a passion/provocation defense and an accident/felony-murder defense is compelling. In the passion/provocation context, the defendant admits that he caused the death of the victim but attributes the killing to passion or provocation; thus, the dispute between the prosecution and the defense boils down to whether the defendant killed the victim in a premeditated, deliberate manner or whether he did so only after being provoked in some way. Likewise, in the accident/felony-murder context, the defendant admits that he caused the death of the victim but attributes the killing to an accident; thus, the dispute between the prosecution and the defense boils down to whether *430the defendant killed the victim in a premeditated, deliberate manner or whether he did so accidentally. In both contexts, the jury is confronted with a choice of premeditation or lack thereof, and, in both contexts, the consequences of the jury’s choice is enormous, namely, whether a penalty phase will be triggered. How the Court can dismiss that similarity is puzzling.
Obviously felony murder and other murder offenses need not be presented as coequal alternatives to purposeful-or-knowing murder in all cases, regardless of the evidence. A rational basis must exist in order for the jury to consider felony murder as a viable option. Trial courts should evaluate the facts and circumstances of the evidence and defenses presented in determining whether alternative homicide offenses should be presented simultaneously to death-eligible murder. In this ease, such a rational basis existed.
The trial court nevertheless ignored that requirement and instructed the jury that it should not deliberate on felony murder without first deliberating on — and, implicitly, acquitting or convicting of — purposeful-or-knowing murder. That presentation could have exerted a coercive influence on the jury by inducing it to consider purposeful-or-knowing murder without giving equal consideration to defendant’s sole defense, namely, that the killing had been accidental and that he consequently was guilty only of felony murder. The error, however, may have been harmless because the trial court clearly informed the jury that it could convict of felony murder while acquitting of purposeful-or-knowing murder and that such a result would avoid a penalty phrase.
Yet, the harmlessness of the error does not justify the Court’s disposition of this claim. Jurors must be provided “every opportunity to convict of the charge not carrying the death penalty.” Cannel, supra. Moreover, jurors must possess a clear understanding of the “legal effect” of their guilt-phase determinations, including their effect on death-eligibility. Mejia, supra, 141 N.J. at 485, 662 A.2d 308; Brown, supra, 138 N.J. at 517, 651 A.2d 19. Although the Court wisely reaffirms Purnell and the essentiality *431of charging non-death-eligible alternative homicide offenses, ante at 358-365, 369, 700 A.2d at 321-325, 329, by fashioning a per se rule that felony murder is nothing more than a traditional lesser-included offense of purposeful-or-knowing murder that can be given secondary status regardless of the facts of the case, the Court does great disservice to those aspects of our capital jurisprudence that are designed to promote sound and informed decisions.
C.
The Court also rejects the possibility of a nonunanimous murder verdict in which jurors are divided between purposeful-or-knowing murder and felony murder. Under that nonunanimity theory, if some jurors believe that a defendant intended to kill while other jurors believe that he killed accidentally in the course of a felony and, hence, is guilty of felony murder but not purposeful-or-knowing murder, then the result is a non-death-eligible murder conviction. The Court’s conclusion is inconsistent with our requirement that juries have “every opportunity to convict of the charge not carrying the death penalty.” Cannel, supra.
New Jersey’s murder statute has a distinctive structure in that it initially was drafted without capital punishment in mind. When the Legislature later adopted the death penalty in 1982, it simply engrafted death-eligibility onto the already-existing definition of murder. See Gerald, supra, 113 N.J. at 89-90, 549 A.2d 792 (noting discrepancies between statements by key legislators in enactment of death penalty and actual structure of statute once death-eligibility provisions were incorporated); John J. Farmer, Jr., The Evolution of Death-Eligibility in New Jersey, 26 Seton Hall L.Rev. 1548, 1564-72 (1996). The statute defines “murder” as death resulting from one of five mental states: intent to cause death; intent to cause serious bodily injury; knowledge that death will occur; knowledge that serious bodily injury will occur; or the mental state necessary to support a predicate felony giving rise to felony murder. N.J.S.A. 2C:11-3a. Those five forms of “murder” are not distinguished by degree, and they all carry the same *432noncapital sentence, namely, between thirty years and life imprisonment with thirty years of parole ineligibility. N.J.S.A. 2C:11-3b.
After comprehensively defining “murder” without distinguishing among the five mental states that may give rise to it, the statute defines death-eligibility. N.J.S.A. 2C:11-3c. It does so by simultaneously exempting felony murder and most forms of accomplice-liability murder. Ibid. This Court added a class of cases to that exemption when it invalidated the application of the death penalty to murderers who did not specifically intend to kill their victims or have knowledge that death would occur. Gerald, supra, 113 N.J. at 69-92, 549 A.2d 792. Although that decision was superseded by constitutional amendment and legislation, L. 1993, c. 111, § 1 (codified at N.J.S.A. 2C:11-3i), Gerald still exempts from death-eligibility any murder unaccompanied by a specific intent to kill that occurred before December 3, 1992. Loftin, supra, 146 N.J. at 349, 680 A.2d 677.
With the superimposition of death-eligibility onto the murder statute, the Court was faced with the question of how to distinguish between death-eligible and non-death-eligible murder. After all, while the murder statute equated a number of mental states as various forms of the general crime of “murder,” the death-eligibility provision of the statute distinguished among those mental states and made even further distinctions (e.g., principal versus accomplice).
The Court responded by emphatically relying on the original definition of “murder,” without distinguishing among the differing mental states, and by viewing the death-eligibility provision as a simple penalty-phase triggering mechanism, without relevance to the underlying substantive definition of “murder.” As we stated in Mejia, supra:
Like the “by your own conduct” requirement, the “intent to kiH” requirement “is not an element of the offense of murder [but is] merely a triggering device for the death-penalty phase of the trial.”
[141 N.J. at 486, 662 A.2d 308 (quoting Gerald, supra, 113 N.J. at 99, 549 A.2d 792) (citations and quotations omitted).]
*433See also Brown, supra, 138 N.J. at 520, 651 A.2d 19 (“Although murder indictments must specify whether the murder is alleged to have been committed by the defendant’s own conduct, ... the purpose of that requirement is only to indicate whether the alleged crime is one punishable by death.”). Consequently, we have not required jury unanimity with respect to the particular theory or factual basis underlying a murder conviction because the murder statute makes no such distinction in its definition of “murder” and because more than one theory or factual basis can constitute “murder.” Instead, we simply have required that all jurors agree that the defendant is guilty of some form of murder, that is, murder in one of its prescribed forms, as defined by N.J.S.A. 2C:11-3a. E.g., Mejia, supra, 141 N.J. at 486, 662 A.2d 308 (“A jury need not agree unanimously on a defendant’s mental state when a finding on one of several alternative mental states will satisfy the relevant statutory requirement.”); Brown, supra, 138 N.J. at 520, 651 A.2d 19 (“[T]he accepted view is that to return a criminal conviction, a jury need not be unanimous on the theory of criminal-conduct responsibility if the alternative theories apply to commission of the same criminal act and each of them supports conviction of the same offense....”); Bey III, supra, 129 N.J. at 582, 610 A.2d 814 (“That a jury agree unanimously that the defendant’s state of mind was either purposeful or knowing is sufficient to find a defendant guilty of capital murder.”).
It is only in the context of a capital prosecution when the jury has agreed unanimously that the defendant is guilty of “murder” under N.J.S.A. 2C:11-3a that the jury’s agreement or lack thereof about the particular theory underlying the murder conviction becomes relevant in determining whether a penalty phase will be triggered pursuant to N.J.S.A. 2C:11-3c. If the jury unanimously agrees on a death-eligible theory (e.g., intent to kill by one’s own conduct) or is split among more than one death-eligible theory (e.g., intent to kill by one’s own conduct and knowledge that death will result from one’s own conduct), then the verdict triggers a penalty phase. If fewer than all of the jurors agree on a death-eligible theory, however, then a penalty phase does not occur. See *434Brown, supra, 138 N.J. at 511, 651 A.2d 19 (“[T]he inability of the jury to reach a unanimous decision on the own-conduct determination constitutes a final verdict that results in the imposition of a sentence of imprisonment of at least a thirty-year mandatory term....”).
Given that established decisional background, the question in this ease is whether felony murder should be arbitrarily excluded from our nonunanimity jurisprudence despite the inclusion of every other form of “murder” as defined in N.J.S.A. 2C:11-3a. The Court concludes that it should be excluded because of a perceived legislative intent underlying the murder statute and because of the “chaos” that would occur if felony murder were included.
The Court’s reliance on legislative intent is misplaced. It states that
[t]he fact that the Legislature has established the identical sentence for noncapital purposeful-or-knowing murder as it has for felony murder should not be determinative of whether to require a unified-murder charge. It is the culpable mental state plus the aggravating circumstances that determine death-eligibility and that also distinguish capital murder from felony murder.
[Ante at 360, 700 A.2d at 322.]
After noting that “felony murder is an absolute liability crime because a defendant need not have contemplated or intended the victim’s death,” ibid., the Court concludes that “the Legislature intended that death-eligibility be viewed as the touchstone of moral equivalence”3 in determining which forms of murder can be *435combined to form a nonunanimous murder verdict. Ante at 361, 700 A.2d at 322.
The Court bases its holding — that a non-death-eligible theory of murder cannot combine with a death-eligible theory to form a general murder conviction — on the conclusion that death-eligible and non-death-eligible theories are not moral' equivalents. The support for that conclusion, however, is simply that death-eligibility is the touchstone of moral equivalence, a proposition that serves as both the Court’s premise and its conclusion. The Court thus begs the key question of why death-eligibility, as opposed to other aspects of the murder statute’s structure, is the “touchstone” of moral equivalence.
The Court’s assertion finds equivocal support in the statute’s language and history and no support in this Court’s jurisprudence. The murder statute, as noted, groups all five states of mind together in the category “murder” and provides that all are punishable by the same term of imprisonment. It then exempts felony murder and most forms of accomplice-liability murder from death-eligibility. That structure, standing alone, does not lead to the inevitable conclusion that death-eligibility is the “touchstone” of moral equivalence. If anything, the fact that the Legislature, in enacting the death penalty, maintained the unified definition of murder in N.J.S.A. 2C:11-3a, instead of dividing that definition of murder into “capital” and “noncapital” murder, suggests that it intended for the various forms of murder to remain morally equivalent.
Concededly, the murder statute is ambiguous about whether the Legislature intended for death-eligible and non-death-eligible forms of murder to be moral equivalents. On the one hand, the Legislature grouped all forms of murder together in the definition of “murder” and imposed the same noncapital punishment for them, N.J.S.A. 2C:11-3a & b; on the other hand, it distinguished among them in terms of death-eligibility, N.J.S.A. 2C:11-3c. *436Thus, it is plausible to look to either section of the statute to discern a legislative intent regarding moral equivalence.
The ambiguity in the statute, however, is fully clarified by this Court’s repeated rejection of death-eligibility as the sole determinant or “touchstone” of moral equivalence. In Brown, supra, we held that the jury could reach a nonunanimous verdict about whether the defendant was a (death-eligible) principal or (non-death-eligible) accomplice in the murder. 138 N.J. at 511-18, 651 A.2d 19. Likewise, in Mejia, supra, we validated nonunanimous murder verdicts split between (non-death-eligible) serious-bodily-injury murder and (death-eligible) intent-to-kill murder. 141 N.J. at 486-87, 662 A.2d 308. In Mejia, we based our holding on the irrelevance of the defendant’s specific state of mind to the actual “murder” conviction, instead stressing that as long as all jurors agreed that one of the requisite states of mind for murder was present, a murder conviction resulted; state of mind came into play only to determine whether the jury unanimously had agreed on a death-eligible state of mind, thereby triggering a penalty phase. Id. at 486, 662 A.2d 308; see also Brown, supra, 138 N.J. at 520, 651 A.2d 19; Gerald, supra, 113 N.J. at 99, 549 A.2d 792.
Given our explicit holdings and reasoning in Brown and Mejia, the Court’s conclusion that death-eligibility is the sole “touchstone” of moral equivalence and that felony murder is thus “different” from other forms of murder is totally unconvincing. See Brown, supra, 138 N.J. at 521, 651 A.2d 19 (citing to case that permitted mixing of premeditated and felony murder). Even if the Court were incorrectly to conclude that Brown is distinguishable in that the “own-conduct” determination is independent of the statute’s definition of “murder,” it cannot make that claim about the jury’s intent-to-kill determination, which is clearly part of the definition of “murder” in N.J.S.A. 2C:11-3a. The Court cannot avoid the fact that we already have concluded that the Legislature’s definition of “murder” in N.J.S.A. 2C:11-3a is the touchstone of moral equivalence. Nor can it avoid our repeated holdings that the murder statute defines “murder” as encompassing a *437number of different states of mind and that the jury need not agree on one particular state of mind to support a murder conviction, as opposed to death-eligibility.
In order to avoid that conclusion, however, the Court chooses to rely on assertion to support its claim, stating that
[b]ecause the Legislature has decreed that felony murder is not a death-eligible offense and our capital jurisprudence, stressing the importance of providing a jury with every opportunity to spare a defendant’s life, has been limited to purposeful- or-knowing-murder cases in which the jury had to resolve a factual dispute determinative of whether that murder was death-eligible, we decline to extend that jurisprudence to noncapital murder. Those cases have been restricted to capital murder as defined in N.J.S.A. 2C:11-3c for which notice of aggravating factors has been given. N.J.S.A. 2C:11-3c(2)(e).
[Ante at 361-362, 700 A.2d at 323.]
Again, the question here is why “our capital jurisprudence, stressing the importance of providing a jury with every opportunity to spare a defendant’s life, has been limited to purposeful-or-knowing-murder cases in which the jury had to resolve a factual dispute determinative of whether the murder was death-eligible____” And, again, the Court provides no satisfactory response, instead simply falling back on its arbitrary and unsupported assertion that felony murder is different from non-death-eligible forms of purposeful-or-knowing murder.
The Court comes closest to attempting to justify its excision of felony murder from its unambiguously coequal place in the statute’s definition of murder by relying on felony murder’s “different” mens rea, namely, its status as an “absolute liability crime.” Ante at 360, 700 A.2d at 322. That reliance is undermined, however, by our inclusion of serious-bodily-injury murder in our nonunanimity jurisprudence. Serious-bodily-injury murder is, in reality, a form of “absolute liability” felony murder, with aggravated assault as the underlying felony. See Gerald, supra, 113 N.J. at 88, 549 A.2d 792 (“In that respect [i.e., felony murder only requires an intent to commit the underlying felony as opposed to an intent to kill] felony-murder is strikingly similar to the crime at issue here, the purposeful or knowing infliction of serious bodily injury resulting in death.”). The Court, however, has not excluded *438that “absolute liability” form of murder from its nonunanimity jurisprudence, thus easting doubt on the Court’s new-found distinction between absolute-liability and non-absolute-liability forms of murder.
Moreover, felony murder’s requisite mens rea is not necessarily less culpable than other mens reas. All of the predicate felonies to felony murder (robbery, sexual assault, arson, burglary, kidnapping, and criminal escape) involve either violence or high risk of violence against others; indeed, for a felony murder to occur, the felony must result in the ultimate form of violence, namely, death. Our requirement that, for causation to exist in the felony-murder context, death must have been the “probable consequence of the defendant’s conduct,” Martin, supra, 119 N.J. at 33, 573 A.2d 1359, is further evidence of the high level of indifference to the value of human life that we require to sustain a felony-murder conviction. The distinction between a violent rape or robbery resulting in death and a violent assault that results in death is unclear at best.
As in the case of the Court’s emphasis on death-eligibility as the “touchstone” of moral equivalence, its reliance on mens rea as an indicator of moral culpability is a sharp break from the past and not dictated by the structure of the murder statute. We long have allowed nonunanimous murder verdicts split between and among various mental states (e.g., purpose/knowledge, intent to kill/intent to cause serious bodily injury). Moreover, the statute’s definition of murder in N.J.S.A. 2C:11-3a equates all of the violent mental states as “murder,” thereby undermining the putative legislative intent that the Court reads into the statute. Consequently, the impression is unmistakable that the Court is engaging in jurisprudential and statutory revisionism in order to sustain this death sentence.
The Court also points to the “nearly-universal hallmark[ ] of ... the requirement of a unanimous verdict in criminal cases” and asserts that “the intermingling of death-eligible murder with non-death-eligible murder would create utter chaos when instructing a jury.” Ante at 361, 700 A.2d at 323 (quotations and citation *439omitted). The Court then insinuates that a murder conviction split between felony murder and purposeful-or-knowing murder would fail to lead to a guilty verdict for murder, stressing “[t]he public’s right to see that a criminal prosecution proceeds to a verdict either of acquittal or conviction” and that a nonunanimity instruction “would lead a jury down the wrong path ... to a verdict not permitted under our law.” Ibid.
In concluding that the felony-murder nonunanimity instruction would be “chao[tic]” and somehow unconstitutional, the Court, again, completely ignores our nonunanimity jurisprudence, the nonunanimity practice of other states, and the United States Supreme Court’s explicit validation of such verdicts. In Brown, supra, 138 N.J. at 511-18, 651 A.2d 19, and Mejia, supra, 141 N.J. at 486-87, 662 A.2d 308, we explicitly required “the intermingling of death-eligible murder with non-death-eligible murder,” ante at 361, 700 A.2d at 323, and, to my knowledge, neither “utter chaos,” ibid., nor unconstitutional verdicts have occurred. In fact, in Mejia, supra, we specifically rejected the argument that nonunanimity in that context would violate the “bedrock principle of jury unanimity....” 141 N.J. at 486, 662 A.2d 308 (quotations omitted); see also Brown, supra, 138 N.J. at 518, 651 A.2d 19 (noting that, as in the penalty phase, a nonunanimous guilt-phase verdict regarding death-eligibility does not result in a hung jury or mistrial, instead resulting in a “valid murder conviction”).
Moreover, as the Court notes, numerous other states allow nonunanimous murder verdicts split between premeditated and felony murder, ante at 359, 700 A.2d at 322; see Schad, supra, 501 U.S. at 641-42, 111 S.Ct. at 2502, 115 L. Ed.2d at 571-72 (canvassing states that allow such nonunanimous verdicts), including Washington and Kansas, where felony murder is not death-eligible, ante at 359 n. 1, 700 A.2d at 322 n 1. As one court has stated:
[Premeditated murder and felony murder are not separate or different offenses. The statute merely provides alternative methods of proving the deliberation and premeditation required for a conviction of first degree murder____ A prosecution under the felony murder rule merely changes the type of proof necessary to establish a violation of the statute. Proof that a homicide was committed in the *440perpetration of a felony is tantamount to premeditation and deliberation which otherwise would be necessary to constitute murder in the first degree.
[State v. McCowan, 226 Kan. 752, 602 P.2d 1363, 1370-71 (1979), cert. denied, 449 U.S. 844, 101 S.Ct. 127, 66 L.Ed.2d 53 (1980).]
Finally, the United States Supreme Court has approved of murder verdicts split between premeditated and felony murder, Schad, supra, 501 U.S. at 630-45, 111 S.Ct. at 2496-2504, 115 L. Ed.2d at 564-74 (plurality opinion); id. at 648-52, 111 S.Ct. at 2505-07, 115 L.Ed.2d at 576-78 (Scalia, J., concurring in part and in judgment), which should assuage the Court’s fear that such verdicts are “not permitted under our law.” Ante at 363, 700 A.2d at 324 (quotations and citation omitted).
Regarding the “utter chaos” that a felony-murder nonunanimity instruction allegedly would cause, resulting in the “reversal of any murder or felony-murder conviction,” ante at 363, 700 A.2d at 324, the Court’s fears are greatly exaggerated. As noted, numerous states permit nonunanimous verdicts split between felony murder and premeditated murder and have done so for many years. See Schad, supra, 501 U.S. at 651, 111 S.Ct. at 2507, 115 L.Ed.2d at 577-78 (Scalia, J., concurring in part and in judgment) (“Submitting killing in the course of a robbery and premeditated killing to the jury under a single charge ... was the norm when this country was founded ... and remains the norm today.... [It is] a practice as old as the common law and still in existence in the vast majority of States....”). One imagines that if “utter chaos” had resulted, many states would have abandoned the practice.
The Court’s chaos argument is further undermined by the proposed verdict sheet that defendant submitted in this case, which is quite clear and comprehensible. Slightly modified, it states:
MURDER
We the jury unanimously find the defendant
not guilty of murder_
guilty of murder_
(defendant caused the death of L.G, purposely or knowingly, or he caused the death of L.G. during the commission of or attempted commission of, or flight *441after committing or attempting to commit the crime of kidnapping and/or sexual assault)
If your answer is not guilty, go to AGGRAVATED MANSLAUGHTER on the next page.
If your answer is guilty answer the following questions:
SELECT ONE OF THE FOLLOWING:
1. We find unanimously that the defendant purposely or knowingly committed the murder._
2. We find unanimously that the defendant caused the death of L.G. during the commission of or attempted commission of or flight after committing or attempting to commit the crime of kidnapping and/or sexual assault.
3. We unanimously find both 1 and 2__
4. We are unable to agree unanimously on either 1, 2, or 3, but all 12 of us find 1, 2, or 3 and at least 1 of us finds 2 but not 1 or 3. _
That verdict sheet is entirely understandable and provides the jury with a clear framework within which to work. It allows the jury, once it has reached a unanimous determination that the defendant is guilty of “murder,” to decide whether the murder was (1) purposeful-or-knowing murder but not felony murder, (2) felony murder but not purposeful-or-knowing murder, (3) both purposeful-or-knowing murder and felony murder, or (4) one or the other, but not both. Thus, the jury can implement our consistent description of the murder statute as defining the general crime of “murder” followed by a triggering mechanism (i.e., mens rea) to determine death-eligibility.
Moreover, the fact that a jury can convict of both purposeful-or-knowing murder and felony murder because they are not mutually exclusive is of no moment. Although accomplice-liability murder and principal-liability murder may be mutually exclusive, neither intent-to-kill murder and serious-bodily-injury murder nor purposeful murder and knowing murder are mutually exclusive. If one intends to kill, then she also intends to cause serious bodily injury, and if one has the purpose of killing, then she also possesses the knowledge that death will result if she completes her purpose. That two forms of murder may overlap has not been dispositive in our nonunanimity jurisprudence. Instead, we have focused on the difference in the “legal effect” (i.e., death-eligibility *442or lack thereof) of the different theories in determining whether a nonunanimity instruction is required.
Despite the ability of trial courts to inform and to guide juries about the possibility of reaching a nonunanimous verdict regarding death-eligibility, the Court determines that juries must reach unanimity on each theory of murder in order to produce a valid murder conviction. When combined with the Court’s erroneous holding that it is permissible to require juries to convict or acquit of purposeful-or-knowing murder prior to considering felony murder, ante at 363-370, 700 A.2d at 324-327, the unanimity requirement may cause holdout jurors, who believe that a defendant is guilty of some sort of murder but not necessarily purposeful-or-knowing murder, to cave in to majority pressure and to convict of purposeful-or-knowing murder rather than risk not convicting of any form of murder.
The Court’s rejection of the nonunanimity instruction is inconsistent with the murder statute, our capital jurisprudence, principles of fairness and due process, and the practice of many other states. Similar to the Court’s treatment of the ultimate-outcome instruction and simultaneous presentation of the two forms of murder, its analysis and dismissal of this claim have no support in either logic or precedent.
II
Although defendant’s guilt-phase death-eligibility process was fundamentally flawed, death-eligibility, by itself, does not determine whether a capital defendant will be sentenced to death. A defendant who is denied a fair determination of death-eligibility still has the opportunity to persuade the penalty-phase jury that his life should be spared. In this case, however, an egregious error by the trial court in submitting mutually exclusive aggravating factors to the jury, prosecutorial misconduct in encouraging and compounding that error, and the accidental submission of a gruesome autopsy photograph to the jury, combined to inject a degree of arbitrariness and irrationality into the penalty phase *443that no court should tolerate. Paralleling its treatment of the guilt-phase issues in this case, however, the Court, after conceding that errors occurred, employs an analysis that falls far short of supporting its conclusion that those errors were harmless. I believe that the errors were not harmless beyond a reasonable doubt and that, individually and cumulatively, they require that defendant’s death sentence be vacated.
A.
The submission of both the depravity and escape-detection aggravating factors (in addition to the felony-murder factor, whose existence defendant essentially conceded) to the jury was one of the most contentious issues in this capital prosecution. As the Court notes, ante at 381, 700 A.2d at 333, on numerous occasions, defendant moved to dismiss the depravity factor because the finding of lack of motive required to support it was inconsistent with the definite motive required to find the escape-detection factor. Defendant’s motions created confusion on the part of the trial court and induced the prosecutor to make a number of telling statements that demonstrated her knowledge of the impermissibility of the submission of both factors.
The trial court’s position was consistent throughout the proceedings, namely, that the jury could find motive (i.e., to escape detection) or lack thereof (i.e., depravity) but could not find both. The court based its belief that the two factors could be submitted to the jury, despite their mutual exclusivity, on a misreading of the precise wording of our holding in State v. Moore, 122 N.J. 420, 471, 585 A.2d 864 (1991), in which we noted the “logical contradiction” in submitting the two factors.
Although the trial court’s unfortunate error stemmed from a misreading of our precedents, the prosecutor’s “errors” were not so benign. She initially expressed agreement with the trial court that the two factors were mutually exclusive, but she subsequently changed her position, stating that “Pm not sure I haven’t proven both. There is no reason for the killing of this child, none. That’s *444the depravity. He did it because he wanted to. And he didn’t want to get caught. I think they are both here.” After trial, the prosecutor clarified her rationale for charging both aggravating factors:
Now, Samuel Moore says that those are mutually exclusive aggravating factors. But I would submit to you that this is the exception to the rule. I would submit to you that even though the Supreme Court gives us that direction, that in this ease they have to give us the opportunity, because this is the rare case where it was fun and part of the fun was to get away with it.
After the trial court, with the encouragement of the prosecutor, submitted both factors, in complete contravention of “the Supreme Court[’s] ... direction,” the court instructed the jury that it could not find both factors, because they were mutually exclusive. The court never instructed the jury that it could reject both factors. The prosecutor, after commenting on the mutual exclusivity, told the jury in summation:
He either did it because it involved depravity [of] the mind, ... which really talks about from society’s point of view, a totally meaningless, senseless, from the point of view of society, killed her. And in the alternative, ... to keep from getting caught____ And you will be able to decide and it’s not hard, I think, between the two. They are an alternative. One is basically an absence of motive and the other is motive.
She also stated: “The question remaining is which of the other two aggravating factors [i.e., other than the felony-murder factor] is here,” and “so you have to pick one [i.e., either escape-detection or depravity].” Again, neither the court nor the prosecutor qualified those statements by telling the jury that it had a third alternative, namely, rejection of both factors.
After its penalty-phase deliberations, the jury unanimously rejected the depravity factor and unanimously found the existence of the felony-murder and escape-detection factors. In weighing the aggravating factors against the mitigating evidence, the jury explicitly found that the two factors, when combined, outweighed the mitigating evidence but that, individually, they did not.4 Thus, *445only with both aggravating factors did the State obtain a death sentence.
The Court correctly holds that the trial court committed a blatant and serious error when it submitted both the depravity and escape-detection factors to the jury. Ante at 383, 700 A.2d at 334. We have held on numerous occasions that the lack-of-motive component of the depravity factor may not be submitted along with a factor that requires a finding of motive. Coyle, supra, 119 N.J. at 233-35, 574 A.2d 951; State v. Rose, 112 N.J. 454, 529-33, 548 A.2d 1058 (1988). Moreover, as the Court notes, we specifically have applied that principle to the depravity/escape-deteetion context. S. Moore, supra, 122 N.J. at 471, 585 A.2d 864. Trial courts that preside over capital cases must be knowledgeable about our capital jurisprudence, given that a defendant’s life hangs in the balance.
Despite the Court’s correct holding that the trial court committed error, it nevertheless classifies the error as harmless, glossing over the error’s factual context, omitting much of the prosecutor’s outrageous conduct, and ignoring several important cases that point toward a finding of prejudice.
In State v. Christener, 71 N.J. 55, 362 A.2d 1153 (1976), the trial court submitted both murder and manslaughter to the jury, even though insufficient evidence existed to support the murder charge. The jury acquitted the defendant of murder and convicted him of manslaughter. After recognizing that submission of the murder charge was error, we confronted the error’s potential harmless*446ness resulting from the jury’s rejection of the erroneously submitted offense. We held that the error was prejudicial because “the possibility that the jury, in the absence of sufficient evidence to sustain a first degree murder charge, may have reached a compromise verdict suggests that [defendant] may have suffered prejudice by that instruction in spite of his manslaughter conviction.” Id. at 69-70, 362 A.2d 1153.
In State v. Thomas, we clarified Christener by stating that
[o]ur holding in Christener does not stand for the proposition that reversal is mandated every time a judge charges a jury about a crime for which there may be insufficient evidence to support a conviction. The entire discussion of Christener centered on the prejudicial effect of “overcharging,” or instructing the jury on a crime more serious than is warranted by the evidence. In this case, the mistaken charge was not an overcharge. Rather, it related to alternate theories of establishing the same offense____
[76 N.J. 344, 365, 387 A.2d 1187 (1978).]
We concluded in Thomas that, to show prejudice, a defendant would have to show that the mistaken charge led to jury confusion. Ibid.
Since Christener and its clarification in Thomas, we have applied Christener analysis in the context of overcharging of aggravating factors. In State v. DiFrisco, 137 N.J. 434, 502, 645 A.2d 734 (1994) (DiFrisco II), cert. denied, — U.S. -, 116 S.Ct. 949, 133 L. Ed.2d 873 (1996), we held that the erroneous submission of the escape-detection factor was harmless because the jury had rejected the factor and had found that the other factor alone outweighed the mitigating evidence. We distinguished Christener by reasoning that the prejudice to the defendant in that case had grown out of the mutually exclusive nature of murder and manslaughter. Ibid. In other words, because the Christener jury had to consider the mutually exclusive murder and manslaughter charges in conjunction with one another, it was more likely to reach a compromise verdict of manslaughter where it might otherwise have acquitted of all charges. On the other hand, we reasoned that, when the jury is confronted with two non-mutually-exclusive aggravating factors, it must consider each one independently of the other. We concluded that it would be speculative to *447assume that the jury’s consideration of one independent aggravating factor affected (via compromise) its consideration of the other independent factor. Ibid.
In Rose, supra, we previously had noted the distinction between independent and mutually exclusive aggravating factors in the Christener context:
Unlike the mutually exclusive choices between first-degree murder and manslaughter ... the jury’s function in the penalty phase of a capital case is first to assess, independently of each other, the sufficiency of proof of the aggravating and mitigating factors. Its rejection of one such factor neither compels nor inhibits its determination that another factor exists.
[112 N.J. at 533, 548 A.2d 1058 (emphasis added).]
Thus, the essence of Christener prejudice is that when the jury is overcharged, be it with an unsupported greater offense or aggravating factor, the inquiry must focus on the potential for jury confusion and compromise, focusing on the interdependence of the properly and improperly charged offenses or factors. Mutual exclusivity, as evidenced in Rose, is an important consideration in evaluating the possibility of jury confusion and compromise because, where the jury is told that it cannot choose both options and that accepting one option automatically means rejection of the other option, the jury may feel constrained to choose one of the options, thus overlooking the choice of rejecting all of the options.
In the context of aggravating factors, an evaluation of Christen-er prejudice entails consideration of which factors were submitted and the potential for confusion. For example, if the felony-murder factor is properly submitted but the murder-for-hire factor is improperly submitted, the jury can be expected to consider each without reference to the other. The same cannot be said, however, of aggravating factors that are inextricably linked to one another, as both DiFrisco II and Rose recognized.
In this case, the depravity and escape-detection factors were mutually exclusive, thereby triggering Christener concerns. The jury had to consider the two factors jointly, and its deliberations on one necessarily affected its deliberations on the other; if it found that one existed, it had to reject the other. The jury easily *448could have been confused about the relationship between the two and may have believed that it had to choose one or may have compromised by choosing one.
That rather abstract vision of prejudice was translated into reality by the actions of the trial court and especially the prosecutor. Both stressed to the jury that it could not pick both factors, but they never explicitly informed it that it could reject both. The jury thus may have believed, consciously or subconsciously, that it needed to pick one of the two. The danger of jury compromise was quite real. The prosecutor’s inaccurate statements strengthened the subtle, but clear, message that the jury “ha[d] to pick one.” Supra at 372, 700 A.2d at 328. (“He either did it because it involved depravity [of] the mind, ... [a]nd in the alternative, ... to keep from getting caught — ”). The jury easily could have interpreted those remarks, which went uncorrected and unclarified by the trial court, as instructions to pick one or the other; after all, either defendant had a motive, or he did not have a motive.
Nor does the jury’s uncontroversial acceptance of the felony-murder factor obviate a finding of prejudice. As noted, the jury explicitly found that the two aggravating factors (felony-murder and escape-detection) outweighed the mitigating evidence only when combined and that, individually, they did not. Thus, had the jury found only the felony-murder factor, it would not have imposed death.
Defendant was further prejudiced by the erroneous submission because of his resulting inability to argue against one of the mutually exclusive factors without implicitly arguing in favor of the other. For example, to argue that depravity was not present necessarily implied that defendant had a motive, which the jury reasonably could have interpreted as a concession of the existence of the escape-detection factor. Conversely, an argument that defendant’s motive was not to escape detection, while failing to present another motive, could have been viewed as a concession of lack of motive, ie., depravity. The erroneous submission of the *449depravity factor thus effectively placed defendant in a catch-22 in his efforts to confront either of the two factors.
Defense counsel’s penalty-phase summation reflected that dilemma by urging the jury to reject both factors because it was unclear whether or not defendant had a motive. Jurors, however, reasonably could have discerned from that subtle argument the implication that, although it was unclear which of the two factors was present, one of the two had to be present because defendant killed either for a reason or for no reason — there could be no in-between. By precluding counsel from strenuously arguing against either of the factors, the erroneous submission placed him in the untenable position of implying to the jury that one of the two had to be present. The State should not be able to abuse its discretion in charging aggravating factors by placing defense counsel in such a precarious and untenable posture.
We thus are faced with a situation in which the trial court, over defendant’s strenuous and repeated objections, erroneously submitted an aggravating factor to the jury. That error, by itself, could support a finding of prejudice because of the mutually exclusive nature of the depravity and escape-detection factors. The error then was greatly exacerbated when the prosecutor inaccurately told the jury that it had to pick one or the other and when defense counsel was precluded from effectively advocating against either one. The jury’s explicit statement on the verdict sheet that it would not have sentenced defendant to death had it rejected both factors is proof of the prejudice that defendant may have suffered as a result. That course of events constitutes reversible error.
In classifying the error as harmless the Court engages in the most cursory analysis, reasoning that because the “rejection by the jury of one aggravating ... factor does not compel or inhibit its determination that another factor existed,” it would be “highly speculative” to conclude that defendant was prejudiced. Ante at 384, 700 A.2d at 334. The Court then undermines its own argument by stressing the trial court’s instruction to the jury that *450if it found one factor, it had to reject the other. Ante at 384, 700 A.2d at 335. What the Court appears to miss, however, is that the instruction to the jury that the two factors were mutually exclusive certainly “inhibit[ed] its determination that another factor existed,” ante at 384, 700 A.2d at 385, and, given the prosecutor’s repeated, inaccurate remarks, may have “compelDed] ... its determination that another factor existed,” ibid. One need not “speculat[e],” ibid., to draw that inference.
Nowhere does the Court confront the heart of this claim, namely, that the overcharging and its attendant circumstances created a reasonable possibility that the jury believed, either consciously or subconsciously, that, while it could not choose both factors, it “ha[d] to choose one,” given that defendant either had a motive or did not have a motive. Nor does it so much as pay lip service to the potential that the circumstances of this claim, from the prosecutor’s improper remarks to counsel’s paradoxical penalty-phase summation to the jury’s explicit findings,, had to prejudice defendant.
I do not assert that the circumstances here make it certain or even likely that the jury was induced by the cumulative errors to find the escape-detection factor. That, however, is not the standard. Instead, the measure is whether a reasonable possibility existed that, absent the multiple errors, one juror would not have found the factor. By that measure, one must conclude that the error generated a fatally prejudicial result.
B.
Eight hours into the penalty deliberations, the jury sent a note to the trial court, asking if jurors could use the photo exhibits of L.G. for the “sole purpose of determining whether the aggravating factors outweigh the mitigating factors.” Evident from the note was that the jury already had evaluated the existence of the aggravating and mitigating factors and that it wanted to view the photos during the subjective balancing portion of the penalty phase. The trial court, after giving a cautionary instruction, allowed the jury to use the photos during its balancing process. *451Fifty minutes later, after defense counsel claims to have heard “raised voices” coming from the jury room, the jury returned a death sentence. After the verdict, Exhibit S-158, or “the crotch picture,” as the prosecutor so eloquently termed it, was discovered in the jury room; that photo was so graphic that, prior to the penalty phase, the State had agreed to exclude it, as it had agreed to exclude all of the highly graphic photos.
I concur in the Court’s holding that the accidental submission was error because it “deprived defendant of an opportunity to confront that evidence,” ante at 394, 700 A.2d at 340, but I strongly disagree with its two-sentence analysis of the alleged harmlessness of the deprivation. The Court bases its finding of harmlessness on S-158’s admission during the guilt phase, defendant’s resulting opportunity to confront the photo during the guilt phase, the trial court’s cautionary instruction about the use of the photos, and the rejection of the depravity factor. Ibid. Yet, although those considerations may have reduced to a degree the prejudice that defendant suffered, they did not render the error harmless beyond a reasonable doubt.
It is important to consider the context of the accidental submission in order to assess its prejudicial impact. The use of the photos during the guilt phase was very different from their use during the penalty-phase balancing process. In the guilt phase, the jury viewed the photos solely to determine the factual issues relating to defendant’s guilt or innocence. Its guilt-phase viewing of the photos, moreover, had occurred substantially prior to its penalty deliberations. In the penalty-phase balancing process, by contrast, the jury was in the final stages of a highly sensitive and subjective determination of blameworthiness that clearly could be influenced and easily swayed by such a gruesome photo.
We know that the jury’s penalty determination was a close one. First, as I have noted, the jury determined that only the combined aggravating factors outweighed defendant’s mitigating evidence and that, individually, they did not. Supra at 371-372, 700 A.2d at 328-329. Second, we know from the post-trial hearing examining *452the propriety of Juror Maria Hollenback’s employment at the Monmouth County Prosecutors Office, that one of the juror’s had been reluctant to agree to a death sentence.
Given what we know about the jury’s use of the photo, we can readily envision a plausible situation in which defendant was prejudiced by its accidental submission. The jury already had deliberated for eight hours before viewing the photo, and, from the jury’s note, we know that it already had determined which aggravating and mitigating factors existed. When it began to use the photos, therefore, it was already in the critical, subjective balancing process. Given Hollenback’s statement that one juror had been reluctant about sentencing defendant to death, the eleven pro-death jurors may have desired to use graphic photos (including the inadvertently submitted S-158) to convince the holdout that defendant deserved to die. After the trial court responded affirmatively to the jury’s question about whether it could “review the photographs of the victim for the sole purpose of determining whether the aggravating factors outweigh the mitigating factors,” the pro-death jurors may have used S-158 to convince the holdout that he or she should switch. The jury’s determination that the individual aggravating factors did not outweigh the mitigating evidence may reflect that tentative verdict.
That set of possible circumstances is sufficiently plausible, considering the unusual amount of information that we have about the jury’s balancing process, to preclude a finding that the accidental submission was harmless beyond a reasonable doubt. Had the photo not been submitted, the jury would have been left with a number of autopsy photos that were not particularly graphic (as per the parties’ stipulation). Without the exhibit, therefore, the pro-death jurors may not have been able to persuade the potential holdout juror to switch, thus resulting in a life sentence. The exhibit was highly inflammatory; with its graphic display of L.G.’s spread legs and bloody genitalia, it may have been just enough to provide that late push toward a death sentence. That possibility *453is sufficient to cast a reasonable doubt on the harmlessness of the error.
The Court’s reliance on defendant’s opportunity to confront the photo does not rescue this death sentence. That counsel did not confront the photos during the guilt phase is irrelevant because of the difference between the guilt and penalty phases and because defendant did not contest anything except for intent during the guilt phase. Regarding the penalty phase, given the stipulation that no graphic photos would go to the jury, defense counsel may have made a deliberate choice not to confront them. Although it could be argued that counsel’s failure to confront them during the penalty phase is dispositive because of the trial court’s erroneous failure to limit the jury’s consideration of guilt-phase evidence during the penalty phase (i.e., counsel should have known that the jury could rely on its recollection of the photos regardless of their physical submission during the penalty phase), counsel may have made the reasonable determination that, given the time interval between the jury’s use of the photos during the guilt phase and the penalty deliberations, it would be unwise to remind jurors of the graphic photos. Had counsel been aware that S-158 would be submitted to the jury, however, he very well may have determined that it would be advisable to defuse its prejudicial impact during his penalty-phase summation.
In any event, we should not be so presumptuous as to guess what defense counsel’s strategy would have been had he known that the inflammatory exhibit would be in the jury room. The fact remains that he did not know and thereby was denied any opportunity to confront it in the context of the penalty phase.
We repeatedly have recognized the differences between the guilt and penalty phases in terms of the greater care that trial courts must take during the latter in ensuring that highly prejudicial evidence does not infect the jury’s sensitive penalty determination. State v. Dixon, 125 N.J. 223, 249, 593 A.2d 266 (1991); State v. Johnson, 120 N.J. 263, 299, 576 A.2d 834 (1990); State v. Pitts, 116 N.J. 580, 638-39, 562 A.2d 1320 (1989). In this case, we *454know that the gruesome and emotionally-charged photo was viewed during the most subjective and sensitive portion of the penalty phase, namely, the balancing process. Moreover, we know that the penalty determination was a close one. As one court has stated:
Modern day trials are factually presented in open court before the iron curtain descends upon the jury room. We cannot tolerate prejudicial factual intrusion into that sanctum lest our courts return to darker days of our jurisprudential history. The dagger of hidden evidence must not be taken from its scabbard for the first time in the jury room to wound the defendant; and unless its piercing effect is only skin deep and without prejudice to the anatomy of the trial, we must apply a constitutional salve.
[United States v. Howard, 506 F.2d 865, 866 (5th Cir.1975).]
This revolting and powerful piece of evidence was not “factually presented in open court,” instead having been “taken from its scabbard for the first time in the jury room.” Because it had the reasonable potential to affect the precarious balance in which defendant’s life hung, I would “apply a constitutional salve” and vacate defendant’s death sentence.
III
This case is a prime example of the severe constitutional infirmities that plague the administration of capital punishment in this State. Through the trial court’s errors, the prosecutor’s misconduct, and this Court’s refusal to view the trial for what it was — a proceeding profoundly lacking in fairness and due process — defendant has been condemned to die.
If we are to subject individuals to the ultimate violence, namely, the deliberate taking of life in the name of the people, then, at the very least, we should require that adequate safeguards be in place to ensure that that violence is not inflicted in an arbitrary manner. In this case, however, the Court dismisses and even legitimates substantial arbitrariness in both phases of the trial. Ranking with the Court’s failure to right the injustice of defendant’s death sentence is the grave harm that it inflicts on the progress that we have made over the last fifteen years in rooting out arbitrariness in the administration of capital punishment in New Jersey.
*455The arbitrariness of this death sentence becomes apparent when the errors infecting defendant’s conviction and sentence are viewed in their entirety. Defendant was found death-eligible by a jury that was unaware of the consequences of accepting his non-death-eligible theory of the case and that was encouraged to reach unanimity on the State’s death-eligible theory of the case before even thinking about defendant’s theory. Once found death-eligible, defendant’s penalty jury was told, both implicitly and explicitly, that it had to choose one of two alternative aggravating factors. After it potentially was induced into choosing more aggravating factors than it would have chosen with a proper presentation of aggravating factors, the jury’s balancing process, including a probable holdout juror, was infected by the unconfronted submission of a highly inflammatory photo.
At every stage of this capital prosecution — the determination of death-eligibility, the evaluation of aggravating factors, and the balancing of aggravating and mitigating factors — the questions were close. Yet, at every stage, something occurred to ensure that defendant would proceed to the next stage. However ugly the facts of this case are and however much the victim suffered at defendant’s hands, defendant was constitutionally entitled to a fair trial. That he did not receive.
I dissent.
POLLOCK, J., joins in Part I of this opinion.
O’HERN, J., joins in Parts I and IIA of this opinion.
STEIN, J., concurring in result.
For affirmance — Chief Justice PORITZ, and Justices GARIBALDI, STEIN and COLEMAN — 4.
For reversal — Justices HANDLER, POLLOCK and O’HERN — 3.
The Court must make its newly minted version of the ultimate-outcome instruction purely prospective because, even under the Court’s new, limited ultimate-outcome rule, the trial court in this case erred badly. As noted, after the trial court refused to instruct the jury about the exact sentence for felony murder, it refused to tell the jury anything about the sentence for felony murder, including the fact that it carried a substantial sentence that was equivalent to the noncapital sentence for purposeful-or-knowing murder. That refusal was wholly inconsistent with the Court's holding today that juries must receive at least a fair degree of information about the sentencing results of their guilt-phase determinations. Ante at 378-379, 700 A.2d at 331-332.
The facts of Purnell further undermine the Court's conclusion that felony murder, regardless of the specific facts of a case, is simply a traditional lesserineluded offense of purposeful-or-knowing murder and not an alternative offense. In Purnell, supra, the defendant stabbed the victim fifteen times in the neck, chest, and abdomen. 126 N.J. at 528, 601 A.2d 175. Without more, that factual scenario does not appear to provide a rational basis both for conviction of felony murder and acquittal of purposeful-or-knowing murder, which is exactly what would be required as a rational basis to charge a traditional lesserineluded offense. The Court, however, required that felony murder be presented precisely because of its non-death-eligible status.
"Moral equivalence" is the standard that both the United States Supreme Court and this Court have adopted for determining whether two mens reas are similar enough in terms of traditional notions of blameworthiness for the government to base a conviction for a single crime on the two mens reas without having to specify which one. See Schad v. Arizona, 501 U.S. 624, 643, 111 S.Ct. 2491, 2503, 115 L. Ed.2d 555, 573 (1991) (plurality opinion) ("If, then, two mental states are supposed to be equivalent means to satisfy the mens rea element of a single offense, they must reasonably reflect notions of equivalent blameworthiness or culpability, whereas a difference in their perceived degrees of culpability would be a reason to conclude that they identified different *435offenses altogether."); Bey III, supra, 129 N.J. at 581-82, 610 A.2d 814 (applying Schad standard).
The verdict sheet's explicit question about how many of the aggravating factors that the jury found were necessary to outweigh the mitigating factors is a *445practice that does not appear to be widespread in New Jersey, but it is one that trial courts should employ, whether or not the parties request it. Requiring juries to make that specification yields two benefits. First, it forces the jury to engage in deeper analysis of and reflection on the balancing process, particularly the precise weight to be assigned to each factor. Second, it allows for more effective and efficient appellate, review by this Court in that we can assess more accurately the prejudicial impact of errors relating to aggravating factors. For example, if a jury determines that each aggravating factor individually outweighs the mitigating evidence, then the erroneous submission of an aggravating factor is more likely to be harmless.