dissenting.
Gary Marsh, a night clerk at a gas station in Lawrenceville, was fatally shot at work during an apparent robbery. When found, he *372was laying among coins on the floor in the small office of the gas station, his keys were in the door, and the cash drawer was empty on the counter. Approximately ninety dollars was missing, but otherwise nothing in the small office was disturbed.
Defendant, twenty-six years old at the time, was arrested four days after the killing. Charged and prosecuted for capital murder, he was found guilty, on July 8,1994, of purposeful or knowing murder by his own conduct, with the requisite intent to kill the victim. On December 6, 1994, a separate jury returned a death sentence. The penalty-phase jury determined that the State had proven beyond a reasonable doubt three statutory aggravating factors: N.J.S.A. 2C:ll-3c(4)(a) (conviction for prior murder); N.J.S.A. 2C:ll-3c(4)(f) (murder committed to escape apprehension for another offense); and N.J.S.A. 2C:ll-3c(4)(g) (murder committed during the course of a robbery). All of the jurors concluded that the aggravating factors outweighed the mitigating factors, and defendant was sentenced to death. On direct appeal, defendant’s conviction and death sentence were affirmed. State v. Loftin, 146 N.J. 295, 680 A.2d 677 (1996) (.Loftin I).
The present appeal concerns the proportionality of defendant’s death sentence. The Court holds that defendant’s sentence is not disproportionate under the current standards and methodology for determining proportionality in capital eases. In reaching that determination, it eliminates one component of that methodology, namely, a statistical analysis based on the numerical preponderance of aggravating and mitigating factors. The Court rules further that proportionality review should be studied to determine whether its methodology should be revised. Anticipating such a revision, the Court also decides that a statutory amendment to the capital murder statute, which drastically changed the universe of eases that constitutes the basis for comparisons of defendants, need not be applied in this case; its constitutionality, therefore, need not be determined. Finally, the Court acknowledges that proportionality review and the evidence adduced in its implementation are material in determining whether racial discrimination *373exists in the administration and application of the death penalty. The signs of discrimination are undeniable, so the Court rules that racial discrimination should be the subject of continuing study. Nevertheless, the Court concludes that the evidential findings do not demonstrate discrimination sufficiently pervasive to establish constitutional violations.
I disagree with the Court’s disposition of each of the important issues in this case. First, because the Court has acknowledged that our current method of proportionality review requires further revisions, I take issue with the application of proportionality review to defendant’s case at this time. Given that the Court has chosen to review defendant’s sentence, I conclude that the evidence considered within the framework of current proportionality review standards indicates that defendant’s death sentence is disproportionate. Further, consideration of the constitutionality of the 1992 amendment should not be avoided or postponed on the assumption that a determination of its constitutionality may be obviated by any changes in proportionality review contemplated by this Court: the amendment, which changes and drastically limits the analysis required for meaningful proportionality review, is blatantly unconstitutional. Finally, and most importantly, the evidence presented by defendant demonstrates strongly and inescapably that a constitutionally impermissible risk of racial discrimination exists in the imposition of the death penalty in this State. We should declare the capital-murder statute unconstitutional because it discriminatorily singles out black persons for death.
I, therefore, dissent.
I
Rejecting the notion that “[ajpparent disparities in sentencing are an inevitable part of our criminal justice system,” McCleskey v. Kemp, 481 U.S. 279, 312, 107 S.Ct. 1756, 1778, 95 L.Ed.2d 262, 291 (1987), this Court has proclaimed that “our history and traditions would never countenance racial disparity in capital sentencing.” State v. Marshall, 130 N.J. 109, 207, 613 A.2d 1059 *374(1992) (Marshall II), cert. denied, 507 U.S. 929, 113 S.Ct. 1306, 122 L.Ed 2d 694 (1993) (emphasis added). The Court made this studied observation:
New Jersey would not tolerate a system that condones disparate treatment for black and white defendants or a system that would debase the value of a black victim’s life. Whether in the exercise of statutory proportionality review or our constitutional duty to assure the equal protection and due process of law, we cannot escape the responsibility to review any effects of race in capital sentencing,
[Id. at 214, 613 A.2d 1059 (citations omitted).]
In this case, the Court is confronted with extensive evidence indicating that racial discrimination infects this State’s capital punishment regime. The statistical record before us extends the evidence of racial discrimination from our earliest proportionality reviews, and takes into account subsequent capital cases, the current case, and several others pending. The evidence demonstrates that penalty-phase juries sentence a disproportionate number of black defendants to death, and that prosecutors capitally try a disproportionate number of cases involving white victims. The probability of being sentenced to death is thus positively correlated with whether the defendant is black and whether the victim is white.
The Court rules that the evidence is not sufficient to offend the constitution. See ante at 316, 724 A.2d at 160. I believe, to the contrary, that the statistics strongly indicate an impermissible risk of racial discrimination. A reliable determination of whether imposition of the death penalty is infected with racial discrimination need not, and indeed should not, be made solely on the basis of findings of statistical significance. The death-worthiness of a capital defendant as measured by culpability, the central and ultimate inquiry in a capital prosecution, is an inquiry that is inherently moral. Determinations of death-worthiness serve as the crucial indicators of whether or not invidious discrimination, has been a likely factor in the decision to impose the penalty of death. The criminal penalty in capital cases is the execution of the defendant, and the heart of the inquiry is whether persons most likely to be put to death are executed because they are black. *375Accordingly, the judgments that we reach as to whether individual defendants deserve to die cannot rest solely on pure scientific abstractions. Rather, the evaluation of the evidence that informs those judgments should be based on reasonable standards of knowledge that appropriately reflect real-life understanding. That knowledge must be extrapolated from an appraisal of the scientific material that not only draws on acceptable social scientific methodology, but also on common sense and human experience informed by the history that chronicles invidious discrimination in criminal prosecutions in this country.
While the statistics alone may not make the ease with scientific certainty, the totality of the evidence before this Court, measured by reasonable standards of knowledge appropriate for reaching sound judgments that assure ultimate fairness in capital-murder prosecutions, documents the risk of racial discrimination in the application of this State’s death penalty. So imposed, the death penalty violates the New Jersey Constitution and must be invalidated.
A.
This Court has consistently recognized that proportionality review is the tool by which to monitor whether capital-murder death penalty “proceedings are vulnerable to the influence of impermissible considerations.” State v. Ramseur, 106 N.J. 123, 326, 524 A.2d 188 (1987). “The prevention of ‘any impermissible discrimination in imposing the death penalty’ ” is one of the core purposes of proportionality review. Marshall II, supra, 130 N.J. at 135, 613 A.2d 1059 (quoting Ramseur, supra, 106 N.J. at 327, 524 A.2d 188). Uncovering racial discrimination is perhaps proportionality review’s most significant role. State v. Bey, 137 N.J. 334, 418, 645 A.2d 685 (1994) (Bey IV), cert. denied, 513 U.S. 1164, 115 S.Ct. 1131, 130 L.Ed.2d 1093 (1995).
Eleven years ago, this Court confronted empirical evidence indicating that New Jersey’s capital punishment system was racially discriminatory. The Court recognized the constitutional *376threat posed by that evidence and particularly acknowledged the relevance of the history of racial discrimination in capital punishment. Nevertheless, the Court refrained from finding invidious discrimination, promising instead that it would continue to accept evidence on this issue and “attempt to monitor the racial aspects of the application of the [Death Penalty] Act.” Ramseur, supra, 106 N.J. at 182, 524 A.2d 188.
Five years later, this Court, in its first proportionality review of a death sentence, was again faced with evidence indicating “that there may be a discrepancy in capital-sentencing rates that correlate to the race of the defendant or the race of the victim.” Marshall II, supra, 130 N.J. at 207, 613 A.2d 1059; see also David C. Baldus, State v. Robert Marshall: Death Penalty Proportionality Review Project, A Report to the New Jersey Supreme Court (Sept. 24,1991) (Baldus Report). In response, the Court rejected the United States Supreme Court’s suggestion that “ ‘[a]pparent disparities in sentencing are an inevitable part of our criminal justice system.’ ” Marshall II, supra, 130 N.J. at 207, 613 A.2d 1059 (quoting McCleskey, supra, 481 U.S. at 312, 107 S.Ct. at 1778, 95 L.Ed.2d at 291). “[W]ere we to believe that the race of the victim and race of the defendant played a significant part in capital-sentencing decisions in New Jersey,” the Court iterated, “we would seek corrective measures, and if that failed we could not, consistent with our State’s policy, tolerate discrimination that threatened the foundation of our system of law.” Id. at 209, 613 A.2d 1059.
In 1994, a second proportionality review case presented new statistics that, yet again, indicated that prosecutors and juries impermissibly considered the race of defendants and victims when imposing the death penalty. Bey IV, supra, 137 N.J. 334, 645 A.2d 685. The Court stressed that uncovering racial discrimination is perhaps proportionality review’s most significant role, id. at 418, 645 A.2d 685, and stated that it remained committed to eliminating any racial disparity either by seeking corrective measures or, if that failed, by taking more drastic steps, id. at 389, 645 *377A.2d 685 (citing Marshall II, supra, 130 N.J. at 209, 613 A.2d 1059). The Court portended that even though the disparities seen in McCleskey were not significant under the Federal Constitution, they “could be significant under the New Jersey Constitution.” Ibid, (citing Marshall II, supra, 130 N.J. at 210, 613 A.2d 1059).
Since then, the Court has considered two more proportionality review cases that included claims of racial discrimination, and, while acknowledging the real threat posed by discrimination, has nevertheless perceived the underlying statistical evidence as insufficient to demonstrate invidious racial disparities. State v. DiFrisco, 142 N.J. 148, 662 A.2d 442 (1995) (DiFrisco III); State v. Martini 139 N.J. 3, 651 A.2d 949 (1994) (Martini II), cert. denied, 516 U.S. 875, 116 S.Ct. 203, 133 L.Ed.2d 137 (1995).
In this case, the Administrative Office of the Courts (“AOC”) and defendant produced statistical data demonstrating for the first time that the race of the defendant and the race of the victim statistically significantly affect the likelihood of the imposition of a death sentence. In response to that evidence, the Court appointed a Special Master, Richard S. Cohen, who was given several tasks: (i) to examine the reliability of the AOC’s data regarding racial discrimination presented in the Harris Report and the Loftin Report; (ii) to utilize trial judges experienced in New Jersey’s death-penalty system to assess and rank by culpability all penalty-phase cases; and (iii) to undertake precedent-seeking review of those death-penalty cases falling within the mid-range of culpability (the range at which defendants had claimed and the AOC’s data had shown the most profound racial effect). The Special Master then issued findings and recommendations. See Richard S. Cohen, Report to the Supreme Court of New Jersey (Jan. 27,1997) {Special Master Report).
The studies undertaken by the AOC and defendant, together with those of the Special Master, provide evidence that racial disparities in the imposition of death sentences have not merely persisted, they have grown starker. All the data point toward the highest risk of the imposition of a death sentence when a black *378defendant kills a white victim, as was the case with Donald Loftin. This risk is sufficiently great that it can no longer be discounted.
1.
In the Loftin Report, and more recently in the Chew/Cooper/ Harvey Report, the AOC undertook analyses to measure the effects of race on the probability of the imposition of a death sentence. The reports reflect three logistic regressions that are designed to control for other pertinent variables impacting on deathworthiness, thereby producing results that can compare the sentences of defendants of different races who are equally culpable.
In the three models used by the AOC in its Loftin and Chew/Cooper/Harvey Reports (labeled Schedules 2, 5 and 8),1 the racial effect on capital sentencing is statistically significant with regard to several variables. A statistically significant black-defendant effect is reflected in Schedules 2 and 5 in both Reports and in Schedule 8 of the Loftin Report. This result strongly supports the conclusion that juries are more likely to sentence a black defendant to death than an equally culpable nonblack defendant. Although the black-defendant effect in Loftin Schedule 8, which uses the death-eligible universe, is not statistically significant (according to conventional standards), the observed significance is 0.0834, indicating that there is only an 8.34 percent chance that the black-defendant effect is caused by mere mathematical ran*379domness. Consequently, the black-defendant effect in Schedule 8 cannot be ignored.
A statistically significant white-victim effect is reflected in Schedules 2 and 8 of both the Loftin and Chew/Cooper/Harvey Reports. The effect’s statistical significance is strongest in Schedule 8, the only schedule that encompasses prosecutors’ charging decisions.
Table 1
AOC Statistical Significance Reports
Loftin Report
Variable Coefficient2 Observed Significance1
Schedule 2
White Victim 1.3555 0.0412
Black Defendant 1.4522 0.0171
Schedule 5
White Victim 0.5157 0.5535
Black Defendant 2.3796 0.0066
*380Schedule 8
White Victim 1.0699 0.0240
Black Defendant 0.9520 0.0418
Chew/Cooper/Harvey Report
Variable Coefficient Observed Significance
Schedule 2
White Victim 1.2922 0.0434
Black Defendant 1.1758 0.0376
Schedule 5
White Victim 0.9114 0.2490
Black Defendant 1.8423 0.0169
Schedule 8
White Victim 1.0517 0.0234
Black Defendant 0.7791 0.0834
The Special Master concludes that the raw data presented in these reports are a “mixed bag,” and that the AOC’s statistics are not dependable or accurate. Special Master Report, supra, at 43-45. In his view, “the Court is no closer than it was in the past to statistical evidence of a race effect,” and should continue to study the issue. Ibid. The Court accepts this evaluation of the evidence, and adopts the recommendation that the matter be subject to further study. See ante at 303, 315, 724 A.2d at 154, 160.
It is important to note that the Special Master does not conclude that the statistical evidence fails to indicate racial discrimination. Accordingly, while in his view the evidence does not prove the existence of racial bias, it also does not prove “that the system operates without bias.” Special Master Report, supra, at 34. Although the results showing a race effect “are not individually vivid,” the Special Master acknowledges that they do “seem to trend, to a degree not revealed by the statistical evidence, toward showing a race effect.” Ibid. In particular, he recognizes that the statistics regarding transracial cases reflect “disturbing trends *381... that may describe a troublesome racial imbalance, and clearly deserve intensified attention.” Id. at 44.
I believe the Court misjudges the strength of the evidence of invidious discrimination in capital prosecuting and sentencing decisions. The Loftin Report, updated with the data from the Chew/Cooper/Harvey Report, confirms earlier trends indicating that both the defendant’s arid the victim’s race have a statistically significant effect on whether a person who commits a capital murder is sentenced to death.
I begin with an examination of the raw numbers as analyzed by the Special Master. Referring to the Loftin Report, the Special Master points out that prosecutors selected almost equal numbers of black defendants and nonblack defendants for prosecution (seventy-three and seventy-four, respectively), although there were more death-eligible black defendants than nonblack (203 and 159, respectively). Special Master Report, supra, at 8. The Special Master then claims that reverse race discrimination occurs at the front end of the process, cancelling out the discrimination against black defendants at the back end of the system. He goes so far as to assert that “[t]he numbers could hardly have come out more race-neutral____” Id. at 9. The conclusion, or implication, that the numbers indicate that capital-murder prosecution is “race-neutral” is unsound and untenable.
The figures do not negate the presence of discrimination. The Special Master’s contention that discrimination of one sort in prosecutorial decision making and discrimination of the opposite sort in jury sentencing cancel out, creating a race-neutral outcome, is correct only as a matter of arithmetic; it could not be more amiss as a matter of law. The Special Master opines that “[i]f racial imbalance in penalty trial death verdicts occurred as an historical phenomenon, it followed, and might have been associated with, a reverse racial imbalance in capital prosecutions.” Id. at 11. Even if the most culpable defendants regardless of race were charged capitally and a higher percentage of those were nonblack *382defendants, that result could in no way “offset” any improper and discriminatory decision made by penalty-phase juries to sentence black defendants to death. Alternatively, if defendants’ race improperly influenced the decision to prosecute, discrimination against nonblack defendants in an effort to “offset” racial discrimination occurring in the penalty-phase is equally intolerable. Such a result can never be characterized as a “neutral outcome.”
Notwithstanding the expressed proposition that two different, potentially offsetting forms of race discrimination cause little need for concern, I agree with the Special Master’s conclusion that “[t]he limited scope and significance of these findings should be made clear.” Id. at 9. The Special Master and the Court are correct to point out that the findings show only that “nonblacks were 30% brought to penalty trial more frequently than blacks, and black penalty trial defendants received death sentences 27% more often than nonblack defendants.” Ante at 307, 724 A.2d at 156 (quoting Special Master Report, supra, at 9 (footnote omitted)). Although I find these discrepancies troubling on their own, the reason for the limitations of the findings is that they fail to take into account the race of the victims and the culpability of the defendants. When these two factors are considered, the statistics regarding both prosecutorial and sentencing decisions become even more troubling, demonstrating anything but a neutral or benign outcome.
Prosecutors selected over forty-eight percent of death-eligible black defendants who killed nonblaek victims (33/68) for capital prosecution, while none of the nonblack defendants who killed black victims (0/6) proceeded to a penalty trial. See tbl. 2, infra, at 383, 724 A.2d at 196. Prosecutors capitally tried forty-eight percent of nonblaek-vietim eases, but only twenty-eight percent of black-victim cases. Ibid. Overall, eighteen percent of black-defendant, nonblack-victim cases resulted in death sentences compared with zero percent of nonblack-defendant, black-victim eases. Ibid.
*383The disparity is even more troubling when the raw numbers are examined in light of the average culpability levels of each group provided by the AOC. Black killers of nonblack victims have an average culpability rating of 1.34, while nonblack killers of black victims have an average culpability rating of 1.67 — yet none of the defendants in the latter category were capitally prosecuted. Ibid.
Table 2
Special Master’s Findings 4
Racial Effect on Capital Prosecution and Sentencing by Culpability Average
Death-Eligible Cases Capitally Prosecuted Penalty Trial Cases Sentenced to Death Death-Eligible Cases Sentenced to Death Avg. Culpability 5
All cases 41% (147/362) 34% (50/147) 14% (50/362) 1.36
Black Defendants 36% (73/203) 38% (28/73) 14% (28/203) 1.32
Nonblack Defendants 47% (74/159) 30% (22/74) 14% (22/159) 1.43
Black Victims 28% (40/141) 40% (16/40) 11% (16/141) 1.32
Nonblack Victims 48% (107/221) 32% (34/107) 15% (34/221) 1.39
Black Defendants, Black Victims 30% (40/135) 40% (16/40) 12% (16/135) 1.30
Black Defendants, Nonblack Victims 49% (33/68) 36% (12/33) 18% (12/68) 1.34
Nonblack Defendants, Black Victims 0% (0/6) 0% (0/0) 0% (0/6) 1.67
Nonblack Defendants, Nonblack Victims 48% (74/153) 30% (22/74) 14% (22/153) 1.42
In addition, contrary to the Special Master’s conclusion, the eight percent disparity between black- and nonblack-defendant death sentences, see ibid., is worrisome. Although an eight percent discrepancy may seem facially mild, the figure becomes more significant when considered in context. The Special Master erroneously assumes that picking defendants for a death sentence *384is analogous to picking balls at random out of a sack stocked equally with two balls of different colors, and that the eight percent racial disparity in sentencing is no greater than one would get by randomly picking balls since it is not statistically significant. However, the analogy fails to account for culpability, the factor that should dictate a jury’s decision to sentence someone to death. One would not expect the fifty defendants selected for death necessarily to include twenty-five black defendants and twenty-five nonblack defendants, even if the pool of defendants were equally split between black and nonblack defendants, and even if no racial discrimination occurred. Rather, the expectation is that the fifty defendants picked would be the fifty most culpable defendants, regardless of their race.
Overall, nonblack defendants are more culpable than black defendants. See ibid. This finding is supported by the AOC culpability measurements broken down by level as well. Using the Special Master’s universe, the AOC analysis shows significant disparities with nonblack defendants tending to be more culpable in all five culpability levels. See Appendices (Apps.) A-l to -3, infra, at 446-447, 724 A.2d at 228. These disparities appear in greater significance at the higher culpability levels. In light of this fact, the eight percent disparity between black and nonblack defendant death sentences appears more significant and more ominous than the Special Master suggests.
The majority contends that “the Special Master understood the importance of the culpability rankings but found them to be unsound.” Ante at 307, 724 A.2d at 156. The recognition, however, that reliable culpability rankings do not exist, and can therefore not be used to test the hypothesis that the racial disparity is not statistically significant, does not render the Special Master’s ball analogy any more appropriate. The analogy fails unless we know something more about the balls than simply their color.
Further, even if one accepts the Special Master’s contention that the AOC’s culpability rankings are methodologically unrelia*385ble, Special Master Report, supra, at 19-20, the judges’ culpability ratings also indicate that nonblack defendants are, on average, more culpable than black defendants.6
The Special Master concludes that the statistical evidence is not adequate to demonstrate systemic racial discrimination. In doing so, he fails to give adequate recognition not only to the current statistics, but also to the constant nature of the disparity in death sentences between the races. Although the annual rate of death sentencing for black and nonblack defendants tends to fluctuate within any one year, the cumulative rates of death sentencing show a disparity between the races from 1984 on, and a rather constant, nonfluctuating disparity from 1986 to the present. In fact, over the last eight years, the disparity in death-sentencing rates has remained between six and eight percent.
In sum, the raw data reveal tangible and frightening disparities between black and nonblack defendants in the likelihood that prosecutors will seek, and in the rate that penalty-phase juries will impose, the death penalty. The disparities adversely affect black defendants, killers of white victims, and black defendants who kill white victims. Seen against the backdrop of prolonged racial discrimination that has historically and persistently plagued our nation, see infra, at 296-801, 724 A.2d at 150-153, we cannot trust that capital defendants are being selected to die by a system as benign as the one posited by the Special Master.
b.
Discrimination is also evident in the results of the multiregression analyses conducted by the AOC, which allow for examination of the race factor while controlling for culpability. The multiregression analysis reveals that race is a more important factor than most details of the crime, defendants’ backgrounds or records. *386See App. A-2, infra, at 447, 724 A.2d at 228.7 Differences in percentages of black and nonblack defendants sentenced to death are as high as thirty-three percent when the culpability levels are divided into five categories with equal ranges, see App. A-2, infra, at 447, 724 A.2d at 228, and thirty-six percent when the culpability levels are divided into levels with equal numbers of cases. Compare App. A-3, infra, at 447, 724 A.2d at 228 with App. A-1, infra, at 446, 724 A.2d at 228.
While these findings, as construed by the Special Master, are not statistically significant by themselves, the results comport with historical realities and the discrepancies continue over time with no sign of abatement. See infra at 296-302, 724 A.2d at 150-153. The Special Master recognizes that the AOC’s models are, despite shortcomings, both informative and useful. Special Master Report, supra, at 42-44. The AOC’s models indicate that a defendant’s race likely plays a very important role in penalty-phase juries’ decisions to impose death sentences instead of life sentences. In fact, defendants’ race is more important than almost every other variable in the models, including the presence or absence of almost every aggravating and mitigating factor.8 When viewed in light of the AOC’s findings that race plays a statistically significant role in capital sentencing, the Court’s decision to reject the contention that our system of capital punishment is in all likelihood discriminatory is unacceptable.
*387By way of explanation for his conclusion that there is not adequate evidence of racial discrimination in capital prosecution and sentencing, the Special Master criticizes the AOC’s culpability models, finding several methodological “defects that repel ultimate reliance.” Id. at 44. He believes the AOC’s statistical methodology is flawed because of the excess of factors compared to the number of death verdicts in the universe (a problem known as “overfitting”), and because of the “ambiguities” in some of the data selections and codings made by the AOC. The Special Master contends that the small size of the raw data pool renders a theoretically sound statistical methodology unsuited to the purpose for which it is being used. Id. at 20. Finally, the Special Master criticizes the process used by the AOC, known as “purging,” to correct the impermissible effects of certain variables on culpability determinations. Id. at 30. In my opinion, these criticisms, while not without merit, do not justify rejection of the AOC’s statistical evidence.
The Special Master’s assertion that the small size of the universe renders the AOC results unreliable is an important one. Id. at 20. The database contains only 147 penalty-trial cases and only fifty death-sentenced cases. On the other hand, an almost limitless number of variables may account for why different juries decided to impose death sentences in those fifty cases. The difficulty in creating a model to account for the historical occurrences is that the model may become so detailed as to have no predictive ability. Id. at 27-28. Although this is indeed a valid concern, the paucity of cases cannot, under the circumstances, become dispositive in the Court’s decision to ignore tangible evidence of systemic racial discrimination. The black man executed purely because of his race finds no solace and gains no redemption from the argument that had his case been examined ten years later — confirming what we reasonably now know — when the pool of comparison eases was larger, he would have been spared. This Court cannot, on the ground that the population size is too small, continue to ignore the risk that racial discrimination is infecting this State’s imposition of the death penalty.
*388Given his concerns regarding the validity of the AOC’s models, the Special Master appointed Dr. John Tukey as a statistical consultant to examine the AOC’s methodology and to develop some alternative models. Id. at 2. Dr. Tukey created three “parsimonious” models of his own in an attempt to limit the number of variables used by the AOC from thirty-two (used in Schedule 5) to a more acceptable number (between five and ten). Two of these models do not show a statistically significant race effect. For that reason, Dr. Tukey concludes that defendant has not relentlessly documented racial discrimination in this State’s imposition of the death penalty. John W. Tukey, Report to the Special Master 10 (Jan. 27,1997) (Tukey Report).
However, Dr. Tukey’s third model, which appears to be the most methodologically sound,9 reveals a statistically significant effect of the black-defendant variable on the likelihood that a defendant will be sentenced to death. The Special Master not only discounts Dr. Tukey’s results indicating a possible racial effect in two of three parsimonious models, but then also fails to give any recognition to the statistically significant racial effect in the third. His criticism, therefore, that the AOC models fail because of overfitting, seems purely technical: when provided with models that solve the overfitting dilemma, the Special Master still refuses to recognize compelling evidence of race discrimination.
*389Dr. Paul Allison, hired by the Public Defender’s Office, reconfigured Dr. Tukey’s parsimonious models slightly, maintaining their parsimonious quality, but adding the white-victim variable to the analysis. Appendix to Defendant’s Supplemental Brief in Answer to State’s Response to Report of the Special Master 8, 10 (Mar. 11, 1997) (Appendix to Defendant’s Supplemental Brief). The results were quite significant. Not only did Dr. Allison find a statistically significant racial effect on death penalty sentencing (correlating to the race of the defendant) in Model 3, as Dr. Tukey did, but the data also showed a statistically significant black-defendant effect in Dr. Tukey’s Model 2: that is, the percentage of black-defendant penalty-trial cases resulting in a death sentence was statistically significantly greater than that for nonblack-defendant eases. Ibid.10
The Special Master criticizes the models based on what he considers to be coding problems within the universe. 11 As a result of these coding “inconsistencies,” the Special Master revises his initial contention that the consistent trends in the data indicating the presence of race discrimination are worrisome. However, this revision, adopted by the majority, ante at 304, 724 A.2d at 154, is largely unexplained.
*390The Special Master himself states that while he finds some of the AOC coding decisions to be inconsistent, he does “not mean to equate ‘inconsistent’ with erroneous, or ‘consistent’ with preferable. These are alternate ways of sorting and neither one is ‘true’ or ‘false.’ ” Richard S. Cohen & Dr. John W. Tukey, A Study of Statistical Evidence of Race Bias in Penalty Trials (May 4, 1997) (Supplemental Report).
In the Supplemental Report, the Special Master recodes the data to make the racial and ethnic groupings more consistent. The results of these recodings do not undermine the conclusion that the indications of race discrimination are very real given the consistency of the results across models. On the contrary, if anything, the Supplemental Report supports the contention that the risk of race discrimination is profound and has infected our system of capital punishment. When the Special Master recodes the data to account for apparent inconsistencies in the AOC groupings, more, rather than less, significant disparities in sentencing between the races are revealed.12
The majority points out that these stark differences in the findings resulting from the recoding of the raw data are responsible for the Special Master’s reluctance to reiterate his concerns about the consistency of trends across models. See ante at 304-306, 724 A.2d at 154-155 (citing Supplemental Report, supra, at 10) (noting that “the modification of the Supplemental Report highlights the dangers inherent in the improper use of statistics” because trends are irrelevant when “separate analyses operate on the same data”). The further implication of that observation is that it may never be possible to create a statistical model that is satisfactory. The Special Master never identifies the coding deci*391sions as flawed. However, even if' he had, his recoding would presumably have been constructed to correct these flaws or, as he calls them, inconsistencies. Therefore, even if the AOC trend data are all based on the same flaw, these trends are not refuted by the “corrected” analysis — they are, rather, enhanced. The majority’s contention that the trends may not be trusted would rest on solid ground only if the Special Master’s recodings were considered methodologically superior to the original ones and refuted the trends. Neither is the case here.
Coding decisions must be made in every model — if the mere fact that the results will be different if the coding decisions are reworked is ample evidence that we cannot give credence to our results (even when the various coding schemes are equally legitimate), no model will ever meet the Court’s approval. The majority’s standard for reliability, therefore, is untenable. I find it compelling that the Public Defender’s, Dr. Tukey’s and, indeed, the Special Master’s recoded models, all fail to challenge the notion that racial discrimination is playing a role in this State’s imposition of the death penalty.
Finally, both Dr. Tukey and the Special Master raise questions about the propriety of the purging process which was used here to attempt to correct the effects of impermissible factors such as race on determinations of culpability. Special Master Report, supra, at 30. The purging process is a sound one. As initially recommended by Professor Baldus, the models from which the AOC computed culpability scores initially contained variables for the races of the defendant and the victim. After computing the black-defendant and white-victim effects, those variables were dropped from the equation. If the AOC had not purged the race-based variables, the culpability scores might have been tainted by racial discrimination. If the models never took race into account, the effects of racial discrimination, which are statistically significant in those models controlling for culpability, would be inaccurately attributed to other variables. Thus, the purging process is both *392sensible and necessary to compute accurate culpability scores that are not tainted by racial discrimination.
Even, however, if one rejects the validity of the purging process, the AOC’s models that show race to have a statistically significant effect were not purged; only the models that were used to provide culpability estimates were purged. Further, de-purging the models, while lowering the racial discrepancies evident in the culpability models (because, in effect, one is leaving in race as an impermissible aggravator), still indicates a substantial race effect. See id., tbl. D (revised) (showing non-purged as well as purged results). Thus, the purging process does not explain why the AOC’s culpability measurements show a marked race effect with a disparate impact on black defendants.
c.
In finding insufficient evidence of racial bias, the Special Master also examines the results derived from culpability measurements undertaken by judges. Special Master Report, supra, at 45. The judges’ culpability ratings were based on summaries of the cases prepared by the AOC, from which the race of the victim and the defendant were omitted. The Special Master concludes that the “judges’ raceblind culpability ratings were completely race-neutral,” id. at 23-24, and that the judges’ study shows no racial effect, id. at 43-45; see also App. B-1, infra at 447, 724 A.2d at 228.
The Special Master finds that the judges’ culpability measurements differ from the AOC’s statistically-generated culpability scores of death-eligible defendants’ culpability, and he therefore concludes that the AOC’s statistical analyses are inaccurate. “Unless one believes that the ratings produced by the panel of 50 judges were distorted by faulty methodology or are unworthy of weight for some other reason,” he concludes, “one has to view the judges’ ratings as challenging the accuracy of the computer outputs, and as undermining the significance of their purported showing that nonblack defendant[s] were on average more culpable....” Id. at 25.
*393The Court gives significant weight to this part of the Special Master’s study to prop up the position that the current evidence of discrimination is not persuasive. See ante at 309-310, 724 A.2d at 157-158. I believe that the judges’ culpability ratings neither detract materially from the AOC’s data nor justify disregarding the AOC’s statistical findings. In my opinion, the judges’ culpability ratings support defendant’s racial discrimination claim, albeit less strongly than the AOC’s data. Moreover, the methodology used in the Judges’ Survey is not flawless.
The existence of disparities between the culpability of black and nonblack defendants is essential to the claim that the death penalty is disproportionately imposed on black individuals. The AOC’s culpability measurements support the position that, overall, black defendants are less culpable than other defendants and that for defendants of equal culpability, regardless of the level of culpability, a black defendant is more likely to be sentenced to death than a nonblack defendant. For all defendants in the penalty-trial universe, the average culpability was 3.30 for a black defendant and 3.38 for a nonblack defendant. For defendants sentenced to death, the average culpability was 3.68 for a black defendant and 3.77 for a nonblack defendant. Special Master Report, supra, at 23.
In addition, the Special Master’s analysis of the judges’ culpability ratings may not reveal the true level of disparity in culpability between the races. The analysis is problematic because it focuses only on average ratings.13 When broken down by culpability level, *394the judges’ results are disturbing. See Apps. B-l to -4, infra, at 447-448, 724 A.2d at 228-229. Specifically, they demonstrate marked disparities between black and nonblack defendants at the higher culpability levels, with the greatest disparity evident in level five (which is defendant’s culpability level). Although slight reverse disparities appear at the lower culpability levels, the disparities are small and are due at most to a single additional death sentence at those levels. See App. B-4, infra, at 448, 724 *395A.2d at 229. In any case, the AOC culpability levels suitably reflect juries’ determinations of death-worthiness14 and the findings should therefore be given appropriate weight.
d.
At the very least, all of the findings indicate that the statistical disparity (thirty-eight percent of blacks (n=28) compared with thirty percent of nonblacks (n=22) sentenced to die) is utterly unexplainable when viewed in terms of relative culpabilities. The Special Master’s conclusion that a raw number of disparities between black and nonblack defendants sentenced to death can be explained by randomness fails to take into account the culpability levels of defendants in each racial category. Although the differences in culpability are not statistically significant, they buttress, rather than contradict, the AOC’s findings of race discrimination.
Unquestionably, the models created by the AOC, Dr. Tukey, and the Public Defender are all flawed in some very important ways, as are the judges’ culpability rankings. Nevertheless, they have residual probative worth and are useful indicators of the types of variables that have an impact on decision-making regarding the death penalty at various stages in the process. The fact that every single one of them indicates that race impacts capital sentencing cannot be shrugged off as a meaningless coincidence.
The significance of the racial effect becomes even clearer and more compelling when it is compared to the findings in McCleskey, *396which, as this Court earlier sensed, could serve to demonstrate racial bias in the imposition of the death penalty under the State Constitution. See Marshall II, supra, 130 N.J. at 210, 613 A.2d 1059.
In McCleskey, the statistical evidence strongly bore out the impact of the race of victims (but not the race of defendants), as a significant factor in determining who had received death sentences. The race of the victim was found to be a more important factor than most of the details of the offense or the defendant’s past criminal conduct. McCleskey, supra, 481 U.S. at 321, 107 S.Ct. at 1782, 95 L.Ed.2d at 297 (Brennan, J., dissenting). At the front end of the system, the data revealed that prosecutors sought the death penalty in seventy percent of eases in which a black killed a white, but in only nineteen percent of cases in which a white killed a black. On the other end, only one percent of defendants who killed black victims received a death sentence, while eleven percent of defendants who killed white victims were sentenced to death. Id. at 286, 107 S.Ct. at 1764, 95 L.Ed.2d at 274-75; id. at 326-27, 107 S.Ct. at 1785, 95 L.Ed.2d at 300-01 (Brennan, J., dissenting).
An examination of the defendant’s race alone revealed a higher rate of sentencing for white defendants (seven percent) when compared to that of black defendants (four percent). Id. at 286, 107 S.Ct. at 1764, 95 L.Ed.2d at 275. Overall, twenty-two percent of black killers of white victims received death sentences while only three percent of white killers of blacks were sentenced to death. Id. at 286-87,107 S.Ct. at 1764, 95 L.Ed.2d at 275. In the mid-range of culpability, where the largest disparities were seen, in one model fourteen percent of defendants in black victim cases were sentenced to death, while thirty-four percent of white victim cases resulted in death sentences. Id. at 287 n. 5, 107 S.Ct. at 1764 n. 5, 95 L.Ed.2d at 275 n. 5; id. at 325, 107 S.Ct. at 1785, 95 L.Ed.2d at 300 (Brennan, J., dissenting). These data revealed that killers of white victims were 4.3 times as likely to receive the death penalty as equally culpable killers of blacks. Id. at 287, 107 *397S.Ct. at 1764, 95 L.Ed.2d at 275; id. at 321, 107 S.Ct. at 1782, 95 L.Ed.2d at 297 (Brennan, J., dissenting).
The data presented in Loftin reveal racial effects driven by the race of the victim and the defendant similar in degree to disparities in McCleskey. The data show a forty-nine percent disparity between the percentage of death-eligible black defendants capitally prosecuted for killing nonblack victims and that of nonblack defendants prosecuted for killing black victims. See Tbl. 2, supra, at 383, 724 A.2d at 196 (comparing forty-eight percent to zero percent). In McCleskey, the disparity was fifty-one percent. McCleskey, supra, 481 U.S. at 286, 107 S.Ct. at 1764, 95 L.Ed.2d at 274-75; id. at 326-27, 107 S.Ct. at 1785, 95 L.Ed.2d at 300-01 (Brennan, J., dissenting) (comparing seventy percent to nineteen percent). Overall, the disparity between the percentage of black defendant, nonblack victim cases sentenced to death and that of nonblack defendant, black victim cases was eighteen percent, see tbl. 2, supra, at 383, 724 A.2d at 196 (comparing eighteen percent to zero percent); in McCleskey, the corresponding disparity was nineteen percent, id. at 286-87, 107 S.Ct. at 1764, 95 L.Ed.2d at 275 (comparing twenty-two percent to three percent).
The comparison is justified further when one examines the results of the multiregression analyses to those in McCleskey. Differences in percentages of black and nonblack defendants sentenced to death are as high as thirty-three percent when the culpability levels are divided equally, see App. A-2, infra, at 447, 724 A.2d at 228, and thirty-six percent when the culpability levels include equal numbers of eases, see App. A-3, infra, at 447, 724 A.2d at 228; App. A-l, infra, at 446, 724 A.2d at 228. In McCleskey, the greatest disparity was twenty percent when defendants were examined by race and with probability level. Id. at 287 n. 5, 107 S.Ct. at 1764 n.5, 95 L.Ed.2d at 275 n. 5; id. at 325, 107 S.Ct. at 1785, 95 L.Ed.2d at 300 (Brennan, J., dissenting).
*398In fact, in some instances, the AOC data reveal a significant defendant-race effect on the imposition of the death penalty that exceeds that found in McCleskey. In McCleskey, the most compelling statistic was that killers of white victims were 4.3 times more likely to be sentenced to death than equally culpable killers of black victims. Id. at 287, 107 S.Ct. at 1764, 95 L.Ed.2d at 275; id. at 321, 107 S.Ct. at 1782, 95 L.Ed.2d at 297 (Brennan, J., dissenting). Here there is evidence that black defendants are 6.1 times more likely to be sentenced to death than equally culpable nonblack defendants (comparing forty-three percent to seven percent). See App. A-3, infra, at 447, 724 A.2d at 228.
Given this Court’s prior avowal that racial disparities in capital sentencing comparable to those demonstrated in McCleskey will not be tolerated, its present unwillingness to acknowledge the strength of the evidence before it is deeply distressing. The Court blinks in the face of the truth: discrimination based on race maintains a persistent presence in the administration of the death penalty.
B.
The predicament this case presents is profound: To discover the existence and extent of the risk of invidious discrimination in the administration of capital punishment, the Court’s need for facts surpasses the capacity of the scientific community to furnish them. See Alan B. Handler, The Judicial Pursuit of Knowledge: Truth and/or Justice, 41 Rutgers L.Rev. 1, 26 (1988). To acknowledge that dilemma is not a concession that reliable facts are not available and cannot be gathered. Nor does it require that we rely on factual determinations steeped in ignorance or speculation. The singular danger posed by this daunting intellectual challenge is acceptance of the premise that only punctilious and pristine statistical analyses can be the basis for factfinding.
The path to decisions soundly based on reliable and usable facts must necessarily be determined by our destination, by what we *399are seeking to learn and trying to accomplish. Here, the fundamental matter to be learned is whether a risk of pernicious systemic discrimination inheres in the administration of the death penalty; we are striving to avoid the execution of persons in the presence of that risk. The determination that discrimination exists, and that persons may be executed notwithstanding that risk, is legal, not scientific. The insistence on scientific perfection and statistical nicety to a degree that obscures the obvious and forces the rejection of facts that are at hand misperceives the nature of the judicial inquiry and, because life itself hangs in the balance, fundamentally disserves the cause of justice.
Nowhere in the law is more at stake or is there a greater need for positing a definitive resolution on a sound and understandable basis than in a capital case, even if that resolution errs by falling on the side of fairness rather than accuracy. See, e.g, Woodson v. North Carolina, 428 U.S. 280, 303-04, 96 S.Ct. 2978, 2991, 49 L.Ed.2d 944, 961 (1976) (“[Djeath is a punishment different from all other sanctions.”); Furman v. Georgia, 408 U.S. 238, 286, 92 S.Ct. 2726, 2750, 33 L.Ed.2d 346, 376 (1972) (Brennan, J., concurring) (“Death is a unique punishment in the United States.”); Loftin, supra, 146 N.J. at 368, 680 A.2d 677 (recognizing “unique nature of the death penalty”). “Because of the range of discretion entrusted to a jury in a capital sentencing hearing, there is a unique opportunity for racial prejudice to operate but remain undetected.” Turner v. Murray, 476 U.S. 28, 35, 106 S.Ct. 1683, 1687, 90 L.Ed.2d 27, 36 (1986). “The heightened concern in a capital ease for whether a sentence is disproportionate ... derives from the finality of the result and the risk that the proceedings are vulnerable to the influence of impermissible considerations.” Ramseur, supra, 106 N.J. at 326, 524 A.2d 188. “[T]he fact that we must always act without the illumination of complete knowledge cannot induce paralysis when we confront what is literally an issue of life and death.” McCleskey, supra, 481 U.S. at 335, 107 S.Ct. at 1789, 95 L.Ed.2d at 306 (Brennan, J., dissenting).
*400The intellectual difficulties in scientifically proving the existence of racism and the extent of the risk of discrimination should not defeat our search for the truth. Those difficulties cannot obscure wrongs that appear undeniable, and that, as a matter of law, policy and justice, demand acknowledgement. In Rubanick v. Witco Chemical Corp., 125 N.J. 421, 593 A.2d 733 (1991), for example, this Court recognized “the extraordinary and unique burdens facing plaintiffs who seek to prove causation in toxic tort litigation,” id. at 433, 593 A.2d 733, and concluded that in certain settings, “[t]he scientific method ... fails to address or accommodate the needs and goals” to be achieved by a judicial determination, id. at 436-37, 593 A.2d 733. Because “plaintiffs in toxic-tort litigation, despite strong and indeed compelling indicators that they have been tortiously harmed by toxic exposure, may never recover if required to await general acceptance by the scientific community of a reasonable, but as of yet uncertain, theory of causation,” id. at 434, 593 A.2d 733, we held that strict scientific standards may be loosened in this arena, id. at 449, 593 A.2d 733.
As with environmental pollutant cases, there are strong statistical indications here that an impermissible factor — racism—is inextricably linked to the imposition of the death penalty in our criminal justice system. Due to the finality of the death sentence and the concern that capital punishment is infected with racial discrimination, this Court should not hesitate to find support from the AOC’s data and logistic regressions, despite the imperfections. Our toxic-tort holdings have always insisted on an assessment of the soundness of the methodology employed to demonstrate causation. See ibid, (holding that sound methodology can demonstrate facts even though result may not be generally-accepted by scientific community); see also Landrigan v. Celotex Corp., 127 N.J. 404, 420, 605 A.2d 1079 (1992) (hMdmg ln~toxic-tort context that scientifically accepted methodology that included a review of the scientific literature by the expert to “establish both the ability of [the toxin] to cause [the harm] and the magnitude of the risk that it would cause that result” provides adequate basis for expert’s *401“conclu[sionj that decedent’s exposure more likely than not had been the cause of his colon cancer.”).
The majority takes issue with the conclusion that the methodology employed in the statistical models at our disposal is adequate to meet the standard of reliability established in our toxic-tort cases. See ante at 313-314, 724 A.2d at 159 (“[H]ere the statistical models we use for proportionality review do not even meet the more lenient standard accepted by the Court in toxic-tort eases.”). The Court, however, takes the tort analogy too literally. Our acceptance of a level of proof in toxic-tort eases that is less than a scientific certainty is relevant to show that, in contexts in which actual scientific proof is difficult to harness with any reliable degree of certainty, we may apply a flexible standard when determining causation. The extent to which that standard is adjusted, however, must necessarily depend on the specific circumstances of the proof at issue. Unlike in the toxic-tort cases, the statistical models here are not being used to demonstrate that defendant was the actual victim of race discrimination; nor are they being used to pinpoint one or more specific actors who caused the harm. Rather, the models are used to determine the presence of general trends, indicating an unconstitutional risk that the system, in its entirety, operates in a racially discriminatory manner. Accordingly, whether or not the proof of racial discrimination presented by defendant meets the standard used in toxic-tort cases is not the point.
We must ask ourselves what level of statistical certainty is required before we decide the risk of discrimination is too great. The Court has followed the widely accepted rule of ninety-five percent certainty — that is, until our models demonstrate only a five percent possibility that the result achieved is due to chance rather than to the race of the defendant and/or the victim, the Court will not halt executions on the basis of race discrimination. In effect, the Court has turned the convenient measure of “statistical significance” into a talismanic rule. Despite its widespread *402acceptance in social science research, this cut-off point is, for the most part, arbitrary:
‘Statistical significance’ is a good example of something that looks to many people like simple fact. Statistical significance — the cut-off point scientists use to say ‘more than this much evidence counts as proof, and less than that much doesn’t’— when you get right down to it is arbitrary. Scientists use that cut-off line because it has worked pretty well. But that does not mean someone ... cannot or should not question a particular conclusion that something is or is not proved.
[Alice Dreger, A Call for Passion in the Realm of Discovery, N.Y. Times, Jan. 12,1999, at F4J
Accord Robert J. Levine, Ethics and Regulation of Clinical Research 933 (2d Ed.1986) (‘We have chosen arbitrarily to say that something is true when probability is less than 0.05 that it could have occurred by chance ... ”); Michael D. Maltz, Deviating from the Mean: The Declining Significance of Significance, 31 J. Res. Crime & Deling. 434,440 (1994) (“Statistical significance does not imply substantive significance, and most researchers know this — but this does not stop them from implying that it does”); Joseph Sanders, From Science to Evidence: The Testimony on Causation in the Benediction Cases, 46 Stan. L.Rev. 1, 15 (1993) (“There is nothing magic about an Alpha of .05 ... An Alpha of .10 is not inappropriate, nor is an Alpha of .01”).
Legal scholars, judges and practitioners have been quick to accept the stringent standards set by scientists, perhaps with good reason, in order to avoid claiming causal relationships that do not exist, and thereby avoid holding innocent parties responsible for harms they did not commit. Other disciplines, ones that have traditionally relied more on scientific certainty than law does, question the value of arbitrary line-drawing and the use of science to the exclusion of common-sense observations in circumstances less troubling than ones of life and death. E.g., Patrick E. Shrout, Should Significance Tests Be Banned? Introduction to a Special Section Exploring the Pros and Cons, 8 Psychological Science 1 (1997) (noting the high rate of error in making predictions about the efficacy of a treatment based on “significance testing”); Louis Uchitelle, A Challenge to Scientific Economics, An Older School Looks At Broad, More Intuitive Picture While Modernists See *403Just the Numbers and Facts, N.Y. Times, Jan. 23, 1999, at B7 (quoting Alfred Marshall, the 19th Century British economist: “ ‘[E]conomics cannot be compared with the exact physical sciences, for it deals with the ever changing subtle forces of human nature.’ ”). In certain instances, we should be prepared to question whether the level of certainty demanded by the scientists should not be modified.15 “The law may be less interested [than science] in avoiding [false positives ] and may thus be more willing to accept a less stringent standard for statistical significance.” Sanders, swpra, 46 Stan. L.Rev. at 15. Would it be so radical to decide that a ninety percent certainty that race is playing an impermissible role in capital sentencing violates the Constitution?
Even if we were to alter the level of significance at which we acknowledge an unacceptable risk of race discrimination in capital sentencing, methodology must, of course, remain important; but the number of studies we now have that employ different methodologies — all of which indicate the same, impermissible trends— should be examined in light of what we know about our nation’s history of race discrimination. When a defendant’s life is at stake, I believe we should be more certain that race is not playing a role in capital sentencing decisions than the generally accepted level of statistical significance allows.
*404As the Court noted in Landrigan, supra, in the toxic-tort context, once expert testimony has been admitted regarding causation,
two criteria, among others, ... are crucial in determining whether a statistical association will give rise to an inference that a particular substance causes a certain disease in people who are exposed to it. The two criteria are the strength of the association and the consistency of any such association with other knowledge.
[127 N.J. at 416, 605 A.2d 1079.]
The studies with which we are confronted show a strong association between race and the imposition of a death sentence. The consistency of this association with our general knowledge is equally riveting. The relevant literature reveals a long and relentless history of racism that has not only the capacity to cause a disproportionate impact on blacks in the administration of the death penalty, but has indeed done so from the era of slavery in this country, and, many argue, to the present. If we look at the AOC studies in light of a presumption that racism exists in our prosecutorial and jury decision making process in death penalty cases, the statistics lead us to the obvious conclusion; when controlling for other risk factors such as culpability, race is playing a constitutionally impermissible role in our State’s administration of the death penalty.
“Evaluation of [defendant’s] evidence cannot rest solely on the numbers themselves. We must also ask whether the conclusion suggested by those numbers is consonant with our understanding of history and human experience.” McCleskey, supra, 481 U.S. at 328, 107 S.Ct. at 1786, 95 L.Ed.2d at 302 (Brennan, J., dissenting); see Theodore M. Porter, Trust In Numbers: The Pursuit of Objectivity in Science and Public Life 7 (1995) (“Even with regard to purely scientific matters, the importance of tacit knowledge is widely recognized. In efforts to solve problems posed from outside the scientific community, informed intuition is all the more crucial.”). Even the majority acknowledges a need to look beyond the numbers, noting that “[w]e have never believed that statistics can replace the process of judging; we do believe that statistical results, properly used and understood, can alert us to the need for *405increased vigilance in our quest for impartial justice.” Ante at 323, 724 A.2d at 164. That monition, expressed as a consideration in examining the methodology for individual proportionality review, applies even more cogently to the determination of the extent of the risk of systemic racial discrimination in capital punishment. In this ease, the Court should recognize that the history and legacy of racism confirm the statistical evidence that racial discrimination influences the imposition of capital punishment. Accord McCleskey, supra, 481 U.S. at 328-30, 107 S.Ct. at 1786-87, 95 L.Ed.2d at 302 (Brennan, J., dissenting) (“Georgia’s legacy of a race-conscious criminal justice system, as well as this Court’s own recognition of the persistent danger that racial attitudes may affect criminal proceedings, indicates that McCleskey’s claim is not a fanciful product of mere statistical artifice.”) (footnote omitted).
The statistics presented by this case indicate discrimination that comports with this Nation’s and this State’s history of racial discrimination in criminal law and capital punishment. “[Ojnly those oblivious to the brutal history of racial discrimination in American law would deny the danger of racial prejudice entering the decisions which lead to the imposition of a death sentence.” Stephen B. Bright, Discrimination, Death and Denial: The Tolerance of Racial Discrimination in the Infliction of the Death Penalty, 35 Santa Clara L.Rev. 433, 438 (1995) (footnote omitted). It is crucial in this setting to recount this tragic history.
“From colonial times to the Civil War, the criminal law in many states expressly differentiated between crimes committed by and against blacks and whites.” Id. at 439. Most of the formal racial distinctions in the law afflicted slaves most severely. Slaves were proscribed from certain activities in which whites were permitted to participate. Randall Kennedy, Race, Crime, and the Law 76 (1997).
Governed by a separate law of crimes, slaves were also subjected to a separate brand of punishment. Slaves, for example, were subject to capital punishment for a wider range of crimes than any other sector of the population. Virginia, for *406instance, defined seventy-three capital crimes applicable to slaves but only one— first degree murder — applicable to whites.
[Id. at 76-77.]
In colonial Georgia, the criminal code provided for an automatic death sentence for blacks who committed murder, while others committing murder could receive a life sentence. McCleskey, supra, 481 U.S. at 329, 107 S.Ct. at 1786, 95 L.Ed.2d at 302 (Brennan, J., dissenting).
Free blacks, like slaves, were also targets of racist criminal laws and faced more severe criminal penalties than whites. Ursula Bentele, Race and Capital Punishment in the United States and Africa, 19 Brook. J. Int’l L. 235, 255 (1993). Several states, including Kansas and Virginia, castrated blacks convicted of raping or attempting to rape a white woman but merely imprisoned whites convicted of similar offenses. Id. at 254 & n. 70. Georgia went one step farther. While white men convicted of rape were merely imprisoned or fined, black men convicted of raping or attempting to rape a white woman were executed. McCleskey, supra, 481 U.S. at 329-30, 107 S.Ct. at 1786, 95 L.Ed.2d at 302 (Brennan, J., dissenting); Bentele, supra, 19 Brook. J. Int’l L. at 254-55.
These antebellum formal racial distinctions in the law were not limited to the South. In the colonial era, New Jersey laws also facially discriminated against slaves and free blacks. Leigh B. Bienen et al., The Reimposition of Capital Punishment in New Jersey: The Role of Prosecutorial Discretion, 41 Rutgers L.Rev. 27, 49 (1988) (recounting New Jersey’s history of executing slaves for crimes for which whites were not capitally prosecuted). In addition, blacks were subject to the death penalty for a wider range of crimes than whites. Id. at 48. Even after the elimination of separate slave courts, only slaves could be sentenced to death for manslaughter, grand larceny and any other felony or burglary. Id. at 49. The racist distinctions carried over into the manner of executions as well. While whites were merely hanged, slaves often had their hands cut off prior to being hanged or burned alive. Ibid.; see Rev. Edwin F. Hatfield, History of *407Elizabeth New Jersey; Including The Early History of Union County 363-64 (1868) (detailing New York reports in 1741 of “Negro Conspiracy to burn the city and to murder the white population,” noting that 154 blacks were imprisoned, fourteen burned at the stake, and eighteen hanged; and that “these [horrible] transactions were- [ ] not confined to New York” but extended to New Jersey where two fleeing blacks were “condemned to the same inhuman fate”).
The abolition of slavery did not end the overt racial discrimination in this Nation’s criminal law. Immediately after the conclusion of the Civil War, Southern states enacted the Black Codes, which reinstated a dual system of criminal prosecution. Kennedy, supra, at 84-85; Bentele, supra, 19 Brook. J. Int’l L. at 255. South Carolina devised a separate court system for criminal trials of black defendants. Kennedy, supra, at 85. As was the case under antebellum laws, blacks were forbidden from engaging in activities in which whites could partake. Id. at 84-85. They were given harsher punishments than whites for committing similar offenses. Bentele, supra, 19 Brook. J. Int’l L. at 255. In Alabama, Mississippi, and South Carolina, only blacks could be executed for raping a white woman. Kennedy, supra, at 85.
Despite the abolition of the Black Codes during Reconstruction, racial discrimination in the administration of criminal justice remained. Several states “expanded the capital offenses in existing law, declared the death penalty discretionary, and then applied it almost exclusively to blacks. The institutionalized discrimination had not altered; equality was no closer.” Kendra Meinert, Criminal Injustice: Continuing Racial Inequities in Death Penalty Sentencing, 22 Sw. U.L.Rev. 1177, 1184 (1993). Accordingly, the death penalty was applied disproportionately against black defendants long after the annulment of de jure racial discrimination in criminal law. See Stan Robin Gregory, Capital Punishment and Equal Protection: Constitutional Problems, Race, and the Death Penalty, 5 St. Thomas L.Rev. 257, 257-58 (1992).
*408New Jersey’s experience with capital punishment is rife with evidence of racial discrimination. It parallels the national experience. See Bienen, supra, 41 Rutgers L.Rev. at 49; Hugo A. Bedau, Death Sentences in New Jersey 1907-1960, 19 Rutgers L.Rev. 1 (1964) (finding that race of the offender was a significant factor in determining the verdicts of death penalty cases in the first half of the twentieth century).
The proven existence of racial discrimination in contemporary capital punishment regimes throughout the century bolsters the conclusion that racial bias infects the administration of capital punishment. The presence of statistical evidence of racial disparities in other states shows that the AOC’s data indicating discrimination in this State are neither unexpected nor isolated. Rather, the discrimination in this State comprises a part of a nationwide pattern of racial discrimination in the administration of the death penalty since Gregg v. Georgia, 428 U.S. 153, 96 S.Ct. 2909, 49 L.Ed.2d 859 (1976), the United States Supreme Court decision that marks the inception of the modern era of capital punishment in this country.
In 1990, the General Accounting Office (GAO) compiled a report on racial discrimination in capital sentencing. The GAO examined twenty-eight studies from various states, including New Jersey, regarding racial disparities in the death penalty. United States General Accounting Office, Death Penalty Sentencing: Research Indicates Pattern of Racial Disparities (Feb.1990), reprinted in 136 Cong. Rec. S6889-90 (daily ed. May 24, 1990). The GAO found that a white-victim effect existed in eighty-two percent of the studies. Ibid. The GAO noted that “[t]his finding was remarkably consistent across data sets, states, data collection methods, and analytic techniques.” Ibid.
Many other studies have found that black defendants and killers of white victims are more likely to be sentenced to death than other similarly culpable defendants. See Bentele, supra, 19 Brook. L. Int’l L. at 256-57 (“[Cjomprehensive studies have shown that the death penalty is more likely to be imposed and *409carried out if the offender is black, particularly if the victim is white.”); Erwin Chemerinsky, Eliminating Discrimination in Administering the Death Penalty: The Need for the Racial Justice Act, 35 Santa Clara L.Rev. 519, 522 (1995) (“Many other studies confirm [the] discriminatory pattern in capital decision-making.”); Meinert, supra, 22 Sw. U.L.Rev. at 1194 (“A variety of studies has indicated that blacks are sentenced to death, at higher proportions than whites, and for less serious crimes.”).
For example, an analysis of 11,425 capital murders from 1977 to 1984, in the thirty-two states where capital punishment was imposed, reveals that the killer of a white victim is nearly three times more likely to be sentenced to death than the killer of an African-American. An analysis of capital sentencing in twenty-three states between 1976 and 1994 found that while African-Americans constitute fifty percent of the murder victims, they constitute only fifteen percent of the victims in death penalty cases.
[Stewart F. Hancock, Jr. et al., Race, Unbridled Discretion, and the State Constitutional Validity of New York’s Death Penalty Statute — Two Questions, 59 Alb. L.Rev. 1545,1560 (1996) (footnote omitted).]
Referring to the plethora of evidence of racial discrimination, a United Nations report determined that race is a “key determinant[ ] of who will, and who will not, receive a sentence of death.” U.N. ESCOR, 44th Sess., Provisional Agenda Item 10, ¶ 148, U.N. Doc. E/CN.4/1998/68/Add.3 (1998). Consequently, we can safely conclude that racial discrimination infects the administration of the death penalty throughout the Nation. See Blair v. Armontrout, 916 F.2d 1310, 1351 (8th Cir.1990) (Heaney, J., concurring and dissenting) (“Race plays an especially influential role in capital sentencing decisions.”); Bright, supra, 35 Santa Clara L.Rev. at 461 (“Sentencing patterns confirm that racial prejudice plays a role in imposition of the death penalty.”). That conclusion comports with our Nation’s history and legacy of racial discrimination. “Indeed, given the past and present realities of racial conflict in the United States, it would be surprising if racial bias did not affect ... capital sentencing.” Kennedy, supra, at 348.
The existence of statistical evidence of racial discrimination in the imposition of capital punishment in foreign jurisdictions should deflate the suspicion that the statistical evidence of racial discrimi*410nation in New Jersey is either inaccurate or the product of mathematical randomness. The experience in other states suggests that the statistical data, revealing that the races of the defendant and victim reflect the reality of racial discrimination in this State’s application of the death penalty, are accurate.
C.
When reviewing defendant’s claim that the administration of the death penalty is tainted by racial discrimination, we must not view the statistics showing such discrimination in isolation. To minimize the impact of indications of race discrimination in light of “this State’s strong and steadfast public policy against [such] invidious discrimination ... would be blind and impervious to the lessons of history.” Taylor v. Metzger, 152 N.J. 490, 510, 706 A.2d 685 (1998). The Special Master notes: “Most of [the statistical data] trend toward showing an unquantifiable race effect, ... [t]he statistical results now before the Court are not comforting, and do not refute claims of race bias,” Special Master Report, supra, at 44; and “[d]espite the undependability of the statistical evidence,” it is “troubling that so much of it tends in the same direction____ What one person thinks he sees in the dusk may be disregarded. What ten people independently think they see may really be there.” Id. at 43.
We should stop questioning our senses and admit, at the very least, that what we see is, in all likelihood, an administration of the death penalty that is racist and unequal. Certainly, these data “relentlessly” suggest a risk of racial discrimination. See McCleskey, supra, 481 U.S. at 314, 107 S.Ct. at 1778, 95 L.Ed.2d at 292 (holding that the statistics must demonstrate a “constitutionally significant risk of racial bias affecting the ... capital sentencing process.”) (emphasis added); see also Turner, supra, 476 U.S. at 36 n. 8, 106 S.Ct. at 1688 n. 8, 90 L.Ed.2d at 36 n. 8 (holding that “the only question” is at what point the risk of racial prejudice influencing a criminal trial becomes constitutionally unacceptable) (emphasis added). Thus, while not all of the results obtained from *411the new data when parsed and segregated are statistically significant, they comport with historical experience and they are consistent across studies, even if not to the same degree in each case.
The majority explains away the compelling statistical evidence by finding fault in the methodology of each model presented. According to the Court: (1) the consistent trends across models based on the raw data are to be disregarded because different coding decisions produce different results, thus highlighting the “dangers inherent in the improper use of statistics,” ante at 305, 724 A.2d at 155; (2) the AOC culpability rankings cannot be relied upon because they integrate the AOC raw data coding decisions and because the size of the database renders the models unsuitable for the purpose of determining the presence of race discrimination, ante at 308, 724 A.2d at 156; (3) the Tukey models, commissioned by the Special Master himself, are flawed because although they cure the parsimony problem, they are not “the most effective means of studying possible racial bias,” ante at 312, 724 A.2d at 158; and (4) Dr. Allison’s models are flawed, again, due to the “suspect culpability rankings and the small size of the database,” ante at 312, 724 A.2d at 158.
The majority’s absolute unwillingness to recognize any non-statistical indications of race discrimination in the capital sentencing process, combined with its finding that none of the models used by the Court to examine systemic discrimination is methodologically reliable, renders our proportionality review in prior cases, as well as in this case, meaningless. The Court has established a level of statistical proof necessary to prove race discrimination that is, in my view, focused on theoretical perfection; and, in doing so, renders this process one that is academically futile and jurisprudentially defeatist. The problem with an uncritical reliance on statistical perfection is that the Court “assume[s] that science can provide [a] precise answer [to the question we are raising], but science never comes packaged this way. The quality of the methodology and data analysis in scientific studies varies, and the results are often contradictory.” Sanders, supra, 46 Stan. L.Rev. *412at 17. To require the level of certainty that we have erected is to annul the reliability of any conclusions. Note, Easing the Fear of Too Much Justice: A Compromise Proposal to Revise the Racial Justice Act, 30 Harv. C.R.-C.L. L.Rev. 543 (1995) (noting that requiring statistical significance for determining the presence of race discrimination in capital sentencing “will have absolutely no effect in jurisdictions where the number of death sentences is so small as to preclude any statistically reliable conclusions”).
Although the majority debunks and denigrates current statistical results indicating the presence of racial discrimination, it is inconceivable that this Court, with the evidence around us, would allow a defendant to be executed. If, as the majority asserts, we have been unable to apply a proper methodology and to construct models that are reliable measures of race discrimination, we should place a moratorium on executions in this State until such models have been created. That said, however, I doubt that the Court will ever be satisfied that “sufficiently reliable,” ante at 313, 724 A.2d at 159 (quoting Rubanick, supra, 125 N.J. at 449, 593 A.2d 733), results have been attained even after its anticipated improvements of proportionality review.
The Court ought not to limit the exercise of its responsibility— its oft-repeated promise to discover and extirpate racial discrimination in capital murder prosecutions — to the most exacting and sterile standards set by academic statisticians. With this State’s first execution in over thirty-five years looming on the horizon, it is necessary to turn to other possible indicators of race discrimination to supplement our statistical findings. Tragically, the Court’s decision risks the execution of people sentenced to death under a capital-sentencing regime rife with racial discrimination. That is a risk we cannot take.
In my opinion, there is an unavoidable risk that racial discrimi- • nation is a potent factor influencing capital prosecutions and death sentences. The State’s capital murder statute is unconstitutional, and defendant’s death sentence should be vacated.
*413II
The Court employs proportionality review in an attempt “to ensure that the death penalty is being administered in a rational, non-arbitrary, and evenhanded manner, fairly and with reasonable consistency.” Marshall II, supra, 130 N.J. at 131, 613 A.2d 1059; see Ramseur, supra, 106 N.J. at 327, 524 A.2d 188; Gregg, supra, 428 U.S. at 206, 96 S.Ct. at 2940, 49 L.Ed.2d at 893. Proportionality review also plays a part in upholding the constitutional requirement that capital punishment be individualized and imposed only subsequent to a sentencing proceeding at which evidence unique to the defendant can be presented to, and considered by, a jury. See Ramseur, supra, 106 N.J. at 331, 524 A.2d 188; Lockett v. Ohio, 438 U.S. 586, 605, 98 S.Ct. 2954, 2965, 57 L.Ed.2d 973, 990 (1978).
In the course of our first reflection on the capital-murder statute and its provisions for proportionality review, this Court recognized that the dual aims of uniformity and individualization created an “inherent paradox in the process.” Ramseur, supra, 106 N.J. at 330, 524 A.2d 188. I warned that “the contrary burdens” would be “extraordinarily difficult, if not impossible, to reconcile.” Id. at 351, 524 A.2d 188 (Handler, J., dissenting). They were, I claimed, “fundamentally contradictory — perhaps unattainable.,.Id. at 347, 524 A.2d 188 (Handler, J., dissenting).
Since then, we have continually sought to enhance the reliability of proportionality review, and perhaps quell the competing constitutional aims of uniformity and individualization, by complementing our statistical analysis with a reasoned review of precedent. In the same pursuit, we have determined that “[a] capital sentence is excessive and thus disproportionate if other defendants with characteristics similar to those of the defendant under review generally receive sentences other than death for committing factually-similar crimes in the same jurisdiction.” Martini II, supra, 139 N.J. at 20, 651 A.2d 949; see DiFrisco III, supra, 142 N.J. at 160, 662 A.2d 442; Bey IV, supra, 137 N.J. at 343, 645 A.2d 685; *414Marshall II, supra, 130 N.J. at 131, 613 A.2d 1059; see also Gregg, supra, 428 U.S. at 205, 96 S.Ct. at 2940, 49 L.Ed.2d at 892 (“[NJo death sentence is affirmed unless in similar cases through the State the death penalty has been imposed generally.”) (citation omitted).
Yet, statistical frequency analysis and precedent-seeking review each suffer from deficiencies that accentuate the impossibility of ensuring both consistent, non-arbitrary application of the death penalty and individualized decision making. I have discussed the “persistent defects of the Court’s proportionality review methodology” on previous occasions. DiFrisco III, supra, 142 N.J. at 212, 245, 662 A.2d 442 (Handler, J., dissenting); see generally Martini II, supra, 139 N.J. at 82-106, 651 A.2d 949 (Handler, J., dissenting) (discussing Court’s use of reversed death sentences in proportionality review universe, failure to identify a standard by which to measure general imposition of death penalty, and inherent subjectivity of precedent-seeking review). Those problems remain and affect the soundness of the review in this case. For instance, the universe of cases to which Loftin was compared should not include those in which death sentences were reversed, accord infra at 416 n. 16, 724 A.2d at 212 n. 16, defendant’s designated comparison category does not include all relevant cases, and the database remains too small to provide a rehable review.
Some problems previously present are now more conspicuous. Defendant’s proportionality review illustrates that precedent-seeking review as conducted by this Court rests upon distinctions so idiosyncratic to defendants’ profiles — is so individualized — that generating a class of defendants by which to accurately measure whether a defendant’s sentence was arbitrarily imposed becomes a near impossibility. Statistical frequency analysis, on the other hand, succeeds in assessing defendants according to a defined set of categories, but lacks any formal definition of when the death penalty is “generally” received. As such, the statistical data compiled for frequency analysis by the AOC are less definitive than they are suggestive as to whether or not a defendant’s *415sentence is proportional. In short, the data merely provide a basis for the Court’s unguided determination of what constitutes a general occurrence. Proportionality review, I maintain, is inherently subjective and therefore unable to provide constitutional legitimacy to capital punishment. Nevertheless, if we are to venture toward this illusory goal, the Court must adhere to a proper designation of what it means for a death sentence to be “generally” received and to some definition of “similarly situated defendants.” Only then will we truly be able to recognize that a death sentence is disproportionate.
Further, given this Court’s decision to appoint a Special Master to evaluate the methods we employ in proportionality review, see ante at 265, 724 A.2d at 135, I find the application of existing models in defendant’s ease disconcerting. To question the integrity of the methods, and yet apply them prior to receiving the Special Master’s recommendations, disregards the crucial role of previous proportionality determinations in our appellate review of subsequent capital convictions and sentences. If the Court is going to declare that the system of review requires evaluation and perhaps reconfiguration, one must assume that if the Special Master’s findings impugn current proportionality review methodology, then any death sentence reviewed by current methods, including Loftin’s, will be reconsidered. Otherwise, the Court should halt all proportionality reviews, beginning with defendant’s, until such scrutiny has been applied. We must recognize that in the absence of a reliable statistical assessment, precedent-seeking review is unable to provide reasonably meaningful proportionality review on its own.
A.
The Court undertakes two types of analyses to assess the proportionality of a defendant’s sentence. Frequency analysis is a statistical method that compares the rate at which categorically similar defendants have been sentenced to death. The second test, dubbed precedent-seeking review, involves a case-by-case *416comparison of defendant’s case to other, factually similar, death-eligible cases.
The perimeter of proportionality comparisons is marked by a universe of similar cases. The universe used by the AOC, which the Court used in the first four proportionality review cases, includes all death-eligible homicides committed since the reinstatement of capital punishment in New Jersey. The Court has assigned a Special Master to review the size of the universe of comparable cases, see id. at 290-291, 724 A.2d at 147-148, but, at present, continues to use a universe comprised of death-eligible defendants, see id. at 288, 724 A.2d at 146.16 After the universe is created, the parameters of proportionality review are defined by what the AOC designates as comparison groups. The AOC has established thirteen comparison groups, see id. at 326 n. 15, 724 A.2d at 165 n. 15, which, for cases that reached the penalty phase, are based on the sentencing jury’s findings regarding statutory aggravating factors; for the majority of the cases in the death-eligible universe, the AOC makes a judgment call. These groupings, which are also used by this Court as a basis for comparison in its precedent-seeking review, have previously been challenged as overly limiting. See, e.g., DiFrisco III, supra, 142 N.J. at 230, 662 A.2d 442 (Handler, J., dissenting). The present case provides an additional example.17
*417B.
Frequency analysis aims to provide a statistical representation of how often individuals who committed offenses similar to defendant have been sentenced to death. There are three different statistical tests used in frequency analysis: the salient-factors test, the numerical-preponderance-of-aggravating-and-mitigating-factors test, and the index-of-outeomes test. This Court has recognized on more than one occasion that, despite the analytic potential. of frequency analysis, the presently small number of sample cases disables the statistical analysis from providing reliable results. See DiFrisco III, supra, 142 N.J. at 171, 662 A.2d 442; Martini II, supra, 139 N.J. at 28-29, 651 A.2d 949; Bey IV, supra, 137 N.J at 350, 645 A.2d 685; Marshall II, supra, 130 N.J. at 173-74, 613 A.2d 1059. I continue to concur with this view. Accord DiFrisco III, supra, 142 N.J. at 237, 662 A.2d 442 (Handler, J., dissenting).
The greatest deficiency of the Court’s frequency analysis is its application of the standard chosen for review. The chronic constitutional infirmity of current proportionality review in large measure emanates from the standards invoked by the Court. This Court’s proportionality review is inherently subjective, evidenced by the oscillation of its ostensible controlling substantive principle. Recognizing the complexity and elusiveness of death-sentencing proportionality, this Court has proffered a number of alternative formulations of its basic principle. For instance, the Court has stated that disproportionality is marked by excessiveness, and that “[a] death sentence is comparatively excessive if other defendants with similar characteristics generally receive sentences other than death for committing factually similar offenses in the same jurisdiction.” Marshall II, 130 N.J. at 153-54, 613 A.2d 1059 (emphasis added) (citation omitted). Hence, a death sentence is not disproportionate if it is .generally received by defendants with similar characteristics who have committed factually similar of*418fenses in the same jurisdiction. Id. at 131, 613 A.2d 1059; Bey IV, supra, 137 N.J. at 343, 645 A.2d 685; Martini II, supra, 139 N.J. at 30, 651 A.2d 949; see also DiFrisco III, supra, 142 N.J. at 171, 662 A.2d 442 (stating “principal inquiry” is whether a case like defendant’s will “generally result in a death sentence”).18 On other occasions, however, the Court has said that in order for a death sentence to be proportionate, it must be: not “aberrational,” DiFrisco III, supra, 142 N.J. at 166, 662 A.2d 442; Bey IV, supra, 137 N.J. at 352, 369, 645 A.2d 685; “not insupportable,” Martini II, supra, 139 N.J. at 22, 651 A.2d 949; not “more like that of similar life-sentenced defendants and less like that of death-sentenced defendants,” Id. at 47, 651 A.2d 949; and, applied to a defendant who is “more likely than other killers to receive the death penalty,” DiFrisco III, supra, 142 N.J. at 171, 662 A.2d 442.
The only standard that has any promise of ensuring a degree of proportionality that would be both reasonably meaningful and *419sufficiently protective is the basic standard that requires that death sentences be generally received by defendants within the ease profile. To properly apply this standard we must be able to identify a class of “defendants with similar characteristics,” who have committed “similar crimes.” We must also have a defined notion of when a death sentence is or is not “generally” received. Further, a definition of generality with which the death sentence is imposed must be applied consistently and objectively in proportionality review. Hence, I have previously stressed that the conclusion that a death sentence is “generally” received necessarily requires some acknowledgment of a numerical majority or preponderance that can then be equated with proportionality. See DiFrisco III, supra, 142 N.J. at 212, 662 A.2d 442 (Handler, J., dissenting); Martini II, supra, 139 N.J. at 90-91, 651 A.2d 949 (Handler, J., dissenting); Bey IV, supra, 137 N.J. at 408, 645 A.2d 685 (Handler, J., dissenting). In terms of meeting constitutional demands for basic fairness, consistency, and uniformity that can prevent the imposition of a disproportionate sentence, numerical preponderance in the range of sixty to seventy-five percent would reasonably express when a death sentence is “generally” received.
Without the doctrinal discipline and certainty of a defined numerical majority consistent with the concept of generality to demarcate proportionate sentences, it is inevitable that arbitrary death sentences of individual defendants will be lost in the verbiage that rationalizes a death sentence as proportionate. The Court continues to argue that setting a numerical standard would introduce unacceptable arbitrariness in proportionality review. See ante at 322, 724 A.2d at 163; Martini II, supra, 139 N.J. at 20, 651 A.2d 949 (declining to set numerical standard for determining when defendants “generally” receive death sentences); Marshall II, supra, 130 N.J at 152, 613 A.2d 1059 (declining to adopt numerical standard finding disproportionality whenever death-sentencing rate was less than eighty percent). Any arbitrariness derived from a defined numerical standard, however, is far less problematic, and more manageable, than that inherent in the unprincipled and unchecked application of a verbal standard.
*420Failure to equate the chosen standard with a numerical preponderance plagues the Court’s analysis of the salient-factor and index-of-outcomes test results in the present case. In defendant’s case, the sentencing jury found three aggravating factors: N.J.S.A. 2C:ll-3c(4)(a) (prior murder); N.J.S.A 2C:l-3c(4)(f) (murdered to escape detection); and N.J.S.A 2C:ll-3c(4)(g) (felony murder). As a prior murderer with three aggravating factors, Loftin has been assigned to the B(l) subcategory for purposes of salient-factor analysis. The statistics show that B(l) defendants overall are more likely to be sentenced to death than death-eligible murderers generally. In all death-eligible cases, seventeen percent of B(l) defendants are sentenced to death, whereas twelve and one-half percent of death eligible defendants are generally. See Table, ante at 329, 724 A.2d at 166.19 But the data manifest no such disparity at the penalty phase, where twenty-five percent of B(l) defendants are sentenced to death, compared with thirty percent of death-eligible defendants generally. See ibid. Considering prior murderers as a composite — those in categories B(l), B(2) and B(3) — in comparison to death-eligible defendants,20 we discern the numbers show that prior murderers are relatively likely to be sentenced to death. At the penalty phase, fifty-six percent of class B defendants are sentenced to death, whereas thirty percent of death-eligible defendants are generally. See Table, ante at 329, 724 A.2d at 167. In all death-eligible cases, thirty-six percent of class B defendants are sentenced to death, *421whereas twelve and one-half percent of death eligible defendants are generally. Ibid.
Based on the data, the Court reckons that “application of the salient-factors test appears to suggest that prior murderers receive the death penalty more frequently than most defendants in the death-eligible and penalty-trial universes and that they are considered by community consensus to be highly blameworthy. In fact, prior murderers are sentenced to death at a higher rate than any other category of murderers.” Id. at 329, 724 A.2d at 167. So much is supported by the statistics. There can be no doubt that, as a prior murderer, Loftin finds himself in a category where imposition of the death sentence is relatively common; in other words, as a prior murderer Loftin has a heightened chance of being sentenced to death. The Court’s conclusion, however, that “the results of this test ... do support a finding of no disproportionality” ibid, (citation omitted), does not entirely follow. A heightened chance does not always reflect general imposition. In the present case, the death-sentencing rate for prior murderers with at least two aggravators, class B(l), is twenty-five percent at penalty trial and only seventeen percent for all death-eligible eases. Contrary to the Court’s assertion, one cannot reasonably conclude that a class of defendants sentenced to death at most one in four times receives the death penalty “generally.” The standard of “generally received” must reflect a numerical preponderance. See supra at 308-309, 724 A.2d at 156-157.
The same problem appears in the Court’s assessment of the statistical results of the index-of-outcomes test. The index-of-outcomes test “compare[s] cases that are factually dissimilar but that are nevertheless comparable from the perspective of the defendants’ blameworthiness.” DiFrisco III, supra, 142 N.J. at 179, 662 A.2d 442 (quoting Martini II, supra, 139 N.J. at 42, 651 A.2d 949). In order to undertake such a comparison, the Court relies on AOC calculations of culpability. The AOC’s calculations are based on a multiple regression technique that, unlike the salient-factor and numerical-preponderance tests, examines rele*422vant nonstatutory, in addition to statutory, aggravating and mitigating factors. Four multiple regressions are involved. Two involve both statutory and nonstatutory factors. Two involve statutory factors only; of those, one regression considers only defendants who reached the penalty stage, while the other looks to the entire universe of death-eligible defendants. After assigning a weight to each factor, the AOC uses regression analysis to compute a culpability score for defendants, i.e., the defendant’s predicted probability of receiving a death sentence.21 These scores, which range from a high of .99 to a low of .00, are broken down into five culpability levels.22 At each level, the AOC determines the actual rate at which the death sentence is imposed on defendants at that culpability level.
The Court errs in the manner in which it weighs the index-of-outcomes test statistics. The Court, in conclusion, relies solely upon the predicted probability of death statistics; not at all upon the death-sentencing rates. See ante at 335, 724 A.2d at 170. The death-sentencing rate statistics should primarily guide the determination because they provide the measure of general imposition of death sentences. Indeed, the question before this Court on proportionality review is whether or not death sentences are generally received by, or imposed upon, similarly situated defendants. The predicted probability of the death sentence statistic does have some value, if only to compensate for deficiencies in the methodology behind the death-sentencing rate statistic; for instance, defendants in culpability level one with a culpability rating of .05 may differ substantially in death-sentencing rate, despite being in the same level, from those defendants on the upper end of culpability level one, who have culpability level ratings such as .19. *423The predicted probability statistic can thus help to identify potential problems of disparate effect at the margins of any given culpability level.
Nevertheless, the constitutional mandate of individualized consideration in capital sentencing requires that we rely on actual decisions by sentencing juries, and the characteristics of actual felons and their crimes, to determine when a death sentence is generally imposed upon similarly situated defendants. Certainly, we could not point to our surprise at defendant’s death sentence as grounds for reversal; nor should our expectation that a certain defendant will receive the death penalty serve as adequate support for a finding of proportionality. The death-sentencing rate statistic, therefore, should primarily guide this Court’s interpretation of the index-of-outcomes test results. Accordingly, the majority errs by failing to do so in reaching its conclusion.
In this case, the Court concludes that none of the index-of-outcomes test regression results demonstrates disproportionality. See id. at 334-335, 724 A.2d at 170. However, the death-sentencing rate statistics in this case are not so dispositive. Regression results using only statutory factors do show that death sentences are generally imposed — by a numerical preponderance — on defendants sharing Loftin’s culpability level: similarly situated defendants whose cases proceed to the penalty phase are death-sentenced seventy percent of the time; for all similarly situated death-eligible defendants, the death-sentenced rate is fifty-seven percent. See Loftin Report, tbls. 13-14, 17, 21-24; see also ante at 333-334, 724 A.2d at 169-170. The results from the statutory and nonstatutory factor regression, however, do not show general imposition: the death-sentencing rate for defendants in the same culpability level as Loftin proceeding to penalty trial is only seven percent; for similarly situated death-eligible defendants generally, the rate is thirty-eight percent. Ibid.
The Court’s analysis of the index-of-outcomes test suffers from yet another deviation: the Court compares Loftin’s numbers *424indiscriminately to those of previous proportionality review defendants, regardless of whether or not those defendants were similarly situated. See ante at 331-334, 724 A.2d at. 168-170. The question we seek to answer in proportionality review is whether defendants situated similarly to defendant generally receive a sentence other than death. Accordingly, in this case, in assessing the salient-factors test, and on precedent-seeking review, the Court limits comparison to prior murderers. That basis for similarity comparisons is not available or possible in index-of-outcomes analysis; the Court’s index-of-outcomes comparison to prior proportionality review defendants cannot be limited to similarities based on material or salient characteristics common among defendants. Perhaps, therefore, this comparison should be confined to those defendants who share the same culpability level (resultant from a particular regression) with defendant. The index-of-outcomes test analysis undertaken by the Court, however, compares Loftin to all prior proportionality review defendants, across culpability levels and without regard of whether of not those defendants share Loftin’s culpability level. If the Court deems DiFrisco, Marshall and Martini, all of whom fall outside the B category, unfit for comparison on precedent-seeking and salient-factors analysis, what justifies a comparison to those same defendants (or Bey as well) when they fail to share defendant’s culpability level resultant from the index-of-outcomes regression?
Even culpability level, however, may be an inappropriate and insufficient measure or indicator of similarly-situated defendants. For example, one may find it unreasonable to say that Marshall, a contract killer, and Loftin, a prior murderer, are similarly situated even if both were classified as culpability level five; they are both highly culpable, but their crimes were nothing alike. The fact that one may hesitate in describing Marshall and Loftin as similarly situated only emphasizes the point that the singular fact that a defendant was the subject of a prior proportionality review does not justify his comparison with the present defendant.
*425Admittedly, the death sentencing rates for prior proportionality review defendants become relevant as the Court makes the same error, continuing to venture inappropriately outside the AOC categories in assessing the frequency analysis tests, in successive cases. In a sense, then, it is not only proper but necessary to compare defendant’s statistics to those of Marshall, Martini, Di-Frisco and Bey. Theoretically, and ideally, however, such comparisons should be reserved for precedent-seeking review. It makes a mockery of the statistical method to compare defendant’s numbers to those of defendants in other categories, just as it distorts the apparently objective standard of disproportionality to conclude that a relatively higher likelihood constitutes general imposition.
All of this suggests that the phrases “generally received” and “defendants with similar characteristics,” as used in this Court’s proportionality review, have deteriorated into a shibboleth, esoteric terms without clear meaning. It was unfortunate for the Court to use “generally received” as a measure if it did not really envision a numerical preponderance as the standard. The standard measuring disproportionality by general imposition of sentences other than death must require, not that Loftin be more likely than others to receive the death penalty, but rather that, for defendants similarly situated, the death sentence be received in a defined preponderance of cases.
C.
The second component of proportionality review is precedent-seeking review. As its name suggests, the precedent-seeking approach involves a traditional case-by-case comparison of similar death-eligible eases. DiFrisco III, supra, 142 N.J. at 183, 662 A.2d 442. In contrast to frequency analysis, precedent-seeking review is not statistical. Notably, the standard for precedent-seeking review is more equitable and less quantitative in nature. The majority expresses the standard as follows: “Although identical sentences are not required even in like cases, the defendant must not have been ‘singled out unfairly for capital punishment.’ ” *426Ante at 335, 724 A.2d at 171 (quoting Martini II, supra, 139 N.J. at 47, 651 A.2d 949, additional citation omitted).
Precedent-seeking review also differs from frequency analysis in that it' considers, in addition to statutory aggravating and mitigating factors, non-statutory factors “rooted in traditional sentencing guidelines.” DiFrisco III, supra, 142 N.J. at 184, 662 A.2d 442. Those traditional factors of criminal culpability fall within three broad categories: defendant’s moral blameworthiness, the degree of victimization, and the character of the defendant. Id. at 185, 662 A.2d 442. The first step of precedent-seeking review is to define defendant’s culpability based on these categories. Then, the Court attempts to identify and detail a comparison group of cases.23 Lastly, the important aspects of defendant’s case are compared to the key facts from the identified group of comparison cases.
If this Court’s application of frequency analysis is ultimately subjective, precedent-seeking review is inherently so. Precedent-seeking review invokes culpability assessments by the Court, which are, essentially, moral judgments. Further, in employing precedent-seeking review, the Court analyzes individual circumstances in such depth, ultimately distinguishing defendants on terms so discrete, that categories by which to measure whether or not a given sentence is arbitrary remain undefined. As a result, the Court does not so much provide the intended quality check of the death sentence, as it does substantively reconsider the jury’s verdict of death.
The subjectivity inherent in the method is evident in the Court’s review of Loftin’s sentence. Looking at the same facts and circumstances and the same class of comparison cases as the majority, one might reasonably reach a conclusion, contrary to the *427Court’s, that defendant’s sentence was indeed disproportionate. If one looks at additional cases similar to defendant’s, neither included in the AOC category nor considered by the Court on precedent-seeking review, the disproportionality of defendant’s sentence is even more pronounced.
1.
Moral blameworthiness as interpreted by this Court is a composite of factors such as motive, premeditation, justification or excuse, evidence of mental defect or disturbance, knowledge of helplessness of the victim, defendant's age or maturity, and defendant’s involvement in planning the murder. Id. at 203, 662 A.2d 442. The Court finds that Loftin’s motive was to steal money and, perhaps, to escape detection; that the robbery, but not necessarily the murder, was premeditated; that there was no evidence of justification or excuse, or of provocation; that defense expert testimony that Loftin had a borderline personality disorder, one juror’s finding of the c(5)(a) mitigating factor of mental or emotional disturbance, and three jurors’ finding of the catch-all mitigating factor on the ground that defendant was suffering from extreme mental disturbance, were offset by the testimony of a State’s expert, who concluded that “defendant exhibited narcissistic and anti-social personality traits and not borderline personality disorder”; that the victim was helpless and defendant knew it; that although two jurors found age to be a mitigating factor, defendant, in fact, “lived as an adult with adult responsibilities”; and that defendant was solely responsible for planning and executing the robbery-murder. Ante at 336-338, 724 A.2d at 171-172.
The Court’s findings are, for the most part, plausible. Different conclusions may be drawn, however, in a certain respect. Given the facts of this case, the claim that defendant murdered to escape detection cannot serve as a reliable basis of distinction on proportionality review. I have previously noted that “every murder committed during the course of a rape, robbery, or other felony listed in c(4)(g) [felony murder aggravating factor], also automati*428cally falls within the scope of the c(4)(f) factor [escape detection] because every criminal hopes to escape detection.” Loftin I, supra, 146 N.J. at 399, 680 A.2d 677 (Handler, J., dissenting). If “every e(4)(g) case is also a c(4)(f) case,” id. at 400, 680 A.2d 677 (Handler, J., dissenting) (quoting State v. Hightower, 146 N.J. 239, 286, 680 A.2d 649 (1996) (Hightower II) (Handler, J., dissenting)), the c(4)(f) aggravating factor fails to “genuinely narrow” the class of death-eligible defendants, and is, therefore, unconstitutional. Id. at 399, 680 A.2d 677 (Handler, J., dissenting) (citing Zant v. Stephens, 462 U.S. 862, 877, 103 S.Ct. 2733, 2742, 77 L.Ed.2d 235, 249 (1983)). Accordingly, as this Court has recognized, the “mere fact that a killing takes place in the course of a felony is not enough to invoke [the c(4)(f) aggravating] factor.” Id. at 377, 680 A.2d 677.
Here, there is no direct evidence that Loftin murdered to escape detection. Cf. State v. DiFrisco, 137 N.J. 434, 500-01, 645 A.2d 734 (1994) (DiFrisco II) (finding record sufficient to support c(4)(f) factor where defendant confessed he was hired to kill victim in order to prevent victim from “rat[ting] on” hirer). Nor has the State presented adequate circumstantial evidence to show that one of the motives for the murder was to escape detection. Cf. State v. Martini, 131 N.J. 176, 282-84, 619 A.2d 1208 (1993) (Martini I) (finding record sufficient to support c(4)(f) factor where victim had past knowledge of defendant and was in defendant’s presence prior to the murder for extended period of time with defendant undisguised). Therefore, I disagree with this Court’s finding that the State presented ample evidence to support a jury finding of the c(4)(f) aggravating factor. Using the c(4)(f) factor as a basis of distinction on proportionality review, see ante at 339, 343, 724 A.2d at 173, 174, only compounds the error introduced at the sentencing phase.
After reviewing defendant’s moral blameworthiness, the Court next examines the degree of victimization of defendant’s crime. In this category, the Court assesses the violence and brutality associated with the murder and any injury to non-decedent vie*429tims. DiFrisco III, supra, 142 N.J. at 205, 662 A.2d 442. As the Court properly recognizes, “[Loftin’s] was not a particularly violent or brutal killing.” Ante at 338, 724 A.2d at 172. Marsh was killed “execution style,” with a single gunshot to the head. There was no other contact. There was no injury to non-decedent victims. Thus, the degree of victimization associated with Loftin’s crime is low, and does not contribute to making defendant death-worthy.
The last aspect of culpability that must be examined is defendant’s character. When assessing defendant’s character, this Court considers prior record, prior acts of violence, cooperation with authorities, remorse, and defendant’s capacity for rehabilitation. DiFrisco III, supra, 142 N.J. at 205, 662 A.2d 442. Although it is undisputed that defendant did not cooperate with authorities, the record is equivocal with respect to other aspects of defendant’s character. Defendant had a prior record, yet had never been convicted or incarcerated prior to this offense. The Court finds any manifestations of remorse by defendant canceled by defendant’s admission that “his ‘mistake’ was being incarcerated, not that he had killed two human beings.” Ante at 338-339, 724 A.2d at 172. Loftin’s expression of remorse in his allocution, however, is significant. In his allocution, defendant stated that he “never tried to deny [his] guilt,” and that the events
are a traumatic tragedy for all that are involved, especially for the victim’s family and friends. The pain and the hardships that occurred, and still are, are beyond forgiveness. I can only say I truly am remorseful.
^ ^ ^
I again express my true remorse to the family and friends of Gary Marsh for causing so much pain and hardships throughout this whole ordeal.
Finally, although there is little evidence of defendant’s prospects for rehabilitation, there is reason to believe that defendant, who had not undergone psychiatric counseling, would respond positively to treatment. The defense psychiatric expert noted that the symptoms associated with. Loftin’s mental condition would be eased in structured situations. See Loftin I, supra, 146 N.J. at *430331, 680 A.2d 677 (testifying that defendant struggled outside structured environment of the military or auto-mechanic school). In sum, the prior murder is the only factor that contributes to defendant’s deathworthiness.
2.
a.
Defendant’s culpability must be compared to that of similarly situated defendants. The Court restricts its review, in accordance with AOC categorization, to prior, or category B, murderers. See Appendix C, ante at 348-371, 724 A.2d at 178-190 (providing case summaries of category B defendants). Besides Loftin, only one defendant classified as B(l), a prior murderer with two or more aggravating factors, has been sentenced to death. The other five B(l) prosecutions all resulted in life sentences (James Koedatich (IB), George Booker (1 and 2), John Fauntenberry, and Richard Feaster (2)). The death-sentenced defendant, James Koedatich, was substantially more culpable than Loftin. Comparing defendant’s crime to the other B(l) defendants, it is evident that defendant is clearly less culpable than most of them.
The degree of blameworthiness and victimization associated with Loftin’s crime were relatively low. Koedatich, motivated by a desire to commit sexual assault, concocted a detailed plan to murder an eighteen-year-old high school student. He stabbed her multiple times and left her bleeding to death in a water retention tank. Booker’s offense was a brutal torture killing of two lovers. Finding the first victim alone, he beat, suffocated and strangled her to death. He then lay in wait for the other. When she arrived, he handcuffed her to her lover before brutally murdering her in a similar fashion. Feaster’s crime, like defendant’s, was a gas station robbery-murder; however, Feaster assaulted his victim with over forty slashes and stabs. Moreover, Feaster claimed he killed merely for the thrill of it. Fauntenberry is the only B(l) defendant whose culpability approximates Loftin’s. Fauntenberry shot a truck driver once in the head, and robbed him. Faunten*431berry, like Feaster, was offered a plea to a life sentence. Loftin also sought to plead guilty to murder in exchange for a life sentence, but the State refused to make an offer.
With respect to defendant’s character, his record is more positive than those of the other B(l) defendants. For instance, the other defendants had more extensive criminal, if not murder, records than Loftin, who had no prior record and had never been jailed before this offense. Although Loftin had murdered before, that murder, which occurred shortly before the incident in question, was not particularly violent. In contrast, Koedatich may be considered a serial killer; he apparently raped and killed three women in a two-month period and had murdered someone previously. Booker not only had a prior murder conviction and a more extensive prior criminal record than Loftin, but killed multiple victims during his offense and physically injured third parties. Fauntenberry, notably, confessed to five other murders. Even Feaster, the only B(l) defendant other than Loftin who had killed only one other person, had prior convictions for simple assault and possession of marijuana. Moreover, only defendant, among the B(l) defendants, expressed remorse. Only Loftin presented evidence of capacity for rehabilitation, although one might conclude that the youthfulness of Fauntenberry and Feaster offered them some hope of recovery. Cooperation with law enforcement is the only category in which defendant cannot be considered the least culpable of the B(l) offenders; Booker and Fauntenberry both confessed, while Loftin, like Koedatich and Feaster, did not.
In sum, the individual who committed the sole B(l) death-sentenced crime, Koedatich, is substantially more culpable and blameworthy than Loftin. Defendant’s offense is less blameworthy than every B(l) case that was tried capitally, but resulted in a life sentence. In fact, Loftin, who had only one prior murder, is less culpable than the two B(l) cases in which the State accepted a plea to non-capital murder.
Despite these facts, the Court deems defendant’s sentence proportionate. “The greatest distinguishing factor between Loftin *432and the B(l) defendants,” the Court finds, is that “each B(l) defendant presented uncontroverted evidence that he suffered from some mental disease or defect.” Ante at 340, 724 A.2d at 173. The Court also relies on the more general conclusion that defendant’s evidence “is not as compelling” as the evidence presented by the other B(l) defendants. See id. at 340, 724 A.2d at 173.
I disagree with the Court’s conclusion. The mere fact that the State chose to present an expert of its own to contradict defendant’s evidence of mental illness does not dispel the mitigating effect of defendant’s troubled life. Loftin’s mental and emotional problems were noticeable as early as the age of six when he burned down his family’s home. When his father abandoned the family, defendant was one of seven children raised by his mother in poverty. His adult life was filled with instability in work and personal relationships — defendant never received psychiatric treatment. A defense psychology expert testified that defendant suffered from moderate to severe symptoms of borderline personality disorder at the time of the offense. Loftin I, supra, 146 N.J. at 330-31, 680 A.2d 677. At least one juror presumably credited the defense psychologist’s conclusion by finding the c(5)(a) (extreme mental or emotional disturbance) mitigating factor. Id. at 322, 680 A.2d 677. Moreover, the seventeen reasons jurors found to support the catch-all mitigating factor, N.J.S.A. 2C:ll-3c(5)(h), included the fact that defendant was upset by the loss of his first son; that he was traumatized by the fire he set as a youth that burned down the family home; and, that he was emotionally impoverished growing up. Ibid. The mere fact that the State’s psychiatric expert recognized that Loftin had “some signs of emotional distress,” and, rather than finding that Loftin suffered from a borderline personality disorder, concluded that Loftin “exhibited narcissistic and antisocial personality traits,” id. at 332, 680 A.2d 677, cannot soundly become a basis for the Court’s decision to place defendant’s case alongside those generally resulting in the death penalty. It is a dubious distinction on which to rest a death sentence. This chimera aside, a comparison of *433Loftin’s case to that of all other B(l) cases reveals disproportionality.
In distinguishing Loftin from the B(2) and B(3) defendants, the Court encounters a set of defendants who, by classification alone, should be less culpable than defendant.24 The Court’s conclusion that Loftin “is more culpable than the B(2) and particularly the B(3) defendants” purely because he is classified in a more culpable category, ante at 341, 724 A.2d at 173, is too hasty. Indeed, distinctions based on class alone dispel all illusions of individualized consideration.
To assess Loftin’s sentence in a proper manner, the Court must rely upon distinctions that are less than secure to find his sentence proportionate. With regard to blameworthiness, the Court notes that “aspects” of Loftin’s crime, such as his “pecuniary motive, execution-style murder, which presented no risk to third persons, appears less deathworthy.” Id. at 342, 724 A.2d at 174. To be sure, defendant’s blameworthiness is about average for B(2) or B(3) defendants. Godette and Pennington had the same pecuniary motivation as Loftin. Several of the other B(2) or B(3) defendants had motives that are clearly more blameworthy than defendant’s pecuniary motive. Bey and Vasquez committed their offenses for their own sexual gratification. Muhammad may have been motivated by racial animosity. Purnell and Biegenwald murdered someone after a dispute over drugs; this may have been the case with Muhammad as well. Biegenwald appears to have killed for the mere thrill of it. On the other hand, Coyle, Erazo, Nieves, and Ramseur — all excited by anger stemming from different love affairs — are arguably less culpable. Similarly, Williams *434murdered his own mother, presumably after a family dispute; however, he also stole her ear and sought to benefit financially from her death. Thus, Loftin appears to fall within the mid-range of culpability in terms of motivation. Although several less culpable defendants received death sentences (Coyle, Erazo, Ramseur and Purnell), all of those death sentences were overturned; some of the sentences, such as Ramseur’s, were so grossly defective that they should probably not be accorded much weight.
Loftin suffered from a mental or emotional impairment that affected his conduct. However, this was also true of Bey, Biegenwald, Coyle, Erazo, Godette, Pennington, and Ramseur, most of whom received death sentences. Again, despite substantial evidence of Loftin’s mental impairment, the Court finds defendant’s case distinguishable from the B(2) death-sentenced defendants and from all B(3) defendants that presented similar evidence of emotional disturbance merely because the evidence in this ease was controverted by the State. See id. at 342-344, 724 A.2d at 174-175 (distinguishing Loftin from B(2) death-sentenced defendants); id. at 344, 724 A.2d at 175 (distinguishing Loftin from all B(3) defendants). The Court’s reliance on uncontroverted evidence of mental illness as the distinguishing factor goes too far. In State v. Bey, 129 N.J. 557, 588, 610 A.2d 814 (1992) (Bey III), the State presented rebuttal evidence by a forensic psychiatrist and forensic psychologist who testified that, contrary to defendant’s evidence, he did not suffer from organic brain dysfunction or organic personality disorder, but simply from antisocial personality disorder. The State makes a very similar argument here. See Loftin I, supra, 146 N.J. at 332, 680 A.2d 677. Looking to other measures of blameworthiness, in no other case was the age mitigating factor found, although Bey, Coyle, and Nieves were about the same age as Loftin. The victims of Biegenwald (1), Godette, Muhammed, and Vasquez were clearly helpless, while in Loftin’s case the record is silent on the matter. In sum, little difference is evident when comparing defendant’s blameworthiness to that of the B(2) and B(3) defendants.
*435Defendant’s character also falls within the mid-range of culpability compared to the B(2) and B(3) defendants. The B(2) and B(3) cases include many defendants no worse than defendant. Only Biegenwald had a far more serious prior record — four or five prior murders, mostly of teenage girls. The other defendants murdered only once, although several had criminal records. See, e.g., Biegenwald (armed robbery); Coyle (multiple robberies); Godette (possession of a controlled dangerous substance); Ramseur (weapon possession); Williams (assault and battery, unlawful possession of a weapon, possession of a controlled dangerous substance). Although Biegenwald, Coyle, Muhammed, and Williams expressed an utter lack of remorse, Bey expressed remorse for his offense, and Nieves and Vasquez confessed. Yet, no other defendant produced evidence demonstrating potential for rehabilitation.
Disproportionality is present, though, in the degree of victimization associated with the defendant’s crimes. Several of the B(2) and B(3) cases involved injury or terror to non-decedent victims. Muhammed shot the victim in front of his companion and her father and threatened to kill witnesses. Nieves shot the victim while she was driving with her six-year-old son. Pennington killed a woman in her daughter’s presence. Ramseur threatened to kill the victim’s children. Further, many of the B(2) and B(3) cases involved a great deal of violence. Bey kidnapped, raped, strangled, and stomped on the victim and then continued to beat her after it was apparent that she was dying. Biegenwald (1) kidnapped the victim, shot them four times, and then drove the body to his house so that his roommate could understand the thrill of killing. Coyle shot his victim four times, three while she was laying prone on the ground. Erazo stabbed his victim eight times as she pleaded for her life. Godette strangled and beat his victim six times with a hammer. Purnell stabbed the victim fifteen times and slashed the victim’s throat. Ramseur inflicted thirteen stab wounds on the victim, and may have returned to the scene to inflict more punishment. Vasquez bound, raped, and strangled his victim. Williams stabbed his victim nine times and left the *436victim’s body to decompose in a laundry room. No other crime, except perhaps Biegenwald’s (2), who had a much worse prior record, involved as little victimization as Loftin’s.
Thus, comparing Loftin’s culpability to the B(2) and B(3) subcategories of defendants, who should theoretically be less culpable, we find that defendant is, in fact, one of the least culpable defendants — if not the least culpable — with respect to victimization, arid average with respect to the factors inherent in blameworthiness and character. He is surely the least culpable death-sentenced prior murderer, and one of the least culpable of all class B defendants, whether sentenced to life or death.
b.
In two previous proportionality reviews, the Court has conducted comparisons with defendants that fall outside the defendant’s particular salient-factor grouping. See, e.g., Martini II, supra, 139 N.J. at 50-51, 651 A.2d 949 (comparing multiple categories of defendants because facts of case could qualify defendant for more than one group); Marshall II, supra, 130 N.J. at 175, 613 A.2d 1059 (comparing category of cases proposed by defendant). In the last proportionality review case, the Court reaffirmed its commitment to restricting comparisons to like eases, “adhering to the AOC’s ‘unique assignment’ of each ease to a specific category.” DiFrisco III, supra, 142 N.J. at 169, 662 A.2d 442 (citation omitted). Even so, the Court adjusted the defendant’s comparison group when conducting precedent-seeking review. Id. at 186, 662 A.2d 442.
Understandably, some limits must be placed on what comparisons are made. I believe, however, that limiting comparison to category B in the present ease, is too restrictive. I would assess the propriety of defendant’s death sentence by comparing it with a number of cases outside the category. See App. D. infra at 449-453, 724 A.2d at 230-232 (providing comparison cases summaries of defendants outside category B). Defendant is a prior murderer. Therefore, comparisons should be limited to other cases where the defendant killed more than one person. Nevertheless, *437comparison to cases that involve multiple victims is appropriate. The multiple-murder cases that involved sequential murders are most like defendant’s case because, but for the prosecutor’s discretion to charge the crimes together, the prosecutor could have waited for a conviction on one murder and then proceeded on the second murder charging the c(4)(a) prior murder aggravating factor. In short, four defendants — John Lee Allen, Kim Kuchler, Frank Marsini, and Ronald Mazique — would have fallen within the B category had the State so chosen to prosecute them. See id. at 449-453, 724 A.2d at 230-231. Defendant should also be compared to two A class defendants, Joseph Harris and Hector Sanabria, whose crimes would have placed them within the B category had their subsequent offenses not involved multiple victims. See id. at 449-451, 452-453, 724 A.2d at 230-231, 231-232.
Even the most simplistic comparisons to those cases reveal that defendant’s offense is less blameworthy and defendant less culpable. All of these defendants, except Kuchler, had motivations at least as blameworthy as defendant. Harris sought sadistic revenge for a firing. Sanabria wanted to enhance his criminal drug enterprise. Masini was apparently motivated by a desire for sexual gratification. Allen and Mazique were both looking for money.
Every one of these crimes was more horrific than defendant’s offense. Allen killed a person who was terrified and pleading for his life. Harris hunted down particular victims with lengthy planning and premeditation. Kuchler chopped up his sister and her husband. Marsini sexually assaulted and murdered four elderly victims. Mazique brutally murdered a grandmother and her six-year old grandson with a hammer; and Sanabria shot two different victims.
Each of these defendants had a more substantial criminal record than defendant. Allen murdered two people during the instant offense and had previously murdered another. Harris murdered four people, had previously murdered one, and had *438kidnapped, sexually assaulted, and threatened other victims. Kuchler committed two murders in addition to one prior murder. Masini killed four victims. Mazique’s was a double homicide and he had killed two people previously; and Sanabria, who killed two people, had killed one person previously and was implicated in another murder at the time of the prosecution.
Even so, Allen, Kuchler, Marsini, and Sanabria were permitted to plead guilty in exchange for a life sentence. Although both Harris and Mazique were capitally tried, only Harris received a death sentence. Loftin is clearly less deathworthy than Harris, the sole death-sentenced offender, and less deathworthy than all of the life-sentenced cases. Thus, gross disproportionality is evident in comparisons to this select group of cases.
3.
Apart from defendant’s status as a prior murderer, the facts of this case do not demand the death penalty. Defendant’s sentence is disproportionate.
Nevertheless, it is clear from the Court’s review that Loftin’s is a close case. As the preceding analysis underscores, one might determine that Loftin’s sentence is disproportionate as readily as one may find it proportionate. The fact that conflicting interpretations are equally justifiable inevitably leads to the conclusion that precedent-seeking review is a subjective exercise.25
*439Even if one accepts that the Court must engage in some value judgments in making determinations about proportionality, precedent seeking review suffers from a more fundamental problem. The process involves making critical distinctions between defendants based upon individual characteristics and circumstances. The comparisons the Court is required to make do not always yield obvious or objective results. Pressed to search for possible differences, this validating process, which properly looks to see if similar defendants are generally sentenced to death or life for committing similar offenses, becomes an inquisition of the defendant’s idiosyncrasies.
The results of this distorting inquiry are unfortunate and problematic. The exercise encourages making something out of nothing; microscopic and hypercritical examination in this context *440will produce an exaggerated magnification of minutiae. Further, the institutional investment in individualized review completely overwhelms its interest in achieving uniformity. Precedent-seeking review that strains to make distinctions renders it nearly impossible to assemble a defined class of defendants from which one could make a determination of whether the death penalty was generally imposed. If every capital defendant is so unique, the Court will hardly ever — perhaps never — be able to say that a death sentence is generally imposed on a class of defendants. Precedent-seeking review’s inherent lack of defined categories of defendants thus makes the detection of arbitrariness a moot point — where all defendants are ultimately distinguishable from one another, one’s death sentence can never be an aberration, no less disproportionate.
*439As the majority itself recognizes, a "value judgment is built into practically every measurement” of proportionality. That inescapable entanglement with subjective moralistic judgment renders proportionality part of the problems rather than the remedy to the problems inherent in the statutory scheme. The Court's decision only perpetuates the myth that the death penalty can be imposed in an objective, principled, non-random manner, and that any given death sentence can be validated as a fair, reliable expression of just punishment or community values. The Court thus makes comparison after comparison with absolutely no legal or logical foundation for support. Is it worse to kill for money or for hatred? Is it worse to kill over a woman or over a dog? Is it worse to kill to support a gambling habit or to support a drug habit? Is it worse to kill a relative or a stranger? To pose those questions is to pose insoluble moral conundrums.
[Marshall II, supra, 130 N.J. at 273-74, 613 A.2d 1059 (Handler, J„ dissenting) (citations omitted).]
*440D.
“[N]o death sentence can be validated under a process of review that is extraordinarily vague, rife with contradictions, wildly inconsistent, and inextricably mired in subjective valuations and intuitive moral judgments.” Martini II, 139 N.J. at 109, 651 A.2d 949 (Handler, J., dissenting). Loftin’s case emphasizes these problems. The differences identified by the Court do not explain disparate treatment of Loftin. The Court’s technical discussions lead only to the conclusion that the process is unprincipled. Despite the Court’s best efforts to do so, there is no way to escape the conclusion that the methods cannot be applied consistently and objectively, and that proportionality review in this form cannot lead to a result in which we can be confident.26
*441Although this Court has acknowledged that frequency analysis is unreliable, it nevertheless finds precedent-seeking review sufficient to determine the proportionality of defendant’s sentence in this case. As currently conducted, proportionality review does little to assess the arbitrariness of a defendant’s sentence; if anything, it merely adds an arbitrary step to what is already suspected to be an arbitrary process. Even worse, we apply this suspect methodology in a situation where defendant’s life depends upon the result.
That said, I would not conduct proportionality review of the defendant’s death sentence until this Court has adopted a system for proportionality review from which we can confidently deduce that a defendant was fairly sentenced to death. In anticipation of the Special Master’s report, I refrain from commenting on what such a methodology might be. At a minimum, however, we must assure that defendants will be compared to a defined class of those similarly situated, and that the relative likelihood of a death sentence within a class is taken into account rather than the relationship between the defendant and the class. We must also be certain that the comparable class of defendants is drawn from a universe of all those death eligible, not merely those sentenced to death.
In that regard, we should finally acknowledge that the Legislative amendment to N.J.S.A 2C:ll-3e, L. 1992, c. 5 (eff. May 12, 1992), which limits the universe of cases available for proportionality review to those in which a death sentence has been imposed, is unconstitutional. I have previously expressed doubts regarding the amendment, see DiFrisco III, supra, 142 N.J. at 218 n. 6, 662 A.2d 442 (Handler, J., dissenting), Martini II, supra, 139 N.J. at 83 n. 1, 651 A.2d 949 (Handler, J., dissenting); Bey IV, supra, 137 *442N.J. at 402 n. 1, 645 A.2d 685 (Handler, J., dissenting), but logic, experience, and this Court’s precedent now compel the conclusion that the 1992 amendment effectively abolishes proportionality review as a meaningful procedural safeguard against the arbitrary and capricious imposition of the death penalty. It is inescapably clear that proportionality review based on a universe that does not comprehensively encompass all cases deemed to involve similar defendants who are death-eligible under a uniform and reasonable standard would be constitutionally deficient. Postponing a ruling on the constitutionality of the amendment only compounds the uncertainty and confusion that already surrounds our imposition of capital punishment.27
*443I am in complete agreement with the forceful reasoning of the Court that, under the separation of powers doctrine, the Court’s power, derived from the its constitutional authority to review convictions and death sentences in capital cases, to determine the necessity and standards for proportionality review of death sentences, is paramount. See ante at 279-285, 724 A.2d at 142-145. Given the terms of the amendment at issue, however, and the enhanced protections this Court has deemed necessary to provide the capitally accused, such reasoning must ultimately lead to the conclusion that the Legislature has exceeded its constitutional authority in this case.
The New Jersey Legislature’s 1992 amendment of N.J.S.A. 2C:ll-3e, in pertinent part, limits proportionality review “to a comparison of similar cases in which a sentence of death has been imposed.”28 The amendment thus restricts proportionality review *444to comparisons with those few capitally-charged cases that not only reached the penalty phase, but also resulted in the death penalty. This Court has expressly addressed and rejected the possibility that a universe restricted to cases in which the death penalty has been imposed could serve as a basis for meaningful and principled proportionality review; none of the three purposes that this Court has identified for proportionality review — assessing disproportionality, ensuring appropriate exercise of prosecutorial discretion, and monitoring impermissible racial discrimination either by prosecutors or by juries — could be served by a proportionality review methodology so restricted. Marshall II, supra, 130 N.J. at 137, 613 A.2d 1059 (“[T]he purposes to be achieved by proportionality review require that the universe include clearly death eligible homicides in which the prosecutor elected not to seek the death penalty.”).
■ In addition to frustrating the purposes of proportionality review, the amendment renders the very methodology of proportionality review unworkable. Our proportionality review thus far has depended on comparisons to non-death-sentenced cases. Logic would dictate that it would be pointless to conduct proportionality review otherwise. Indeed, “[a] death sentence is comparatively excessive if other defendants with similar characteristics generally receive sentences other than death for committing factually similar offenses in the same jurisdiction.” Id. at 153-54, 613 A.2d 1059 (emphasis added) (citation omitted). Measuring societal consensus and determining whether defendants with similar characteristics generally receive life sentences will be impossible without comparison to life-sentenced cases.29 Therefore, it cannot sensibly *445be assumed that the Court’s probable modifications of proportionality review methodology will approach or approximate the restricted universe prescribed by the 1992 amendment.
Ill
The Court, in my view, mishandles the three fundamentally important issues before it. It finds this defendant’s death sentence to be proportionate and, therefore, that the death sentence may be carried out. The defendant’s sentence under current standards for determining death-sentence proportionality cannot be squared with the sentences generally received by similar defendants who committed similar crimes. Those current standards are, I submit, not capable of identifying a proportionate death sentence and, indeed, are unconstitutional. The 1992 amendment of the Death Penalty Act that drastically restricts the basis for comparing capital defendants destroys meaningful proportionality review and cannot meet constitutional standards. It should be declared to be unconstitutional. That determination cannot be influenced by any recommendations to change the standards or methodology of proportionality review as may be proposed by the Special Master, and, consequently, such a determination must not be delayed.
Most importantly, the accumulated evidence before the Court, understood in the light of a sad history and a tragic legacy of invidious and vicious racial discrimination in the imposition of the death penalty, further illuminated by common and recurrent experience, demonstrates the presence of racial discrimination in the *446prosecution of capital murder offenses. That evidence is “significant under the New Jersey Constitution.” Bey IV, supra, 137 N.J. at 381, 645 A.2d 685 (citing Marshall II, supra, 130 N.J. at 210, 613 A.2d 1059). An unconstitutional risk of racial discrimination in the imposition of the death sentence has, in my opinion; been laid bare and exposed. If the evidence of racial discrimination is perceived to be less damning than I believe it to be, the Court should, at a minimum, pause in permitting any executions to go forward unless and until the State definitively demonstrates that racial discrimination plays no role in New Jersey’s administration of the death penalty.
I, therefore, dissent.
Appendix A-l
AOC Culpability Rankings
Percent of Cases Capitally Tried According to Victims’ Race by Culpability Levels1
Culpability Level Black Victim Nonblack Victim Disparity
1 (range .00 to .19) 21% (25/119) 40% (72/178) 19
2 (range .20 to .39) 55% (6/11) 68% (15/22) 13
3 (range .40 to .59) 75% (3/4) 83% (5/6) 8
4 (range .60 to .79) 100% (2/2) 100% (2/2) 0
5 (range .80 to 1.0) 80% (4/5) 100% (8/8) 20
Appendix A-2
AOC Culpability Rankings
Percent of Defendants Sentenced to Die According to Race by Culpability
*447Appendix A-2 — Continued
Levels with Equal Range 2
Culpability Level Black Def. Nonblack Def. Disparity
1 (range .00 to .19) 13% (5/40) 0% (0/30) . 13
2 (range .20 to .39) 40% (4/10) 13% (2/16) 28
3 (range .40 to .59) ■ 57% (4/7) 67% (4/6) -10
4 (range .60 to .79) 100% (4/4) 67% (6/9) ' 33
5 (range .80 to 1.0) 92% (11/12) 77% (10/13) 15
Appendix A-3
AOC Culpability Rankings
Percent of Defendants Sentenced to Die According to Race by Culpability Levels with Equal Numbers of Defendants 3
Culpability Level Black Def. Nonblack Def. Disparity
1 (range .00 to .02) 0% (0/14) 0% (0/15) 0
2 (range .02 to .14) 5% (1/19) 0% (0/11) 5
3 (range .14 to .33) 43% (6/14) 7% (1/15) 36
4 (range .35 to .73). 69% (9/13) 53% (9/17) 16
5 (range .74 to 1.0) 92% (12/13) 75% (12/16) 17
Appendix B-l
Culpability Levels According to Race Based on Judges’ Rankings 4
Culpability Level Black Def. Nonblack Def. Disparity
1 (range 1.00 to 1.40) 0% (0/3) 33% (1/3) -33
2 (range 1.50 to 2.40) 17% (2/12) 7% (1/14) 10
3 (range 2.50 to 3.40) 33% (8/24) 26% (5/19) 7
4 (range 3.50 to 4.40) 52% (15/29) 37% (10/27) 15
5 (range 4.50 to 5.00) 60% (3/5) 45% (5/11) 15
*448Appendix B-2
Culpability Levels According to Race With Equal Ranges
Culpability Level5 Black Def. Nonblack Def. Disparity
1 (range 1.00 to 1.79) 20% (1/5) 33% (1/3) -13
2 (range 1.80 to 2.59) 9% (1/11) 13% (2/15) -4
3 (range 2.60 to 3.39) 33% (7/21) 24% (4/17) 9
4 (range 3.40 to 4.19) 41% (9/22) 33% (8/24) 8
5 (range 4.20 to 5.00) 71% (10/14) 47% (7/15) 24
Appendix B-3
Culpability Levels According to Race With Equal Numbers of Defendants
Culpability Level Black Def. Nonblack Def. Disparity
1 (range 1.00 to 2.40) 13% (2/15) 12% (2/17) -1
2 (range 2.50 to 3.17) 38% (6/16) 42% (5/12) -4
3 (range 3.20 to 3.75) 36% (4/11) 27% (4/15) 9
4 (range 3.80 to 4.17) 35% (6/17) 27% (4/15) 8
5 (range 4.20 to 5.00) 71% (10/14) 47% (7/15) 24
Appendix B-4
Culpability Levels According to Race With Averages Without Highest and Lowest Score 6
Culpability Level Black Def. Nonblack Def. Disparity
1 (range 1.00 to 1.79) 17% (1/6) 20% (1/5) -3
2 (range 1.80 to 2.59) 17% (2/12) 14% (2/14) 3
3 (range 2.60 to 3.39) 30% (6/20) 25% (4/16) 5
4 (range 3.40 to 4.19) 44% (8/18) 27% (6/22) 17
5 (range 4.20 to 5.00) 65% (11/17) 53% (9/17) 12
Appendix C
Comparison of AOC and Judges’ Culpability Rankings
AOC's Culpability Rankings7
*449Level Death-Sentencing Rate
1 7%
2 21%
3 54%
4 53%
5 81%
Judges’ Culpability Rankings8
Level Death-Sentencing Rate
1 17%
2 12%
3 30%
4 45%
5 50%
Appendix D
Comparison Case Summaries: Multiple Murderers not Classified in Comparison Case Category
B:
A. John Lee Allen (life sentence)
Defendant, along with his twin brother and another man, conspired to rob cab drivers. Armed with pistols, they held up a cab driver and forced him to drive to the back of a bar, where they robbed him of $100 and shot him in the head. That same day, the three confederates robbed another cab driver of six dollars, and shot him to death, this time because they feared the driver “was going for something.” Neither victim was armed.
*450Allen’s brother confessed to a wired informant that the first cab driver pleaded for his life, he “was saying ‘don’t hurt me man, I got a wife and kids.’” The codefendant recalled that the cab driver “was so scared he almost shit himself.”
At the time of these offenses, defendant was facing prosecutions for murder, attempted murder, robbery, and aggravated assault on a police officer. He also had two prior convictions for receiving stolen property. He had no known mental or physical problems.
Defendant was tried for non-capital murder and convicted. The AOC coded the c(4)(f) (escape detection), c(4)(g) (felony murder), and c(5)(h) (catch-all) factors as being present.
B. Joseph Harris (2) (1 death sentence; 3 life sentences)
Defendant, a postal worker, went on a rampage. He first went to his supervisor’s house, where, after picking the lock and turning off the phone, he shot and killed the supervisor’s sleeping husband with a silenced handgun. Harris then snuck upstairs and woke up the supervisor. The victim begged him not to shoot her, but to no avail. However, the gun jammed. So, instead, Harris drew his twenty-five inch ninja sword and chased her as she ran through the house. He stabbed her more than ten times, killing her, and leaving her with defensive wounds on her hands. Adding a gas mask and a bullet-proof vest to his ninja outfit, Harris headed to the post office. He shot two co-workers at close range with an Uzi machine pistol, lit and threw something at the police when they arrived, and, after the SWAT team raided the building, surrendered. Defendant also booby-trapped his house, but the bomb squad disarmed the device without injury.
Upon arrest, defendant confessed, asserting that he had been the victim of racial discrimination while employed at the post office and that he had killed his supervisor in revenge for being fired. Harris admitted to the police that he thought about raping his supervisor, and that his plan was to take hostages, then *451murder certain people as they arrived for their shifts. He named five individuals that “deserved to die.”
Defendant had a history of psychiatric problems that were never treated. Based on the ninja outfit, the police tied him to the murder of a businessman after the rapes and brutal assaults of his wife and two young daughters.
Defendant was capitally tried and convicted. The penalty-phase jurors found aggravating factors c(4)(a) (prior murder), c(4)(f) (escape detection), c(4)(g) (felony murder), and mitigating factor c(5)(h), present. Defendant was sentenced to death for the murder of the first victim, and to life imprisonment for the other three homicides.
C. Kim Kuchler (life sentence)
Defendant murdered her two sisters and one sister’s husband following a series of disputes between the parties. In 1985, during one of the disputes, the police were called. They found ten plastic garbage bags emitting foul odors. The bags contained bones and flesh, but the officers, believing the contents to be cow bones, placed the bags at the curb for garbage pickup.
Seven years later, defendant confessed that she had hired three gang members to kill her sister and the sister’s husband. Defendant also admitted that she had continued to use the victims’ money by forging checks. Eventually, defendant admitted that she shot the victims herself and cut up their bodies.
Defendant, thirty-nine years old at the time, was a resident alien from Korea who had been raised in a poor and dysfunctional family. She had no prior convictions, aside from a shoplifting charge and the pending murder charge of her other sister. She had no known history of child abuse or mental illness. She pleaded guilty to two counts of non-capital murder (in exchange for dismissal of all other charges relating to the three murders). The AOC coded the c(4)(c) (vile), c(5)(a) (mental disturbance), *452e(5)(b) (provocation), c(5)(f) (no criminal history), and' c(5)(h) (catch-all) factors as present.
D. Frank Marsini (1, 2 and 3) (life sentence)
Defendant murdered four people in the course of a year. First, he stabbed his eighty-five-year-old aunt to death in her home, and sexually assaulted her. Almost a year later, defendant killed an eighty-three-year-old male and his seventy-eight-year-old wife, for whom he had worked as a carpenter. As in the prior case, the victims were found nude from the waist down. All four victims had defensive wounds on their hands.
Defendant, forty-seven years old at the time of the murders, claimed to have suffered from “detachments of reality.” He had an eighth-grade education and was self-employed as a carpenter.
Defendant pleaded guilty to all four murders, for which he received consecutive and concurrent life sentences. The AOC found the c(4)(g) (felony murder), c(5)(d) (mental disease), and c(5)(h) (catch-all) factors present in all murders.
E. Ronald Mazique (life sentence)
Defendant sought to rob the victim of her tax refund. As defendant approached, the victim, forty-one years old, and her six-year-old grandson began to scream and yell. Defendant struck both victims at least thirty times with a hammer. He then turned on the gas stove, intending to blow up the apartment in order to cover up what he had done.
Defendant had previously committed a double homicide in South Carolina for which he had confessed; he had robbed and assaulted a cab driver as well.
At the time of this offense, defendant, was twenty-one years old, had dropped out of high school, and was unemployed. He admitted to using alcohol and drugs.
The State charged defendant with capital murder. The penalty jury found the c(4)(c) (vile), c(4)(f) (escape detection), and e(4)(g) *453(felony murder) aggravating factors as well as the e(5)(h) (catchall) mitigating factor, but not the c(5)(c) (defendant’s age) or c(5)(d) (intoxication) factors. The jurors could not agree on punishment. Consequently, defendant was sentenced to life.
F. Hector Sanabria (2) (life sentence)
In an attempt to gain a monopoly over drugs sales in Paterson, defendant shot and killed a twenty-eight-year-old and a twenty-nine-year-old man. One of the victim’s reached for his own gun before defendant shot him in the chest, liver, pancreas, and stomach.
Defendant, twenty-three years old, was the father of three children. He left school in ninth grade and had been unemployed. Defendant was in good mental and physical health and denied drug abuse. He was, however, an alcoholic.
Prior to these offenses, defendant had another murder conviction. He was also charged with an additional murder prior to the resolution of the instant offense.
Defendant was charged and tried for non-capital murder. He was convicted and sentenced to life. The AOC coded the e(4)(a) (prior murder), c(4)(b) (creating grave risk of death to another person), and c(5)(h) (catch-all) factors as being present.
For affirmance — Chief Justice PORITZ, and Justices POLLOCK, O’HERN, GARIBALDI, STEIN and COLEMAN — 6.
For reversal — Justice HANDLER — 1.
IN THE MATTER OF THE PROPORTIONALITY REVIEW PROJECT.
ORDER
The Supreme Court having completed its proportionality review of the death sentence of defendant Donald Loftin, State v. Loftin, 157 N.J. 253, 724 A.2d 129 (1999);
*454And the Court having determined that the current proportionality review methodologies are in need of careful review and reconsideration;
And the Court having further determined to appoint a Special Master for the conduct of supplemental proceedings to consider the proportionality review methodologies used by the Court;
And the Court having further determined that the Special Master should develop a record and examine a broad range of factual issues before the Court can rule on the constitutionality of an amendment to N.J.S.A 2C:ll-3e that limits the Court’s proportionality review to a specific group of similar cases in which a jury has sentenced the defendant to death, and, specifically, whether the statutory limitation unduly restricts appellate review by this Court;
And good cause appearing;
It is ORDERED that Superior Court Judge David S. Baime, now serving as a Presiding Judge in the Appellate Division is appointed as a Special Master for the Supreme Court; and that he shall conduct a review, perform analyses, and make findings and recommendations relating to the proportionality review methodology used by the Court since Marshall II, that shall consider and address: the size of the universe of comparison cases; particular issues in respect of individual proportionality review; questions relating to the statistical models used in both individual and systemic proportionality review; and the status of proportionality review as a separate proceeding in death penalty appeals; and it is further
ORDERED that such review and analyses shall encompass the following areas:
(1) The Special Master shall conduct additional fact-finding concerning the proper scope of the proportionality review universe. The Special Master shall make an independent evaluation of the deathworthiness of a sample of cases previously classified by the Administrative Office of the Courts (AOC) as either death-*455eligible or death-ineligible. The “provability” of the selected cases and the presence or absence of aggravating and mitigating factors shall be considered and the results compared to the data-coding decisions made by the AOC. If there is a variance between the survey results and the AOC data-coding decisions, possible causes of the variance shall be identified along with recommendations for improved data-coding procedures. The Special Master shall consider whether a questionnaire should be filled out by the judge in each case and used to improve both the data-collection and data-coding process. Alternatively, if the Special Master determines that the intrinsic difficulties and ambiguities of data-coding death-eligible cases cannot be overcome, the Special Master shall consider the impact of anticipated coding errors on the AOC models;
(2) The Special Master shall review data-coding generally and make recommendations for improvements if appropriate;
(3) The Special Master shall attempt to determine, based on projections about the size of the database over time and other relevant considerations, how long it will take before frequency review results can attain a level of statistical reliability;
(4) The Special Master shall undertake a review of both the strengths and weaknesses of the index-of-outcomes test and make recommendations whether the statistical models can be modified and improved or whether the index-of-outeomes test should be eliminated;
(5) The Special Master shall consider methods by which to select a representative number of eases within the group of similar cases for consideration and comparison to the defendant’s case in the salient-factors test and precedent-seeking review. The Special Master shall examine alternate ease sorting approaches that account for mitigating factors. The Special Master shall assess whether some reduction in the number of case classifications is possible without compromising the principle that only similar cases be compared;
*456(6) The Special Master shall attempt to develop parsimonious statistical models for more reliable regression studies of race effect and shall consider whether the process of purging, ie., the removal of the indirect effects of race from variables that appear to be unrelated to race, produces results that are useful;
(7) The Special Master shall consider Special Master Cohen’s recommendation, submitted in State v. Loftin, supra, that the Court appoint a panel of judges to perform periodic assessments of penalty-trial outcomes, along with the composition and mandate of such an independent judicial panel, as independent verification of the culpability ratings derived from the models;
(8) The Special Master shall develop a factual record and issue findings concerning the desirability of maintaining proportionality review as a separate proceeding or, alternately, conducting proportionality review in connection with a capital defendant’s direct appeal; and it is further
ORDERED that to assist the Special Master in this project, the AOC shall forthwith provide the Special Master, the Public Defender, and the Attorney General with: (1) a list of all penalty-phase cases, broken down by the race of the defendants, the sentences, and the aggravating and mitigating factors found; (2) a brief synopsis of all penalty-phase cases; and (3) copies of the reports on all prior proportionality reviews including, among others, the reports submitted to the Court by Professor Baldus and Special Master Cohen; and it is further
ORDERED that the Special Master shall conduct the review directed by this Order using all available data and reports, together with the assistance of the AOC and a consultant to be appointed by the Special Master with the approval of the Court; and it is further
ORDERED that in respect of the relevant issues, the Special Master may invite participation of interested parties not otherwise participating in the within matter; and it is further
*457ORDERED that the Special Master shall have the authority to conduct hearings, procure technical and judicial expert advice, call witnesses, and direct the AOC and the selected consultant to perform such analyses and provide such advice as the Special Master deems necessary and appropriate to comply with the terms of this Order; and it is further
ORDERED that the Special Master shall promptly undertake the review required by this Order and shall file a report consisting of the Master’s findings and recommendations, as well as the underlying evidence, data, and analyses by May 14,1999; and it is further
ORDERED that the within Order may be modified or supplemented by the Court on the application of the Special Master or on the Court’s own motion; and it is further
ORDERED that nothing in this Order should be construed by the Special Master or the parties to represent a position of the Supreme Court on any issue before it.
Schedules 2 and 5 use the penalty-trial universe, which consists of all death-eligible cases that proceeded to a penalty phase. They include variables for all the statutory aggravating and mitigating factors, the defendant’s and the victim’s socio-economic status, the defendant’s sex, and the defendant's and victim’s race. Schedule 5 also includes nonstatutory aggravating factors, such as the source of the victim’s suffering or whether the defendant mutilated the victim during the murder.
Schedules 2 and 5 are the two regressions that measure jury sentencing decisions only. Schedule 8, the third regression, comprises the same variables as Schedule 2. But, Schedule 8 uses the death-eligible universe, which consists of capital cases that proceeded to a penalty phase as well as death-eligible cases for which prosecutors chose not to seek a death sentence.
The regressors use a coefficient that represents the strength of the variable. A coefficient greater than zero illustrates the positive correlation between the variable, whether the victim is white or the defendant is black, and the outcome, whether a death sentence is imposed. The higher the coefficient, the stronger the white-victim or black-defendant effect.
3 The observed significance, which is also called a p-value, measures the extent of the variable’s statistical significance. A positive correlation may be the product of mathematical randomness rather than actual cause and effect. However, the lower the observed significance, the lower the likelihood that the correlation is the product of randomness. In social science, the standard level of statistical significance is 0.05, at which point the probability that the correlation is random is five percent. When the p-value is at or below 0.05, the effect is considered to be statistically significant. For example, in Schedule 5, the regression that takes nonstatutory aggravating factors into account, the p-value is 0.0066, i.e., the probability that the black-defendant effect is random is 66 in 10,000.
These data were taken from the Special Master’s Table A-2 (a revised version of Table A from the Special Master Report).
The culpability level for these cases was estimated from Model 12 in the Loftin Report using 215 non-penalty and 147 penalty-trial cases.
Nonblack defendants have an average culpability rating of 3.38 and black defendants have an average culpability rating of 3.30. The Special Master found these judges' culpability ratings to be more reliable than those of the AOC. Id. at 23.
As noted, App. A is based on AOC Table 18.1 (App.24-27). Although the odds ratio is flawed, that ratio calculated by the AOC ranges from 4.3 to 10.8, Loftin Report, tech. app. 10, schs. 2 & 5 (logistics), suggesting that the odds of a black defendant receiving a death sentence versus a nonblack defendant are 10.8 to 1 (or 4.3 to 1).
The AOC found, for example, in Schedule 2 (logistics), that the race variable ("blackd”) was statistically more significant than all aggravating factors except c(4)(a) (prior murder), c(4)(c) (torture/depravity), and c(4)(d) (pecuniary gain), and than all statutory mitigating factors except c(5)(d) (capacity to appreciate wrongfulness). In Schedule 5 (logistics), the AOC found the race variable to be more significant than every other aggravating factor. See Loftin Report, tech, app. 10.
The third model is the only one (according the AOC) that, in the process of paring down the total numbers of variables, combines strong variables together in one group and weak variables together in another. Dr. Tukey gives no explanation for his decisions regarding pooling of variables; but his first two models, which group factors of varying strengths together in one variable, do not appear to be consistent with fundamental methodological guidelines.
Dr. Tukey’s own statistical print-outs seem to confirm the common sense conclusion that Model 3 is stronger than the other two. One method of testing reliability is to look at the percentage of concordant and discordant pairs within a model. The higher the concordant percentage, the better the model. Model 3 has the highest concordant percentage and the lowest discordant percentage of any of the three models, indicating that it is capable of accurately predicting the highest percentage of outcomes (over 90%, in this case). Tukey Report, supra, at 26.
The Public Defender also created a three-variable model, the most parsimonious of any of the models evaluated herein. Using the judges' culpability rankings, which the Special Master found to be more reliable than those of the AOC, see Special Master Report, supra, at 25, and the black-defendant and white-victim variables, this mo’del also indicates a strong racial effect: the p-value for the black defendant variable is 0.13. Appendix to Defendant’s Supplemental Brief, supra, at 4. While not statistically significant, this racial effect is strong enough to cause concern.
The Special Master suggests the following coding changes to the universe: add life-sentenced defendants and reversed death-sentenced cases; count retrials of reversed cases separately within the universe; exclude fifteen cases where no aggravating factors were found; count multiple-victim cases as single cases; and consider black-Hispanics as blacks and all other Hispanics as nonblacks. Special Master Report, supra, at 30-42.
The new coding leads to some major differences: (Í) a thirty-six percent racial disparity for death-sentenced defendants as a percentage of death-eligible cases, as compared to a twenty-one percent disparity before recoding; and (2) an eighteen percent racial disparity for death-sentenced defendants as a percentage of death-eligible cases as compared to a fifteen percent disparity before recoding. Letter from Richards. Cohen (Feb. 7, 1997).
The Special Master’s decision to average the judges’ culpability ratings is suspect. If nonblack defendants tended towards the extreme culpability rankings, while black defendants tended towards the mid-range, the Special Master would find the averages to be similar. For example, if there were nine nonblack defendants with culpability levels of 1, 1, 1, 3, 3, 3, 5, 5 and 5, and only the three most culpable defendants were sentenced to death, the average culpability level rating for the nonblacks would be level 3 and the death sentencing rate would be thirty-three percent (3/9). Assume, also, that there were nine black defendants, all in culpability level 3, three of whom were sentenced to death. They too would have an average culpability rating of level 3 and a death-sentencing rate *394of thirty-three percent (3/9). Despite the seeming equality between the races, there is great disparity when the cases are examined by culpability level. For equally culpable black and nonblack defendants in level 3, blacks were sentenced to death thirty-three percent of the time and nonblacks were never sentenced to death. This example, though overly simplified, seems to reflect the reality of New Jersey's capital sentencing scheme. The Special Master's reluctance to dispose of any outliers before averaging the ratings was, as the majority points out, due to a legitimate concern that the number of judges were so few that disposing of any of their results would lend even less credibility to the findings. See ante at 309, 724 A.2d at 157 (quoting Supplemental Report, supra, at 8). However, the legitimacy of this concern does not render the averages reliable, it merely explains that averages computed without outliers may be just as flawed as those computed with outliers. The Court also justifies the Special Master’s methodology by quoting his contention that “there is no reason to believe that the outliers leaned more in one direction than the other, and therefore no indication that the course we took had any effect on the accuracy of the results.” Ibid. The Special Master presents no evidence, however, that the outliers do not lean more in one direction than the other (in fact, he does not even contend that he examined them for the purpose of uncovering such trends).
The Special Master’s averaging of culpability ratings also fails to pick up the huge range in ratings exhibited by the judges. For example, the judges collectively gave Booker, a defendant whom the AOC considers to be one of the most culpable in the universe of death-eligible cases, a low average culpability rating of 2.75; the four judges, however, individually gave the defendant ratings of 1, 3, 3, and 4. Again, by way of example, Harvey (1A) and Harvey (IB) are two cases that have nearly identical summaries. Nevertheless, the culpability ratings (each made by a different set of judges) are quite different. In Rise, even though the summaries for the two trials are almost identical, the case with the greater list of mitigating evidence was ranked by the judges much higher in terms of culpability. These unexpected results and the Special Master’s averaging of the scores make it impossible to uncover underlying inconsistencies and attach meaning to them. While we do not know for sure if an examination of the judges' findings without the use of averages would affect the results significantly, the Special Master’s approach is certainly problematic.
The Special Master, while soundly excluding the actual sentences from the summaries, made a value judgment to include a list of what mitigating and aggravating circumstances were found by the penalty-phase juries (as opposed to merely listing which factors were alleged by the State and the defendant). If racial discrimination infects the ultimate verdict of life or death, the same prejudice would presumably also infect the finding of mitigating and aggravating circumstances. For example, in one case judges were told that the c(4)(h) (murder of a public servant) factor was not present because the penalty jury sentenced the defendant to life, refusing to find the c(4)(h) aggravating factor, despite unequivocal evidence that the victim was a uniformed police officer operating a marked police cruiser with lights flashing while engaged in the performance of his official duties.
Some statisticians argue that the use of “statistical significance” in the context of models involving “apparent populations," which are the basis for proportionality review of capital prosecution sentencing, (in contrast to a sample randomly drawn from a true population), is erroneous. Statistical significance is a value used to measure sampling error. However, “[f]or apparent populations, there is no sampling and, therefore, no clue about whether the data were generated to some sampling plan. To proceed as if the data were generated by random sampling or random assignment is to embrace a fiction.” Richard A. Berk, Bruce Western & Robert E. Weiss, Statistical Inference About Apparent Populations, 25 Soc. Methodology 421, 430 (1995). Professor Berk suggests alternative approaches, but notes that none of these methods is wholly satisfactory. Id. at 452-53. The Special Master should explore the various deficiencies in each approach (including our use of statistical significance) in his examination of our methodologies for proportionality review.
I agree with the Court that the "reasons for choosing the larger universe remain valid today.” Ante at 289, 724 A.2d at 147. Nevertheless, I continue to reserve a limited objection to the chosen universe, namely, that the universe is defective insofar as it includes cases in which a defendant’s death sentence was reversed on appeal. See DiFrisco III, supra, 142 N.J. at 232-33, 662 A.2d 442 (Handler, J., dissenting); Martini II, supra, 139 N.J. at 82-90, 651 A.2d 949 (Handler, J., dissenting); Bey IV, supra, 137 N.J. at 402-07, 645 A.2d 685 (Handler, J., dissenting); Marshall II, supra, 130 N.J. at 253-57, 613 A.2d 1059 (Handler, J., dissenting).
As a prior murderer with three aggravating factors, Loftin is placed in AOC category B, for prior murderers, and subcategory B(l), which identifies prior murderers with two or more aggravating factors. There is no intrinsic reason why Loftin’s culpability should not, in addition, be compared with that of *417multiple-victim murder defendants, classified by the AOC as A(l). See infra at 325-327, 724 A.2d at 165-166.
The majority contends that its standard for measuring disproportionality, namely, that a death sentence is disproportionate if similarly situated defendants generally receive sentences other than death, cannot be equated, or used interchangeably, with the concept that a death sentence is only proportionate if death sentences are generally received by similarly situated defendants. See ante at 322, 724 A.2d at 163. That would mean, according to the Court, that we must "find that death is [a] normal sentence when [it] can never be so." Ibid. The Court misperceives the problem and misstates the answer.
There is nothing complicated or exceptional about the proposition that a class of defendants who do generally receive sentences other than death is a class of defendants who do not generally receive death sentences; this follows from common notions of generality. As such, it is accurate, not disingenuous, to assert that a death sentence is disproportional if similarly situated defendants do not generally receive death sentences.
More fundamentally, the Court must recognize the import of its statement: "[b]ecause New Jersey jurors have been sparing in their imposition of the death sentence, it will never be the case that death would be ‘generally received’ or ‘received in a defined preponderance of cases.’ ” Ibid. If, as the Court claims, death sentences will never be generally received — even by similarly situated defendants — in this State, then all death sentences imposed in New Jersey are, and will be, disproportionate. If death must be "a normal sentence” and, in fact "it can never be so,” then we must accept the verdict that death sentencing cannot be constitutionally imposed.
Throughout this discussion, I do not consider statistics that include defendant’s case among the class of comparison cases. Placing defendant’s case on both sides of the comparison, gauging the proportionality of a defendant's sentence by comparing it to a group of which he is a member, skews the analysis. The Court fails to perceive this problem. Of course, defendant’s death sentence has relevance to the database overall, and is properly considered in the proportionality review of subsequent death sentences.
Many of the AOC-defined subcategories include very few cases. Accordingly, this Court frequently appeals to the composite category, such as class B, for comparison. In this case, for example, the category of B(l) defendants, not including Loftin, is comprised of only four defendants at penalty trial, and only six overall.
It must be noted that the huge confidence intervals in the measurement of predicted probability of death detract substantially from the reliability of the test. See ante at 295-296, 331-332, 724 A.2d at 150, 168.
Unfortunately, depending on the regression performed, there is great fluctuation among the levels at which defendants are placed. This plainly detracts from the credibility of the endeavor as well.
Precedent-seeking review uses as a starting point the same categories of comparison, see Loftin Report, tbl. 7; see also ante at 326 n. 15, 724 A.2d at 165 n 15, that are used in the salient-factors test. See, e.g., Bey IV, supra, 137 N.J. at 367, 645 A.2d 685.
Eight prior murderers classified as B(2) or B(3) received death sentences. These defendants are: Marko Bey (2B), Richard Biegenwald (1A and IB), Brian Coyle (1A), Samuel Erazo (1A), Frank Pennington (1A), Braynard Purnell (1A), and Thomas Ramseur. Eleven B(2) and B(3) defendants received life sentences: Biegenwald (1C and 2), Coyle (IB), Erazo (IB), William Godette, Jihad Mu-hammed, Alerto Nieves, Pennington (IB), Purnell (IB), Carlos Vasquez, and Thomas Williams.
I have previously noted that proportionality review inherently involves subjective assessments of moral culpability and that in undertaking such reviews the Court indeed goes beyond simply determining the proportionality of a death sentence:
The fallacy of proportionality review inheres in its attempt to identify precisely what makes some cases similar and others dissimilar. It inheres in the notion that we, as judges, can identify the qualities that make some murderers deathworthy and others not. The Court's efforts to do that — by assessing moral culpability — amply demonstrate that proportionality review often serves to perpetuate rather than eliminate arbitrary sentencing. Moral culpability determinations may appear to be a more sophisticated and refined exercise than the statutory process by which jurors determine death eligibility and deathworthiness. As I have often said, the latter process does *439not sufficiently restrict the discretion of jurors or direct and guide them to sound and reliable decisions concerning capital murder and life-and-death judgments. The objective of proportionality review is to enable the reviewing court to compensate for such subjectivity and correct its unfair results, as when similar defendants committing similar crimes do not in fact receive the same sentence. The irony of proportionality review is that the reviewing court itself resorts to extra-statutory factors that are also inherently subjective. Because the Court at this end-point of capital prosecution validates or invalidates the defendant's death sentence based on its own judgment of culpability, the Court, make no mistake about it, becomes the ultimate sentencer.
I maintain:
Although I believe proportionality review is required constitutionally, I do not believe that its use can save an otherwise unconstitutional death-penalty statute from constitutional infirmity. Proportionality review has its limits. It cannot keep its own promise — it cannot assure evenhandedness, uniformity, or consistency in the imposition of the death sentence. It does not and cannot overcome the arbitrariness that plagues the imposition of the death *441penalty. Our experience with proportionality review itself now confirms that chronic and profound constitutional deficiencies endure in the New Jersey system for administering capital punishment.
[Marshall II, supra, 130 N.J. at 273, 613 A.2d 1059 (Handler, J., dissenting).]
In the unlikely event that this Court were to apply the 1992 amendment to Loftin, it would have to address the possibility that the amended law as applied to Loftin would violate State and Federal Ex Post Facto Clauses. U.S. Const., art. I, § 21; N.J. Const., art. IV, § 7, ¶ 3. A statute is in violation of the Ex Post Facto Clauses if it serves to
(1) punish as a crime an act previously committed, which was innocent when done; (2) make more burdensome the punishment for a crime, after its commission; or (3) deprive a defendant of any defense available according to the law at the time when the crime was committed.
[State v. Muhammad, 145 N.J. 23, 56, 678 A.2d 164 (1996) (citing Beazell v. Ohio, 269 U.S. 167, 169-70, 46 S.Ct. 68, 68-69, 70 L.Ed. 216, 217 (1925)).]
This Court has recognized that "laws that affect the length of a prisoner's term, or are likely to affect it, passed after commission of the crime, obviously constitute 'additional punishment’" and are ex post facto laws as applied to defendants whose crimes were committed prior to the laws' enactment. Doe v. Poritz, 142 N.J. 1, 45, 662 A.2d 367 (1995). The proportionality review provisions of the 1992 amendment, which are retrospective as applied to Loftin because his offense was committed one week before the effective date of the amendment, are likely to affect the term of his sentence. The 1992 amendment has the potential to render an once-disproportionate sentence proportionate, and to convert defendants who are effectively not death-eligible into death-eligible defendants. In contrast to laws that merely alter the scope of evidence admissible at trial, see Muhammad, supra, 145 N.J. at 57, 678 A.2d 164 (1996) (holding statute that "simply modified the scope of evidence that may be admitted during the penalty phase of a capital case and did not alter any substantive rights of defendant ... would not violate the State or Federal Ex Post Facto Clauses" when applied to defendant who committed offense prior to enactment of statute); State v. Erazo, 126 N.J. 112, 135, 594 A.2d 232 (1991) (holding new law *443permitting the State to introduce evidence regarding statutory prior murder aggravating factor did not violate Ex Post Facto Clauses when applied to defendant who committed offense prior to enactment of statute), the 1992 amendment appears to be precisely the type of law that the ex post facto ban is designed to avoid.
The Legislature had previously amended the proportionality review portion of the death-penalty statute to make proportionality review available only when requested by the defendant. NJ.S.A. 2C:ll-3e, L. 1985, c. 478, § 2 (eff. June 10, 1985). The Court has never ruled on the constitutionality of the 1985 amendment because all six capital defendants eligible for proportionality review have requested it. The Court, however, has recognized the change, see State v. Bey, 129 N.J. 557, 625, 610 A.2d 814 (1992) (Bey III), cert. denied, 513 U.S. 1164, 115 S.Ct. 1131, 130 L. Ed.2d 1093 (1995); State v. Marshall, 123 N.J. 1, 170, 586 A.2d 85 (1991) (Marshall I), cert. denied, 507 U.S. 929, 113 S.Ct. 1306, 122 L. Ed.2d 694 (1993), and expressed some misgivings, see State v. Ramseur, 106 N.J. 123, 193, 524 A.2d 188 (1987) ("We share many of the dissent’s concerns with respect to ... the importance of proportionality review even in the absence of a request by the defendant."). In other cases, the Court has indicated, by implication at least, that providing proportionality review on a request-only basis may be unconstitutional. See State v. Martini, 144 N.J. 603, 614, 677 A.2d 1106 (1996) (Martini III) (mandating post-conviction review in death penalty cases), cert. denied, 519 U.S. 1063, 117 S.Ct. 699, 136 L.Ed.2d 621 (1997); State v. Hightower, 120 N.J. 378, 415, 577 A.2d 99 (1990) (Hightower I) (mandating presentation of mitigating evidence in capital sentencing proceedings); State v. *444Koedatich, 98 N.J. 553, 489 A.2d 659 (1984) (Order) (mandating direct appeal in capital cases); see also Martini III, supra, 144 N.J. at 625, 677 A.2d 1106 (Coleman, J., dissenting) (agreeing with majority "that the death penalty should not be carried out until defendant’s guilt and the lack of disproportionality have been reliably established") (emphasis added).
A 1988 study that detailed the performance of different state courts in proportionality review found only one death sentence reversed in proportionality *445review by a state court that restricted its universe to death-sentenced cases. Steven M. Sprenger, A Critical Evaluation of State Supreme Court Proportionality Review in Death Sentenced Cases, 73 Iowa L.Rev. 719, 738 (1988). In comparison, 32 cases were identified in which courts that included life sentences in their proportionality review universe had vacated death sentences as disproportionate. Ibid.; cf. David Baldus, When Symbols Clash: Reflections on the Future of the Comparative Proportionality Review of Death Sentences, 26 Seton Hall L.Rev. 1582, 1587 (1996) (estimating that, nationwide, "fewer than 75 of 5,000 plus death sentences have been reversed based on excessiveness”).
The data for this chart were provided by the AOC, Table 18A.1, using the Special Master’s universe and the AOC's culpability rankings.
This is AOC's Table 18.1 from Loftin Report (App.24 — 27). Each defendant's culpability level is based on the predicted probability of death computed by a logistic regression employing the penalty-trial universe and comprising statutory aggravating and mitigating factors. The data are incomplete.
This is AOC’s Table 18.2 (App.24-27). Unlike Bey IV, supra, 137 N.J. at 392, 645 A.2d 685, in which culpability level four ranged from .20 to .75, the broadest culpability level in this case ranges from .35 to .73. That is a narrower range, that, unlike culpability level four in Bey IV, groups similar cases.
This is Special Master’s Table D.
The culpability rankings are broken into five levels, each with an equal range of .8.
This is based on Table D.3, of Special Master’s Report, Judges' Culpabilities Percent of Defendants Sentenced to Die by Culpability Levels and Race (w/ equal ranges of culpability).
Loftin Report, Table 21 (logistics); see also Loftin Report, Table 23. Here, the AOC culpability rankings show an increase in the rate of death sentences as the *449culpability ranking increases, as one would expect. The judges' culpability rankings do not show similar correlations between culpability and death sentences.
Special Master's Report, Table D (revised).