dissenting.
In this case, the Court sanctions a death sentence that was dependent on an analysis of blood samples that served to identify defendant as the killer of Irene Schnaps. That evidence, endorsed by the Court today in affirming defendant’s conviction of murder and his sentence of death, was not based on accepted scientific knowledge. Further, the Court deems the evidence to have been reliable even though the hearing conducted by the trial court to determine its admissibility did not focus on its scientific acceptability and reliability. The specific evidence that was admitted to establish defendant’s identity was derived from a novel procedure or methodology of blood analysis known as dot-intensity analysis. Dot-intensity analysis cannot, by any measure, be considered reliable or generally accepted. It is not supported by any authority sufficient to establish its scientific reliability or acceptance within any reputable body or community of scientists.
*234We thus are presented with a capital prosecution that is based primarily on disputed, esoteric, and problematic evidence. It is clear that because that evidence is not reliable, defendant’s convictions and sentence cannot be sustained. The Court’s approval of that evidence as scientifically reliable is a mistake of enormous significance based on profound misunderstandings of the law and facts.
The majority compounds its error by failing to recognize that unfair and improper limitations were placed on defendant’s challenges to the DNA evidence and then by sustaining the use of a problematic “statistical” analysis in conjunction with the evidence to enhance its probative significance. The statistical evidence employs a mathematical calculation that is inexact when used with DNA analysis and, in this case, is based on factually erroneously data. Further, that statistic-based evidence was admitted without a hearing to determine its reliability and through a witness lacking in adjective expertise. Part One of this opinion addresses those issues.
In the course of the trial of this case, other significant errors contributed to defendant’s conviction and death sentence. Among those errors were the trial court’s erroneous jury charge on the Mejia issue, which the Court finds harmless even though it resulted in defendant’s eligibility for the death penalty; the application of the escape-detection aggravating factor, which the Court finds harmless in part and error-free in part even though it clearly contributed to defendant’s death sentence; and the misleading verdict sheet, which the Court also finds harmless although it was clearly capable of dissuading the jury from returning anything but a death-eligible murder verdict. Those issues and others are considered in Part Two.
Part One
The evidence adduced at trial connecting defendant to the murder of Irene Schnaps was purely circumstantial. It consisted of a shoe print found at the crime scene consistent with sneakers recovered from defendant; a single hair with negroid characteris*235tics found underneath the victim consistent with defendant’s hair; an empty watch box found at the scene and a watch of the same type found in defendant’s car; an empty camera box found at the scene and a camera strap found in defendant’s car; and blood evidence. The most damning of the evidence for the defense was clearly the blood evidence.
The blood evidence consisted of both traditional and established enzyme analysis as well as novel and rarely used DNA testing. The traditional blood analysis was hardly determinative because the vast majority of the results were inconclusive or incomprehensible. Only two results were readable. One result showed that defendant could not be excluded as the person who contributed to the blood found at the scene. The other result indicated that the contributor to the blood was African-American, although the enzyme found in the blood could have come from anyone whose lineage derived from racially mixed ancestors.
Because the traditional enzyme analysis was weak and inconclusive, the State extended its blood evidence much further. Like the traditional blood-enzyme analysis, the evidence derived from the extended novel DNA testing showed that defendant could not be excluded as a possible source of the blood found at the scene. Unlike the traditional analysis, however, the State’s extended analysis purported to quantify the results and to produce a statistical figure representing the likelihood that defendant was the actual donor. That statistical construct suggested that only one in 1,400 African Americans could have left such blood at the scene. It was that DNA evidence, as well as embellishing statistics, on which the State relied to secure defendant’s conviction and death sentence.
I
The State’s basic DNA evidence consisted of DQ-Alpha testing and polymarker testing. Those testing procedures are- relatively new, untested, and do not enjoy fidl acceptance. DQ-Alpha testing, however, although not universally accepted, has gained *236approval in numerous jurisdictions, see National Research Council, The Evaluation of Forensic DNA Evidence, 177 & n. 30 (1996) (“NRC Report”), including ours, State v. Dishon, 297 N.J.Super. 254, 687 A.2d 1074 (App.Div.1997); State v. Williams, 252 N.J.Super. 369, 599 A.2d 960 (Law Div.1991). On the other hand, only a few courts have permitted polymarker testing since the testing kit was first marketed in October 1993. See United States v. Beasley, 102 F.3d 1440, 1447 (8th Cir.1996), cert. denied, — U.S. -, 117 S.Ct. 1856, 137 L.Ed.2d 1058 (1997); United States v. Lowe, 954 F.Supp. 401, 417-18 (D.Mass.1996); People v. Pope, 284 Ill.App.3d 695, 220 Ill.Dec. 309, 314, 672 N.E.2d 1321, 1326 (1996), appeal denied, 171 Ill.2d 579, 222 Ill.Dec. 436, 677 N.E.2d 970 (1997); People v. Morales, 227 A.D.2d 648, 643 N.Y.S.2d 217 (N.Y.App.Div.), appeal denied, 89 N.Y.2d 926, 654 N.Y.S.2d 729, 677 N.E.2d 301 (1996). Despite the relatively limited trial record relating to these procedures, the majority endorses both the DQ-Alpha and the polymarker testing as generally accepted and admissible scientific evidence.
The technologies of the DQ-Alpha and polymarker testing have been accepted as a basis for admissible scientific evidence, but only when they have been used as designed. The Court’s description of the first two parts of the DQ-Alpha and polymarker testing — the separation and amplification of the DNA strands — is adequate. See ante at 157-163, 699 A.2d at 615-617. Its description of the third and crucial part of the testing, the analysis of results, however, requires elaboration to understand why the evidence used in this ease — dot-intensity analysis — is wholly unsettled, highly controversial, and of problematic scientific reliability; it is not, as indicated by the majority, simply an obvious and routine extension of the basic DNA testing methodologies.
A.
The process of isolating DNA fragments, amplifying the DNA segments, and analyzing the results using the reverse dot-blot system, is identical for both the DQ-Alpha and polymarker testing *237kits. For both kits, the cell’s nucleus is broken open to release the DNA, and the relevant portions of the DNA are separated out by filtration, chemical cleansing, and the utilization of a centrifuge. The DNA materials are then amplified. During the amplification stage, the DNA strands are broken apart through heating and then allowed to reform with loose DNA matter and primers that are added to the mixture. The materials are then subject to the examination or analysis stage, during which the newly isolated and amplified DNA is flooded over testing strips. The only difference between DQ-Alpha testing and polymarker testing is that the primers used in stage two (the amplification process) and the strips used in stage three (the examination process) are different. Both the DQ-Alpha kit and the polymarker kit were developed and manufactured by the same company.
The DQ-Alpha and polymarker tests were designed so that results are analyzed using “little strips” with “reverse dot/blot technology.” The nylon membrane strips contain dots with allele-specific probes. During the third stage, the amplified DNA product is flooded over the strips. The dots represent the sections of the DNA (“loci”) being examined. The number of loci tested and the number of possible alleles are different in the two tests. DQ-Alpha testing examines one locus, while polymarker testing looks at six loci (five polymarker markers as well as the DQ-Alpha marker). The DQ-Alpha marker has six commonly recognized alleles, while each of the five polymarker markers has two or three possible alleles. On the testing strips, there are dots representing each of the different types of alleles for that particular marker.
The dots represent alleles, not genotypes. A person’s genotype for a particular marker is composed of a pair of alleles, one allele coming from each of the person’s parents. In some instances, DQ-Alpha and polymarker testing can determine which of the different genotypes an individual possesses. DQ-Alpha locus has twenty-one different genotypes, made up of every possible combi*238nation of six possible alleles.1 The polymarker loci, which have only two or three possible alleles each, thus have only three or six possible genotypes.2 If a DNA sample contains only one person’s blood and if the testing kit identifies the presence of two different alleles, then the individual’s genotype can be determined by simply combining the two alleles (everyone has two and only two alleles for each genetic marker; the two alleles, however, can be the same). For example, if the victim’s DNA sample is found to have both the 2 and the 4 DQ-Alpha alleles, then we know the victim’s genotype is a “2,4.” To illustrate further, if the sample reveals only the B HBGG allele, then the suspect’s HBGG genotype is “B,B.”
The State’s experts testified about how the strips used during the analysis stage of polymarker and DQ-Alpha testing work. The strips contain dots representing the different alleles for a particular marker. The dots become more intense (brighter blue) as the enhanced DNA that is flooded over the strips binds or “hooks” to the dots. Dr. Charlotte Word, a scientist at Cellmark Diagnostics, testified for the State that this was an “instamatic process whereby you get a colored development and you will get a blue dot in a place DNA attached.” The methodology for reaching a result is hardly complicated. To reach the result, “you simply read the blue dots off of the strip.” If the dot remains blank or faint, then the allele that the dot represents is considered not present in the DNA sample. Every color development brighter than the control dot indicates the presence of the allele that the dot represents. Perkin Elmer Co., AmpliType PM: PCR Ampli*239fication and Typing Kit Procedures Manual 21 (Dec.1993); Cell-mark Diagnostics, Interpretation of HLA DQa Test Results 1 (Mar. 25, 1992) (“Those dots equivalent to or stronger than the [control] dot are considered positive.”). The testing kits were designed to measure only the presence or absence of certain alleles. That point cannot be overemphasized.
At Cellmark, two scientists read the results of the strips and independently record the results that they observe. The results are then photographed, and the photographs are reviewed by a third person. Dr. Word testified that the strips should be read “as soon as they are developed because that is the appropriate time to read them.” Indeed, the strips should be read immediately, even before they are photographed, because the dots become less intense, quickly fade, and even disappear. The State’s expert emphasized that “the blue dots don’t last very long on the strips.”
The polymarker and DQ-Alpha testing kits, as noted, were designed only to reveal the presence or absence of a particular allele. The State, however, went further than just determining the presence of the various DQ-Alpha and polymarker alleles. Its experts also performed a procedure known as “dot-intensity analysis.” 3 That procedure is untested and unapproved by the scientific community.
Although potentially useful for detecting errors in traditional polymarker and DQ-Alpha testing, dot-intensity analysis was used in this ease to narrow further the class of persons who possibly eould have contributed to the blood found at the crime scene. In effect, dot-intensity analysis, as applied in this case, purported to *240quantify the intensity of those alleles present in a sample. The State hypothesized that by examining the relative intensities of the dots obtained in the DQ-Alpha testing and the polymarker testing on a mixed sample, one could determine the relative presence of certain alleles. Once the relative presence of the alleles was determined, one could “subtract” the alleles from a known contributor to the sample and thereby reveal the unknown contributor’s genotype.
To a certain degree the subtraction principle is a method of elimination and is used in analysis even without dot-intensity analysis. For example, if the test of the HBGG marker revealed the presence of all three possible alleles (A, B, and C) in a mixture containing DNA from two persons and if one of the contributors has a genotype of “A,B,” then we know the other contributor must have a genotype with at least one C allele (i.e., the person could be genotype “A,C,” “B,C” or “C,C”). Dot-intensity analysis, however, goes beyond that relatively straight-forward process. It purports not only to establish the presence of alleles, but also to quantify those alleles that are present.
Dot-intensity analysis is based on a single premise. It assumes that if the testing strip reveals a brighter blue dot for one of the alleles than for the other alleles that the strip indicates are present, then the brighter dot means that that allele is present in greater amounts than the other alleles. Using the same example, if the HBGG marker test strip revealed the presence of all three alleles (A, B, and C), if the A dot is brighter than the B or C dot, and if the first contributor has a genotype of “A,B,” the second contributor is assumed to have the “A,C” genotype. That is because the only genotype that would reveal more A alleles than B or C alleles when combined with the known genotype of “A,B” would be “A,C.” Applying that analysis, the result would not be merely that the unknown contributor has at least one C allele and therefore could have possible genotypes of either “A,C” or “B,C” or “C,C”; rather, that analysis would conclude, because of the *241assumed presence of added or extra A alleles, that the contributor in fact has the “A,C” genotype.
For dot-intensity analysis to work, when alleles are present in the same amounts, the dots on the strips must be of the same intensity. So, in a sample containing a single individual’s DNA, the dot intensities that are obtained in the results should be “relatively balanced.” Dr. Word, testified that “[polymarker testing strips are designed to show an equal dot-intensity when the DNA came from a single individual.” For example, if a single individual’s enhanced DNA is applied to the polymarker test strip containing the allele dots for the GYPA marker and that test strip shows that the individual has the A and the B alleles for the GYPA marker, then the A and B allele dots on the GYPA marker test strip should be of equivalent intensity. The reasoning behind that assumption is that an individual has two alleles for each marker, and that those alleles are equally represented (or, phrased differently, the DNA strand will contain one of each allele — one from each parent — paired on the strand).
The fundamental assumption of dot-intensity analysis is that when it is applied to determine the genotypes of a blood sample and a dot-intensity imbalance occurs, the imbalance is caused by the relative imbalance in the amount of alleles present. The State’s experts, however, offered other explanations for why imbalances may occur. First, there are some variances among different individuals and in the different alleles that may cause slight intensity differences even in a single individual’s sample. Second, in mixtures, “relative probe intensities ... within a locus may become imbalanced in the presence of variant alleles or nonoptimal [test] conditions.” Third, a manufacturing defect or a defect in the procedures may cause differences in dot intensity. Finally, different dot intensities may arise because a particular allele is present in greater or different strength than another allele, which would be true in a mixture. Despite those other explanations, however, dot-intensity analysis assumes that any intensity imbalances are due to the relative presence of the alleles and that the *242relative presence of alleles can be measured by the testing strips even though the strips were designed to disclose only the presence or absence of alleles.
B.
Several samples were tested in this case: a dry blood sample taken from defendant, a sample from a towel found at the crime scene and thought to contain only the victim’s blood, and a sample taken from a boxspring in the victim’s apartment and thought to contain a mixture of the victim’s and the suspect’s blood.4 No blood sample containing only the suspect’s blood was recovered from the scene. Thus, to determine the suspect’s genotypes, the victim’s genotypes had to be compared to the mixed sample, and what was unaccounted for was presumed to belong to the suspect.
The following chart summarizes the results obtained by Cell-mark in the analysis of defendant’s known sample (from the swatch created by police with defendant’s blood), the victim’s known sample (from a towel recovered from the scene), and the unknown mixture (from the boxspring).
Summary of DNA Evidence 5
Marker Mixed Sample Schnaps’ Sample Suspect’s Expected Dot Intensity Susp.’s w/ Dot I. Harvey’s Sample
DQa 2 & 4 2,4 2,2; 2,4; or 4,4 2 = 4 2,4 4,4
LDLR A&B A,B A,A; A,B; or B.B B > A B,B B,B
*243GYPA A & B A,A A,B or B,B A > B A,B A,B
HBGG A, B & C B,B A.C A=B=C n/a A.C
D7S8 A & B A,A A,B or B,B A > B A,B A,B
GC A, B & C A,C A,B; or B,C B,B; A>B C & A,B A,B
This table shows that without using dot-intensity analysis, but considering both the DQ-Alpha and polymarker results, defendant cannot be ruled out as a contributor to the mixed sample if the sample also contains Ms. Schnaps’s DNA (compare “Suspect’s Expected” column with “Harvey’s Sample” column). In some instances, however, the results are not particularly helpful because almost no persons are excluded as contributors. For example, the LDLR marker results show that a person with any of the three possible LDLR genotypes (A,A; A,B; or B,B) could have contributed to the mixed sample. On the other hand, even without dot-intensity analysis, the results from the HBGG marker show that the suspect must have a particular HBGG genotype (A,C).
Using dot-intensity analysis, the State argued that it was able to define much more narrowly the genotypes of the suspect. Analyzing the LDLR, the GYPA, the D7S8, and GC markers and using dot-intensity analysis, the State claimed that only persons with the exact genotypes for those four markers as defendant could have contributed to the sample (compare “Sups.’s w/ Dot I.” column with “Harvey’s Sample” column).
The State did not point to the results of dot-intensity analysis for either the HBGG marker or the DQ-Alpha marker. As for the HBGG marker, the State argued that dot-intensity analysis was unnecessary because the subtraction principle revealed that the suspect must have the A,C genotype. The State did not explain why, when dot-intensity results were analyzed, the results (equal dot intensities for the three alleles) did not comport with the State’s theory. Further, the State did not explain whether or not dot-intensity analysis was possible for the DQ-Alpha marker, but the results of comparing dot intensities (equal intensities for *244the 2 and 4 alleles) would not support the State’s assertion that defendant fit the genetic profile of the suspect.
II
The principal disagreement that I have with the majority concerns the general acceptance of dot-intensity testing. Dot-intensity analysis was the essential evidence relied upon by the State to demonstrate that defendant was in all likelihood the actual person whose blood contributed to the mixed sample found at the scene. The majority properly, if reluctantly, recognizes that dot-intensity testing, as a scientific method, must meet the standard of general acceptance even if DQ-Alpha and polymarker testing are themselves found to be generally accepted scientific tests. The majority, however, misconstrues the distinctive and distinguishing features of dot-intensity testing as a method of analyzing DNA, denigrates many of defendant’s challenges to the testing as not going to the reliability of the procedure, but rather only to its weight, and then, on an embarrassingly deficient record, summarily concludes that the novel scientific procedure passes muster under our long-standing precedent. Dot-intensity analysis as used here — a procedure never before used in any court case, successfully documented in any laboratory, or validated in any scientific study or published literature — has not been shown to be an established and reliable procedure. Further, no foundation for dot-intensity analysis exists in the record, and the results obtained clearly show that such evidence is grossly unreliable. Finally, the analysis rests on a combination of assumptions that renders the evidence so unpersuasive and speculative that it is inadmissible under New Jersey Rule of Evidence 402.
A.
The polymarker and DQ-Alpha testing kits were designed solely to determine the presence or absence of certain alleles. Dot-intensity analysis, however, purports to determine more. It purports to quantify the alleles that are present and thereby to *245identify the specific alleles contributed by each donor to the DNA mixture. The majority only grudgingly rejects the State’s argument that dot-intensity analysis is nothing new and that no independent basis for its admission need be established. Without discussion, it recognizes, without really appreciating, that that difference requires an independent foundation for admissibility. Ante at 175, 699 A.2d at 624. Notwithstanding its concession, the majority then erroneously devalues and mischaracterizes defendant’s challenges to the evidence — challenges to its competency— as merely going to Cellmark’s performance of the polymarker test and as “concerning] not the admissibility, but the weight of the evidence.” Ante at 178, 699 A.2d at 625. That conclusion derives from a distortion of defendant’s claims and from a serious misunderstanding of the distinctive nature and purposes of dot-intensity analysis. Whether dot-intensity analysis is a proper interpretive procedure that is scientifically grounded and sufficiently reliable clearly goes to its admissibility as well as its weight.
Defendant’s claims concern the reliability of dot-intensity testing. Defendant specifically challenges “the reliability of th[e] test kit, its reliability in dealing with a mixed blood sample and the reliability of the propounded technique [dot-intensity analysis] for interpreting the results of a test on a mixed blood sample.” In fact, defendant’s exact claim on this appeal is that dot-intensity analysis is not generally accepted within the relevant scientific community. As noted in defendant’s brief, the dot-intensity “method of interpretation [was] not supported by Dr. Shaler, Dr. Blake, the FBI or even the research study conducted by Roche Molecular Laboratories, manufacturers of the kit in question.” Such challenges do not, as the Court asserts, go only to “Cell-mark’s performance of the polymarker test,” ante at 178, 699 A.2d at 625, and thus do not concern merely the weight of the evidence.
In order to avoid admitting that defendant actually challenged the reliability of dot-intensity analysis, the majority cites four *246cases in supports of its inaccurate assertion that defendant’s challenges to the accuracy of dot-intensity analysis go only to the weight to be attached to such evidence. Ante at 178-181, 699 A.2d at 625-626. Those eases, however, State v. Marcus, 294 N.J.Super. 267, 683 A.2d 221 (App.Div.1996); Fishback v. People, 851 P.2d 884 (Colo.1993); State v. Schweitzer, 533 N.W.2d 156 (S.D.1995); and State v. Kalakosky, 121 Wash.2d 525, 852 P.2d 1064 (1993), do not support the majority’s assertion.
Two of the cases cited by the Court, Marcus and Fishback, concern interpretation of “autorads.” Autorads are the x-ray type results obtained from RFLP-type DNA analysis. To compare RFLP results and to determine whether a match exists, autorads must be analyzed and compared. In fact, the obtaining and interpretation of autorads is a part of the RFLP process, and both courts found that RFLP testing, including the obtaining and interpretation of autorads, was generally accepted. Marcus, supra, 294 N.J.Super. at 278-79, 683 A.2d 221 (noting that “[t]he seventh step in the RFLP analysis involves the interpretation of the ‘autorads’ ... to determine whether the DNA print patterns of each of the probes match”); Fishback, supra, 851 P.2d at 888 (same). Thus, unlike dot-intensity analysis, autorads, as found by both courts, are a generally accepted method of presenting RFLP results, and the interpretation of autorads to reach conclusions regarding the inclusion or exclusion of a person as a DNA contributor is generally accepted. The only relevant disputes in those cases concerned “the ability of an expert to interpret the extra bands on an autorad.” Marcus, supra, 294 N.J.Super. at 291, 683 A.2d 221 (emphasis added). In other words, the dispute was not about what autorads purported to show but only about what to do with anomalous results. Notably, in Marcus, the State’s experts even “testified that the reading of autorads containing apparent extra bands ... is a generally accepted aspect of the field of DNA analysis.” Id. at 290, 683 A.2d 221. The issue here is not whether the reverse dot-blots obtained on the poly-marker strips can reveal the presence of alleles in the mixture— they can. At issue is whether an interpretation made of those *247strips that goes beyond what results that the strips were designed to show — the presence of alleles — is generally accepted as scientific evidence. Cf. Lowe, supra, 954 F.Supp. at 411 (finding that a switch to a new “technique for viewing band lengths [i.e. autorads] ... [that] has absolutely no effect on the validity of the overall methodology [i.e. RFLP]” still must be evaluated “to ensure reliability”).6 Thus, unlike “an expert’s ability to perceive an abnormality on an x-ray,” which concededly “is a matter within the province of the jury,” Marcus, supra, 294 N.J.Super. at 291, 683 A.2d 221, here we must decide, by analogy, whether a doctor’s interpretation of an x-ray can be admitted without restrictions when he testifies to a condition that the x-ray was not designed to reveal. Therefore, while a doctor’s diagnosis of a broken bone from an x-ray may be admissible because it is based on a generally accepted interpretation of a generally accepted test, the doctor’s diagnosis of cancer from that same x-ray ought not to be admitted unless and until the doctor can establish that such a diagnosis from an x-ray is generally accepted.
Nor does the Fishback case support the State’s assertion that dot-intensity analysis was properly admitted despite the absence of independent validation. The court in Fishback, supra, clearly acknowledged that “both the theory and techniques underlying novel scientific evidence [like the DNA test there at issue] must be generally accepted....” 851 P.2d at 891. The court also observed that “[a] standard requiring acceptance of only one or the other could lead to the illogical admission of evidence because the theory underlying that evidence is generally accepted even though the techniques for implementing it are highly suspect or controversial.”*2487 Ibid. Here, the soundness of the dot-intensity analysis is at issue, not whether such a procedure was done in a technically correct fashion.
The other two cases on which the majority relies are similarly unhelpful to its position. In Kalakosky, supra, the court rejected a defendant’s challenge to DNA results based on the specific laboratory procedures utilized. 852 P.2d at 1072. The defendant argued that human errors during the procedures rendered the results unreliable; specifically, the defendant alleged that the DNA samples had been accidently spilled together and that the record of the results had been mislabelled. The court correctly found those issues to be questions of fact, not challenges to the admissibility of DNA testing. Indeed, the defendant did not challenge the accuracy of the results if the process had been correctly performed. Id. at 1072-73. Defendant here, by contrast, challenged the use of dot-intensity analysis accurately to obtain particular results even if the process were performed in a technically correct manner.
The Court also cites the Schweitzer case, which concerned a dispute over an expert’s statistical conclusions on the DNA test results, not the performance of the DNA tests or the interpretation of those results. Unlike in this case, the defendant there “d[id] not dispute the scientific principles upon which [the expert’s] conclusions were based.”8 Schweitzer, supra, 533 N.W. 2d at 159.
*249New Jersey ease law also compels a finding that dot-intensity analysis requires independent validation. Repeatedly and consistently, this Court has held that it is not just the scientific principle but the “technique” or “mode of analysis” used by the expert that must have a sound scientific basis. E.g., State v. Kelly, 97 N.J. 178, 210, 478 A.2d 364 (1984); State v. Cavallo, 88 N.J. 508, 517, 443 A.2d 1020 (1982); State v. Hurd, 86 N.J. 525, 536, 432 A.2d 86 (1981); State v. Cary, 49 N.J. 343, 352, 230 A.2d 384 (1967). The application of that principle in Cavallo, supra, 88 N.J. 508, 443 A.2d 1020, highlights the fallacy of the State’s position that dot-intensity analysis needs no independent validation. We found in Cavallo that although “[tjhere is substantial support for the general acceptance of psychiatric witnesses in court ... the issue here is not the reliability of psychiatric testimony relating to an individual’s psychiatric condition. Rather, the question is the reliability of psychiatric testimony on the likelihood that an individual behaved in a particular manner on a specific occasion.” Id. at 522-23, 443 A.2d 1020. The issues that the Court held must be established prior to admission of the testimony were: “(1) whether psychiatrists agree that rapists share particular mental characteristics and (2) whether psychiatrists can ascertain if an individual possesses those characteristics by examining him.” Id. at 523, 443 A.2d 1020. The Court in Cavallo was obviously concerned with the state of scientific knowledge on each step of the processes presented in the case. Justice Pollock himself reiterated that principle last term in State v. Fertig, 143 N.J. 115, 668 A.2d 1076 (1996). In that case, writing for an unanimous Court, he rejected a per se rule either permitting or disallowing hypnotically-refreshed testimony, and ruled that such testimony would be admissible only if certain conditions that ensured the reliability of the process and the results were followed. Ibid. The Court today appears to depart from the standard accurately and clearly stated in the Fertig case, determining now that so long as the underlying scientific test or methodology is valid, then any “interpretation” purportedly derived therefrom seemingly is admissible, even if the *250reliability of that interpretive procedure itself has not been demonstrated.
The United States Supreme Court also has rejected the State’s the notion that interpretations of scientific tests need not be independently admissible. In Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), the leading case governing the admission of scientific evidence in federal courts, the Court recognized that “scientific validity for one purpose is not necessarily scientific validity for other, unrelated purposes.” Id. at 591, 113 S.Ct. at 2796, 125 L.Ed.2d at 482. In fact, an expert’s opinion must be based on the “methods and procedures of science,” not on “subjective belief or unsupported speculation.” Id. at 590, 113 S.Ct. at 2795, 125 L.Ed.2d at 481. “[I]n order to qualify as ‘scientific knowledge,’ an inference or assertion must be derived by the scientific method. Proposed testimony must be supported by appropriate validation — i.e., ‘good grounds,’ based on what is known.” Ibid.
In following Daubert, courts have held that “each step, from • initial premise to ultimate conclusion” must be examined. Hall v. Baxter Healthcare Corp., 947 F.Supp. 1387, 1401 (D.Or.1996). “In other words, this court need not accept, as scientifically reliable, any conclusion that good science does not permit to be drawn from the underlying data but which, instead, constitutes ‘unsupported speculation,’ or, ... a ‘leap of faith.’ ” Ibid. The Third Circuit recently confirmed such precedent:
Daubert’s requirement that the expert testify to scientific knowledge — conclusions supported by good grounds for each step in the analysis — means that any step that renders the analysis unreliable under the Daubert factors renders the expert’s testimony inadmissible. This is true whether the step completely changes a reliable methodology or merely misapplies that methodology.
[In Re Paoli R.R. Yard PCB Litig., 35 F.3d 717, 745 (1994), cert. denied, 513 U.S. 1190, 115 S.Ct. 1253, 131 L.Ed.2d 134 (1995).]
See generally David E. Bernstein, The Admissibility of Scientific Evidence After Daubert v. Merrell Dow Pharmaceuticals, Inc., 15 Cardozo L.Rev. 2139, 2164 (1994) (noting that under Daubert’s two-part test, “[o]nce a court has determined that the underlying *251studies or data relied upon by an expert witness are reliable and trustworthy, the remaining question is whether the expert has properly extrapolated from those studies or data”). Another federal court’s example of this principle sheds light on the problem with State’s analysis:
For instance, there may be times where an expert relies on published literature and widely accepted, tested theories in forming her opinion, and her ultimate conclusion is clearly relevant to an issue in the case. Yet, if those published theories and studies purport to prove XYZ, and from them, the expert concludes ABC, it may be that the expert’s reasoning process itself is not scientifically valid.
[Cavallo v. Star Enterprise, 892 F.Supp. 756, 761 (E.D.Va.1995), aff'd in relevant part and rev’d in part, 100 F.3d 1150 (4th Cir.1996), petition for cert. filed, 65 U.S.L.W. 3666 (U.S. Mar. 19, 1997) (No. 96-1493).]
Here, the “XYZ” in the example is polymarker testing; the unsupported “ABC” conclusion is that dot-intensity analysis can be performed on polymarker testing results.
The majority utterly ignores a second related underlying issue in dispute. The issue is not only whether it is possible theoretically to perform dot-intensity analysis, but also whether the poly-marker test kit was designed to and was capable of reliably measuring dot intensities as a basis for determining not only the presence or absence of alleles but the relative quantity of alleles present. Cf. Romano v. Kimmelman, 96 N.J. 66, 80, 474 A.2d 1 (1984) (“Once the showing of general acceptability has been made, courts will take judicial notice of the given instrument’s reliability and will admit in evidence the results of tests from the instrument without requiring further proof.”) (emphasis added) (citing State v. Johnson, 42 N.J. 146, 171, 199 A.2d 809 (1964)). For example, it has long been undisputed that radar beams directed at a moving object can accurately gauge the speed of the object. The issue we considered in State v. Dantonio, 18 N.J. 570, 115 A.2d 35 (1955), was whether a “radar gun” could reliably measure the speed of a vehicle. Similarly, in Romano, supra, 96 N.J. 66, 474 A.2d 1, and State v. Downie, 117 N.J. 450, 569 A.2d 242, cert. denied, 498 U.S. 819, 111 S.Ct. 63, 112 L.Ed.2d 38 (1990), the issue was not whether the amount of alcohol present in a person’s breath reflected the amount of alcohol in his or her blood, but rather whether the *252testing device reliably could measure the amount of alcohol on someone’s breath so as to be useftd as a forensic tool. An examination of the ability of the polymarker test kit reliably to interpret mixtures is thus necessary.
B.
The majority’s conclusion that dot-intensity analysis passes the general-acceptance test is most remarkable. First, not even the State has the temerity to suggests that it does; rather, the State wrongly argued that “[[Interpretation of the PM [polymarker] strips, and the resulting conclusions, are matters for the jury to resolve in deciding the weight of the evidence and not for the court to decide as a matter of law.” Having been forced to reject the State’s argument, the majority struggles in vain to sustain the use of this evidence. In order to find that this procedure — a procedure never before utilized in any court case, never successfully documented in any laboratory, and hardly even discussed in any publication or conference — passes muster, the Court wreaks havoc on our precedent and grossly distorts the record. Although the Court appears to pay homage to our vital precedents in this area, see ante at 172-174, 699 A.2d at 622-623, its application of those cases to the sparse facts here is a marked and aberrational departure from principle.
Before turning to the application of the general acceptance test, a brief reexamination of our precedent is necessary to understand the scope of the majority’s blunder. The admissibility of all evidence, but particularly scientific evidence being used in a criminal proceeding, must be “clearly established.” State v. Haskins, 131 N.J. 643, 649, 622 A.2d 867 (1993); Romano, supra, 96 N.J. at 90, 474 A.2d 1; Johnson, supra, 42 N.J. at 171, 199 A.2d 809. Such a high standard is justified because freedom — indeed life — is at stake. State v. Cary, 99 N.J.Super. 323, 333, 239 A.2d 680 (Law Div.1968), aff'd, 56 N.J. 16, 264 A.2d 209 (1970).
Scientific evidence is admissible only if the analysis used has “a sufficient scientific basis to produce uniform and reasonably reli*253able results so as to contribute materially to the ascertainment of the truth.” Kelly, supra, 97 N.J. at 210, 478 A.2d 364; see Romano, supra, 96 N.J. at 80, 474 A.2d 1; Cavallo, supra, 88 N.J. at 520, 443 A.2d 1020; Hurd, supra, 86 N.J. at 536, 432 A.2d 86. In Rubanick v. Witco Chem. Corp., 125 N.J. 421, 593 A.2d 733 (1991), this Court explained that “general acceptance.... entails the strict application of the scientific method, which requires an extraordinarily high level of proof based on prolonged, controlled, consistent, and validated experience.” Id. at 436, 593 A.2d 733; see Windmere, Inc. v. International Ins., 105 N.J. 373, 378 n. 2, 522 A.2d 405 (1987) (“[A] technique must pass from the ‘experimental’ stage and reach a ‘demonstrable’ stage before a court will recognize the approach.”).
In general, there are three ways to prove that evidence is generally accepted and therefore reliable:
(1) by expert testimony as to the general acceptance, among those in the profession, of the premises on which the proffered expert witness based his or her analysis; (2) by authoritative scientific and legal writings indicating that the scientific community accepts the premises underlying the proffered testimony; and (3) by judicial opinions that indicate the expert’s premises have gained general acceptance.
[State v. Haney, 121 N.J. 407, 427-28, 581 A.2d 483 (1990) (Haney I) (quoting Kelly, supra, 97 N.J. at 210, 478 A.2d 364), cert. denied, 499 U.S. 931, 111 S.Ct. 1336, 113 L.Ed.2d 268 (1991).] 9
*254Dot-intensity analysis does not come close to meeting those standards for the admission of scientific evidence. Dot-intensity analysis never has been successfully performed on any DNA sample anywhere, be it in an experiment or in an actual case. Not surprisingly then, no court case, no publication, and no scientist ever has concluded that dot-intensity analysis can and does work reliably and consistently.
So how can the majority possibly claim that such a novel scientific testing procedure is reliable? Despite its ultimate conclusion, it really does not. For example, in the one paragraph of the majority’s opinion devoted to the expert testimony on the general acceptance of dot-intensity analysis, the best the majority can do is to say that one scientist said that, in some (ill-defined) circumstances, unbalanced dot intensities may indicate differing relative presence of particular alleles. Ante at 182-183, 699 A.2d at 627-628. As for publications, the majority notes that “three *255articles discuss dot-intensity analysis.” Ante at 181, 699 A.2d at 627. Although the majority claims that “[n]ot one [of those articles] states that dot-intensity analysis is unreliable,” ibid., even it does not claim that any of the articles actually states or implies that dot-intensity analysis is in fact reliable. In the category of judicial opinions, the best the majority can say is that a single federal district court in New Hampshire, utilizing the more lenient federal standard, “recently accepted as reliable ... the premise underlying this interpretive method.” Ante at 184, 699 A.2d at 628 (emphasis added). One ambiguous opinion tendered by the State’s witness, three off-topic publications that make mention of the technique, and one equivocal bit of dicta do not add up to general acceptance.
Worse, though, the majority is forced to bend the record to obtain even those shreds of support. In fact, as has been and will be iterated and detailed, no court ever has accepted dot-intensity analysis, no scientist ever has defended dot-intensity analysis as a reliable and nonsubjective scientific method, and no literature or professional conference cited by the parties supports such a test.
Not even the State’s scientists defended dot-intensity analysis as a reliable, nonsubjective scientific method. Dr. Word and Ms. Cooper both admitted that the procedure is fraught with errors, and they both repeatedly noted that the results are utterly subjective.10 Dr. Word conceded that although “in many cases interpretations can be made of the mixtures [using dot-intensity analysis, this] certainly [can]not [be done] in all circumstances.” During the hearing and at trial, she defined neither in what circumstances it could be done nor in what circumstances it could *256not be done. In a letter to the prosecutor, she provided a few more details: “It is generally known to forensic scientists that it is usually impossible to determine the types of the individual donors if the DNA contains a mixture from 3 or more individuals using the [DQ-Alpha and polymarker] kits. This may also be true for mixtures of DNA from only 2 sources. However, for 2 source samples, it is sometimes possible to determine the types of each of the donors if there are intensity differences detected.” Letter from Dr. Word, Cellmark Diagnostics, to Mr. Corbin, Assistant Prosecutor, at 2 (Oct. 18, 1994) (emphasis added).
In fact, the extent of Dr. Word’s testimony on dot-intensity analysis was remarkably sparse, filling only a line or two of the transcript of her direct examination. In that brief testimony, she properly opined that unbalanced dot intensities could be an indication of a mixture. From that, she apparently decided — without any explanation or evident support — that she could draw conclusions regarding relative presence of alleles. She went on to admit on cross-examination, however, that many other reasons could explain the intensity differences: “[T]here’s [sic] several possibilities that could give me these dot intensity differences.” In fact, she attributed some dot-intensity differences to conditions other than different relative amounts of alleles. Thus, the majority’s characterization that “the testimony of the State’s expert supports the general acceptance of dot-intensity analysis,” ante at 180, 699 A.2d at 626, is exaggerated.
Moreover, Dr. Word’s testimony must be considered with and balanced against the testimony of other experts. Ironically, the principal supporter of the admissibility of polymarker testing in the first case in the nation to admit such testing was Dr. Robert Shaler. See Morales, supra, 643 N.Y.S.2d at 218. Dr. Shaier, called as a defense expert here, sharply denounced dot-intensity analysis as being “scientifically indefensible” at this time. Thus, not even he, a proponent of the use of polymarker testing in courts and a renowned expert in the field, viewed such analysis as reliable.11
*257As the Court appears to recognize, the scientific literature even discussing — much less endorsing — the validity of dot-intensity analysis is scant. Ante at 181, 699 A.2d at 627 (“[T]hree articles discuss dot-intensity analysis.”). In fact, no literature or professional conference cited by the parties supports the evidence. Cf. ante at 181, 699 A.2d at 627 (“Not one [article] states that dot-intensity analysis is unreliable. The articles go no further than to express caution when interpreting the results through dot-intensity analysis.”).
For example, the Federal Bureau of Investigations, in a widely respected and relied-upon analysis, found that “[b]ecause of the potential for unbalanced allele dot intensities and the limitations for detecting some mixed samples containing equivalent amounts of DNA, caution should be exercised when interpreting evidentiary samples that potentially may be from more than one donor.” F.B.I. Study, supra, 40 J. Forensic Sci. at 52 (emphasis added). That study, undertaken to determine “the limitations of the use of [the] new [polymarker] test,” id. at 45, thus found that even interpretations of the simple presence or absence of alleles should proceed cautiously when mixed samples are involved. In exercising that caution, the F.B.I. did not draw any conclusions from dot-intensity imbalances. The study even found that imbalances existed where none were supposed to exist.12 Id. at 48-49 & fig. 4.
*258The only other available study (only in prepublication format at the time of the hearing below) did not conclude that dot-intensity analysis was sufficiently reliable to warrant its use in an actual criminal case. See George Herrin, Nicola Fildes, & Rebecca Reynolds, Evaluation of the AmpliType PM DNA Test System on Forensic Case Samples, 39 J. Forensic Sci. 1247, 1251 (Sept.1994). That study does not bolster the reasoning or conclusion of the majority, despite the majority’s extended citation of it. Ante at 182-183, 699 A.2d at 627-628. The article does state the proposition that when an examination of the HBGG marker reveals the presence of the A, B, and C allele in a mixed sample (from two contributors) and when the known contributor is genotype A,A, we can conclude that the suspect is genotype B,C. Such a conclusion is presumed from the accepted testing tenet for determining the presence or absence of alleles, and the use of the simple subtraction principle that dictates that if all three alleles are present, and if the victim contributed only one type of allele, the other two alleles must have come from the suspect. The conclusion does not involve, depend on, or validate dot-intensity analysis — that is, the attempt to infer the presence of extra or additional alleles from intensity or “quantity” of alleles already present. One need not look to the intensity of the dots to reach that conclusion. That accepted method, however, was not the analysis that Cellmark performed on the markers in this case.
The State’s and majority’s use of the initial study done by Roche and six other laboratories and the validation work by Cellmark was misguided. Those studies had little to do with dot-intensity analysis. At most, the studies found that unbalanced dots could be “a valuable asset” for determining the existence of a mixture. Fildes & Reynolds, supra, 40 J. Forensic Sci at 285. The study, however, did not test the validity of that approach, nor did it even suggest that the presence of specific alleles could be determined by examining dot intensities. The publication concluded that “[t]he potential for sample mixtures in forensic casework analysis has always required careful and thoughtful interpretation. Individual laboratories will need to develop their own policies for *259the interpretation of mixtures based on their experience and case history information.” Id. at 284. The article does not suggest that any policy for interpreting mixtures has been developed, is scientifically reliable, or has earned general acceptance.
Although the Court asserts that Cellmark has developed such policies to interpret mixtures, ante at 181, 699 A.2d at 627, the Court does not indicate how it reached that conclusion, and it cites to no authority that would evidence its acceptance or reliability. Nothing in Cellmark’s own exhaustive diagnostic manual considers the application of dot-intensity analysis to interpret mixtures. In fact, noting in the record suggests that Cellmark had developed such policies. Rather, the manual states only that the polymarker test results are either positive or negative on the testing strips, indicating the presence or absence of a particular allele. Cellmark Diagnostics, Interpretation of HLA DQa Test Results, supra, at 1.
There is nothing in the entire literature or record to suggest that the polymarker test kit can do anything more than identify alleles and potentially show that a mixture exists in certain circumstances. Certainly, no evidence indicates that a determination of the relative presence of different alleles reliably and consistently can be made by examining polymarker testing strips from a mixed sample. Indeed, as we noted in a different context during our original review of this case, the State “did not provide evidence [for which] anyone in the scientific community other than [the State’s expert] [her]self vouche[d]....” Harvey I, supra, 121 N.J. at 428, 581 A.2d 483. There is simply nothing in the record to suggest that any scientist has ever in fact undertaken dot-intensity analysis successfully to determine the allele makeup of donors to a mixed sample.
The majority also relies on post-trial publications as “enhancing] the acceptance of dot-intensity analysis.” Ante at 183, 699 A.2d at 627. Whether post-trial extra-record publications can be relied on to support general acceptance is discussed later, see infra at 177-181, 699 A.2d at 625-626, but even the few relevant post-trial publications do not support the Court’s conclusion. The *260NRC Report, in the one line devoted to dot-intensity analysis, merely states that “[i]n some cases, it might be possible to distinguish the genetic profiles of the contributors to a mixture from differences in intensities of ... dots in a dot-blot typing.” NRC Report, supra, at 129 (emphasis added). The panel indicated only that it might be a marginally-helpful step that was worthy of further investigation; the scientists certainly did not suggest that dot-intensity analysis was in any sense reliable and admissible in court, though they advocated for the admission of virtually every other type of DNA analysis currently being undertaken. Dr. Word’s own publication and study, performed years after the admission of the evidence here, simply concludes that “[bjased on these studies, the PM kit can be expected to detect readily mixtures of DNA that fall within the criteria listed ... and the studies support the notion that relative dot intensities are a linear function of the number of copies of alleles contributed by each individual in the mixture.” Charlotte J. Word, Teryl M. Sawosik, & David H. Bing, Summary of Validation Studies from Twenty-Six Forensic Laboratories in the United States and Canada on the Use of the AmpliType PM PCR Amplification and Typing Kit, 42 J. Forensic Sci. 39, 46 (forthcoming 1997) (emphasis added). No testimony was adduced here about whether the mixture “fall[s] within the criteria,” nor does Dr. Word’s “notion” about dot-intensity analysis equate with a finding of general acceptance. In fact, in her conclusion, Dr. Word merely stated that the data she obtained supported the finding that dot-intensity analysis could detect mixtures. Ibid. Noticeably absent from her conclusion was any statement that dot-intensity analysis could effectively and reliably interpret mixtures.
Nevertheless, the majority, in concluding that dot-intensity analysis is generally accepted, appears to rely on the literature’s tentative and ambiguous assertions that such testing might theoretically be possible. E.g., ante at 181, 699 A.2d at 627 (noting that “three articles discuss dot-intensity analysis” and that “[n]ot one states that dot-intensity analysis is unreliable,” but ignoring fact that not a single article or judicial opinion states that dot-*261intensity analysis is reliable); ante at 181, 699 A.2d at 627 (speculating that the “esoteric” nature of dot-intensity analysis may explain the lack of publications); ante at 184, 699 A.2d at 628 (“No published article ... rejects dot-intensity analysis[,] ... [so] the trend supports acceptance of the test.”). The Court, of course, cannot possibly reconcile its statements with our long-standing standards that scientific evidence must be “clearly established,” Haskins, supra, 131 N.J. at 649, 622 A.2d 867, and that “general acceptance.... entails the strict application of the scientific method, which requires an extraordinarily high level of proof based on prolonged, controlled, consistent, and validated experience,” Rubanick, supra, 125 N.J. at 436, 593 A.2d 733. The Court’s opinion, reflecting an impossible struggle to document an insupportable conclusion, constitutes more words written about dot-intensity analysis than previously written in every publication, study, court opinion, and thesis combined. Although dot imbalances may well indicate the existence of a mixture, the use of those imbalances has not been refined and stabilized in a testing methodology that reliably, consistently, and clearly can be used to determine the genetic makeup of the contributors. Such analysis at this time remains in a speculative and penumbral realm, still far distant from scientific reliability.
C.
In determining whether dot-intensity analysis is sufficiently rehable to warrant its general acceptance, we can be informed by its application in this case. The evidence adduced in this case discloses that dot-intensity analysis does not yield consistent results. Moreover, to the extent that the State argues that it is consistently rehable, the results obtained would be inconsistent with defendant being the murderer.
Dr. Word testified that the sample from the boxspring contained the 2 and the 4 allele for the DQ-Alpha marker. She also noted that the victim’s DNA sample revealed that the victim had a “2,4” genotype for the DQ-Alpha marker. She further testified that *262the dot intensities on the DQ-Alpha test strip were “fairly, roughly equal,” indicating, according to the dot-intensity analysis theorem, that the 2 and the 4 alleles were present in roughly equal amounts. That being so, and the victim’s DQ-Alpha genotype being “2,4,” we would expect the suspect’s genotype also to be “2,4.” Cellmark, however, determined that defendant’s genotype is “4,4.”
If the sample contained the victim’s blood (“2,4”) and defendant’s blood (“4,4”), the intensity of the “4” dot should be significantly stronger than the intensity of the “2” dot if the dot-intensity theorem is valid. That was not the case. Dot-intensity analysis thus apparently would foreclose defendant as a possible contributor to the blood mixture found at the scene.
The majority appears unconcerned about that contradiction. First it notes that “the record contains no discussion of the application of dot-intensity analysis to the results of the DQ Alpha test,” ante at 180, 699 A.2d at 626, and then — without support — it changes its tune by saying that such analysis does not work, ante at 189-198, 699 A.2d at 631-632. We are forced to speculate whether or not dot-intensity analysis works on either the DQ-Alpha testing strips or on the polymarker testing strips, because there is insufficient evidence to establish its reliability and to support the general acceptance of the technique on either of the testing kits. The majority evades that inconsistency simply by deciding that whether or not dot-intensity analysis works on the DQ-Alpha is irrelevant to its analysis of the general acceptance of dot-intensity analysis of the PM probes. As the majority correctly points out, however, the DQ-Alpha testing procedure is factually similar to the polymarker testing procedure. Ante at 161, 699 A.2d at 617. Nothing in the record suggests that the slight differences between DQ-Alpha testing and polymarker testing would affect dot-intensity analysis.
Moreover, as acknowledged by Dr. Word during the pretrial hearing, the DQ-Alpha marker was retested using the PM testing strips. See also United States v. Shea, 957 F.Supp. 331, 334 *263(D.N.H.1997) (noting that the polymarker testing kit utilized in this case amplifies the polymarker loci as well as the DQ-Alpha locus); Herrin, Fildes, & Reynolds, supra, 39 J. Forensic Sci. at 1247 (noting that the AmpliType PM PCR amplification and typing kit amplified the five polymarkers and the DQ-Alpha marker). It is the polymarker testing strips that the State and the majority argue are capable of showing significant dot imbalances. For dot-intensity analysis to be valid, the analysis must work on the polymarker strip and with the DQ Alpha dot blots on that strip. There is no explanation in the record or the literature as to why the results for this particular marker would be any less conclusive than for the five polymarker markers.
Recognizing that dilemma, the majority then digs up extra-record authority for why dot-intensity differences may be irrelevant on the DQ-Alpha probes. Ante at 179-180, 699 A.2d at 625-626. Amazingly, the majority reads the authority as questioning dot-intensity analysis on the DQ-Alpha probe, yet it uses almost the exact same authorities and same ambiguous language in support of its claim that dot-intensity analysis on the polymarker loci is generally accepted. For example, the strongest support for the accuracy of dot-intensity analysis — on any loci — comes from a study done on the DQ-Alpha probes. See Edward Blake, Jennifer Mihalovih, Russell Higuchi, Sean Walsh, & Henry Erlich, Polymerase Chain Reaction (PCR) Amplification and Human Leukocyte Antigen (HLA)-DQa Oligonucleotide Typing on Biological Evidence Samples: Casework Experience, 37 J. Forensic Sci. 700 (1992). That study notes that with DQ-Alpha testing, “mixtures can be identified and interpreted based on relative dot intensities” in certain circumstances. Id. at 706. That is the only study on any loci to conclude that dot-intensity analysis can interpret, and not just identify, mixtures. The support for concluding that dot-intensity differences matter on the DQ-Alpha probes is much stronger than the support for the majority’s conclusion that dot-intensities matter on the polymarker probes. The bias of the majority’s reading of the authorities is starkly evident.
*264Other testing strips for the polymarker markers themselves indicate that dot-intensity analysis is subjective and does not work properly. The polymarker testing of the HBGG marker revealed that the boxspring mixed sample had all three possible alleles for that marker: A, B, and C. Dr. Word and Ms. Cooper testified that the dot-intensity test revealed that the three alleles were “roughly balanced” and of the “same intensity.” Assuming the mixture came from only two individuals (an assumption that must be true for the State’s analysis to work) and assuming that the mixture contains roughly equal amounts from each individual (another required assumption), there is no possible combination of genotypes that would have resulted in equal amounts of the three different alleles.13 In fact, if defendant’s blood (with HBGG genotype “A,C”) and the victim’s blood (with HBGG genotype “B,B”) were in a mixture, dot-intensity analysts would assume that the B dot would be brighter on the testing strip. It was not.
One possible explanation for that apparent inconsistency is that homozygous genotypes (where the two alleles composing the genotype are the same) may show up on the test strips as having the same intensity as single alleles. However, even that speculative explanation suffers, because it defies the logic of the polymarker testing, which separates the genotypes (including homozygous ones). In any event, that explanation has been cast into doubt by a study that showed that “the allele dot of a homozygous profile is more intense than the dots of a heterozygote profile.” F.B.I. Study, supra, 40 J. Forensic Sci. at 12 (emphasis added). There*265fore, the results of the polymarker test for the HBGG marker should have shown a B dot that was brighter than the A and C dots if defendant’s blood was a part of the mixture. The test did not show that result. The Court, for its part, continues to disregard the fact that such results undermine the reliability and hence the admissibility of dot-intensity analysis.14
Other results also indicate the problems with dot-intensity analysis. Each individual, we know, has two alleles (which together compose a genotype) for each particular marker. Therefore, if a sample of DNA were extracted and analyzed using the described testing procedures, and if the dot-intensity thesis is correct, a sample composed of just one individual’s DNA should reveal, on the marker being examined, equal dot intensities for the alleles of that person. In other words, a person’s DNA sample that had the A,B genotype for the GYPA marker should show equal dot intensities for the A and B alleles and should be blank for the C allele dot. That was not so in this case.
On defendant’s sample, both experts for the State, Dr. Word and Ms. Cooper, admitted that “[t]here’s a possible intensity difference ... with GYPA” and that “[the A dot] is somewhat darker than the B.” That “variance,” Dr. Word stated, is a common one, particularly among African Americans. That “variance,” and any other such variances that the State opts to acknowledge, may explain the result obtained here, but the variances destroy the integrity of dot-intensity analysis.15 To the degree that there are naturally occurring variances in dot intensi*266ties that cannot be predicted prior to the analysis, dot-intensity analysis disintegrates as a reliable and useful test.
A real-life example demonstrates the problem. The GYPA results from the mixed sample showed the A dot as darker than the B dot and show no color reaction at the C dot. Ms. Schnaps’s sample showed her genotype for the GYPA marker to be A,A. Thus, the other contributor to the sample could be either genotype A,B or B,B. Based on the fact that the A dot was darker than the B on the mixed sample, the State concluded that the suspect must be genotype A,B, because otherwise the A and B dots on the mixed sample would have been of equal intensity. The point is simply that variances, like the one at the GYPA marker that caused the A dot on defendant’s strip to be brighter than the B dot, upset the analysis. Taking that example, if A dots for the GYPA marker are sometimes more intense than B dots even when equal amounts of the A and B alleles are present (as must have been true in defendant’s unbalanced sample), then the suspect here could have had either genotype A,B or B,B. Both possibilities could reveal, when mixed with an A,A genotype (Ms. Schnaps’s), a more intense A dot than B dot. Hence, dot-intensity analysis is useless unless somehow the variances are factored into the analysis. Variances were not accounted for here. In fact, the State concluded that the suspect must have GYPA genotype A,B — a result that, due to the variance at this exact marker, is not necessarily accurate.16
The example also demonstrates the problems with not knowing the precise amounts of DNA that each person contributed to the mixture. If the victim’s blood was in heavier concentration than *267the suspect’s blood, then the A dot may be more intense than the B dot for the mixed sample regardless of whether the suspect’s genotype is B,B or AB. Thus, the Court’s assertion that Cellmark made only two assumptions in its analysis, ante at 164, 699 A.2d at 618, is also incorrect.
Besides the dot-intensity differences on the polymarker strip at defendant’s HBGG marker, a possible disparity was also spotted on the GC locus. Further, examining the victim’s polymarker strips, the defense expert found a clear imbalance in the GC marker, and imbalances in the control17 on the GYPA marker. In total, of the six strips where dot intensities can be compared18 and where no dot-intensity imbalances are supposed to exist because the alleles are present in equal amounts, two markers show clear imbalances, two other markers show slight imbalances, and two markers show no imbalance. The accuracy rate of this procedure would then hover somewhere around or below the fifty-percent range.19
*268The problems found here are not problems unique to this case. The F.B.I. study found similar difficulties with dot imbalances where there were supposed to be none. F.B.I. Study, supra, 40 J. Forensic Sci. at 12-13 (“The exact percentage of samples that exhibit unbalanced allele dot intensities is difficult to determine because the determination of unbalanced intensity is somewhat subjective.”). The F.B.I. study concluded that the dot-intensity imbalances were not a problem because “none of these observations of unbalanced alleles resulted in an incorrect type.” Id. at 13. As previously noted, however, the F.B.I. was not trying to interpret the relative presence of alleles from mixed samples where incorrect intensity imbalances would affect the conclusions. In short, the imbalances did not affect the conclusions that the F.B.I. was willing to draw, but such imbalances would deconstruct the dot-intensity analysis in this case, which sought to determine the relative presence of certain alleles.
Dr. Shaler accurately summed up the problems with dot-intensity testing: “[It] is a flawed system in that the dot intensities are already unbalancedf.] [T]hey can’t use un[]balanced alleles to derive conclusions regarding types present or not present.” He further noted that Cellmark technicians got “differences in dot intensities when they’re supposed to be equal and since they’re starting off with something which is unbalanced and they end up with something which is unbalanced.... [Y]ou can’t do that.” In sum, the patent unreliability of dot-intensity analysis requires its exclusion. See Hurd, supra, 86 N.J. at 536-37, 432 A.2d 86 (noting that expert testimony must be excluded if there is not even a reasonable likelihood that such testimony was reliable).
Not only do the results obtained here establish the gross unreliability of this evidence, but the entire practice of visualizing and weighing dot intensities to determine the makeup of a mixture is unavoidably subjective. A subjective test, especially one that is immune from later challenge, should not be admissible evidence in these circumstances. The standard for the admissibility of scientific evidence is designed to ensure that the testing procedure *269“relies primarily upon objective factors for reaching a conclusion, with subjective factors playing only a minimal role in the analysis.” Windmere, supra, 105 N.J. at 385, 522 A.2d 405. Even the State seems to have understood the problem. In questioning Dr. Shaler, the prosecutor noted:
PROSECUTOR: Now, would it be fair to say that the review of any of these polymarker dots or alleles the intensity is a subjective opinion of an individual, correct?
DR. SHALER: Pretty much.
PROSECUTOR: So depending upon the individual they may have different interpretations?
* * * *
DR. SHALER: I don’t feel these kinds of interpretations are reliable. Nobody to my knowledge has undergone a blind examination to test their abilities to make these kinds of intensity difference decisions.
Ms. Cooper, one of the State’s experts, also admitted that “intensity differenee[s] ... can be subjective.”
The subjective nature of this proof is shown by the evidence here. Cellmark’s own laboratory technicians could not agree on whether or not some dot intensities were imbalanced. In analyzing the sample taken from the boxspring, one of the two technicians found three of the dot-intensity results to be imbalanced while the other reader found four of the strips to show imbalances. The defense expert found five of the- strips to be imbalanced. No court can accurately assess those conclusions, because the dots fade and begin to disappear almost immediately.20 With such gross disparities in an admittedly subjective test, the Court cannot conclude that dot-intensity analysis is reliable.21
*270D.
Even if dot-intensity analysis were presumed to be generally accepted, the evidence should have been rejected here because the assumptions underlying it were so numerous, complex, problematic, and potentially flawed as to render it substantially more confusing and prejudicial than probative in this specific context. See N.J.R.E. 403; Cavallo, supra, 88 N.J. at 520, 443 A.2d 1020 (holding that expert evidence, even if admissible under N.J.R.E. 702, must be excluded if it poses the danger that “prejudice, confusion and diversion of attention exceeds its helpfulness to the fact finder”). Expert testimony, especially testimony such as this complex DNA evidence, must be scrutinized with particular care because of the “aura of special reliability and trustworthiness” that surrounds it. State v. Berry, 140 N.J. 280, 299, 658 A.2d 702 (1995) (quoting United States v. Young, 745 F.2d 733, 765-66 (2d Cir.1984), cert. denied, 470 U.S. 1084, 105 S.Ct. 1842, 85 L.Ed.2d 142 (1985)). Such scrutiny is even more important in a capital case. See State v. Johnson, 120 N.J. 263, 296-99, 576 A.2d 834 (1990) (finding, in a capital case, blood-spatter evidence to be more prejudicial than probative).
For dot-intensity analysis to work, the blood being analyzed must come from two and only two persons. One of those persons must be the victim (or a different known (or typeable) person), and the other contributor must be the suspect. The victim’s blood and the suspect’s blood must be present in roughly equivalent amounts, and the two blood samples, when mixed, must not interact or affect one another. Those assumptions are logical leaps that may or may not be justified in a particular situation. *271Here, however, it is not possible to establish the validity of the assumptions because they were not even defended by the State.
If one assumes that dot-intensity analysis works, one must conclude that in the present case, the dot-intensity results obtained for the mixed sample in the HBGG marker preclude the possibility that only two people contributed to the sample. Dr. Word testified that she could not tell whether the boxspring sample contained blood from more than two sources. She erroneously opined, however, that “[t]he dot intensities fit together nicely for there being two sources.” Nevertheless, she conceded that, “[t]here are probably some combinations of three sources that you could fit together to give us these same results.”
Although her analysis was correct, her assumption that the dot intensities fit together was erroneous. As previously noted, the results from the HBGG marker (balanced dot intensities for the A, B, and C alleles) do not comport with the State’s conclusions that the blood was a mixture of defendant’s DNA (HBGG genotype “A,C”) and the victim’s DNA (HBGG genotype “B,B”). See discussion supra at 164-168, 699 A.2d at 618-620. One way that the HBGG test strip dots for the A, B, and C alleles could be balanced is if three individuals contributed to the DNA sample. The three individuals could be the victim (“B,B”), the defendant (“A,C”), and a third person (“A,C”). Two individuals could not have produced the results obtained unless a different assumption the State made was wrong.22
The State assumed that the two contributors to the sample contributed roughly equivalent amounts of blood and DNA. Relaxing that assumption, explains the anomalous HBGG result in that the suspect actually contributed twice the amount of DNA to the *272sample as did the victim. The relaxation of that assumption, however, renders other results suspect. Moreover, the need to speculate only highlights the fact that the assumptions made by the State, while necessary to the validity of the dot-intensity analysis, are suspect and even provably wrong. Moreover, altering assumptions to fit the results obtained assumes defendant’s guilt, rather than establishing his guilt. The assumptions, however, must be established independently of the results.
One other assumption warrants brief discussion because the State actually found problems with it. For dot-intensity analysis to be valid, one must assume that the DNA samples from two different people do not affect each other when mixed together. But, Dr. Word admitted that relative probe intensities become imbalanced in the presence of variant alleles. As previously noted, a variant allele was found in defendant’s GYPA marker. See discussion supra at 166-168 & n. 14, 699 A.2d at 619-620 & n. 14. Despite that fact, the State relied on the GYPA dot-intensity result. Moreover, a variance probably existed at the GC loci as well. See Cooper DNA Opinion, supra, at 19-20.
Those assumptions bear directly on the admissibility, not merely the weight, of dot-intensity analysis. A full hearing on the assumptions and the entire validity of the dot-intensity analysis should have been held. That hearing was necessary to explore the inconsistencies in both the State’s experts’ comments and in the actual results obtained. The uncritical admission of this evidence in a capital trial without even remotely establishing its validity is an egregious wrong.
E.
The burden is on the proponent of evidence clearly to establish its admissibility. Windmere, supra, 105 N.J. at 378, 522 A.2d 405. Although the majority claims to recognize that rule, see ante at 171, 699 A.2d at 622, it continually ignores that burden and notes that defendant did not challenge the evidence, that defendant did not present more favorable evidence, or that defendant based his *273conclusions on the same evidence, ante at 178, 179, 191, 699 A.2d at 625, 626, 632. The fact that defendant did not adequately or effectively challenge the evidence does not relieve the State of its burden of clearly establishing the admissibility of novel scientific evidence. Nor is the trial court relieved of its obligation to weigh and to determine the admissibility of the evidence prior to its admission.
Most surprising is the Court’s admission that this evidence is “esoteric.” Ante at 183, 699 A.2d at 628. Esoteric and novel scientific evidence should undergo the most rigorous scrutiny because that is the evidence that is least likely to have been subject to the type of testing and peer review that we demand of all scientific evidence.
In fact, it is our court’s obligation, in reviewing the admission of novel and complex scientific evidence, to assure itself that such evidence is reliable and generally accepted. See, e.g., Kelly, supra, 97 N.J. at 214, 478 A.2d 364 (recognizing such an obligation and acknowledging that the Court has power to order the trial court to conduct additional hearings to evaluate the claims of competing experts and to determine the reliability of the scientific evidence). As such, we must undertake a de novo review of that aspect of the record and can supplement the record with judicial opinions, scientific articles, and our own analysis. See Lindsey v. People, 892 P.2d 281, 289-90 (Colo.1995) (ruling that admission of novel DNA evidence “is a question of law” subject to de novo review); Brim v. State, 695 So.2d 268, 274 (Fla.1997) (holding that trial court’s ruling admitting DNA evidence is subject to de novo review and is “reviewed as a matter of law rather than by an abuse-of-discretion standard”); People v. Miller, 173 Ill.2d 167, 219 Ill.Dec. 43, 60-62, 670 N.E.2d 721, 738-40 (1996) (McMorrow, J., specially concurring) (“[T]rial court decisions regarding the threshold question of whether a scientific technique has achieved general acceptance in the relevant scientific community should be subject to de novo review.”), cert. denied, — U.S. -, 117 S.Ct. 1338, 137 L.Ed.2d 497 (1997); Taylor v. State, 889 P.2d 319, 331-*27432 (Okla.Crim.App.1995) (noting that the court conducts “an independent, thorough review” and does not “simply ask whether an abuse of discretion was committed”); State v. Lyons, 324 Or. 256, 924 P.2d 802, 805 (1996) (“Notwithstanding the usual deference to trial court discretion, we review that ruling on the admissibility of scientific evidence de novo.”) (internal citations omitted); State v. Cauthron, 120 Wash.2d 879, 846 P.2d 502 (1993) (‘We review the trial court’s decision to admit or exclude novel scientific evidence de novo.”); see also State v. Alberico, 116 N.M. 156, 861 P.2d 192, 204-06 (1993) (refusing to abandon abuse-of-discretion standard for admission of scientific evidence, but recognizing that such a standard “lends itself to the criticism that it will lead to inconsistent results in lower courts throughout the state”). The Florida Supreme Court, in adopting the reasoning of Justice McMorrow of the Illinois Supreme Court, see Miller, supra, 219 Ill.Dec. at 61, 670 N.E.2d at 739 (McMorrow, J., specially concurring), recently explained why this was so: “Foremost is the fact that the general acceptance issue transcends any particular dispute.” Brim, supra, 695 So.2d at 274 (citing Jones v. United States, 548 A.2d 35, 40 (D.C.App.1988)).
The majority accepts that our review of this evidence is de novo when such evidence is admitted against a criminal defendant. Ante at 165-168, 699 A.2d at 619-620. Indeed, the Court frequently relies on post-hearing judicial opinions, articles, and analyses to shore up its position. See, e.g., ante at 177, 699 A.2d at 625 (“Based on the record, as well as on posttrial publications and judicial opinions, we conclude that the trial court correctly allowed [such evidence].”); ante at 183, 699 A.2d at 627 (“Recent publications enhance the acceptance of dot-intensity analysis.”); ante at 175, 183, 699 A.2d at 624, 627 (relying on NRC’s post-trial report to support position); ante at 175-176, 699 A.2d at 624 (citing post-trial cases to support general acceptance of polymarker testing).
Despite the majority’s acceptance of de novo review, the majority carefully circumscribes such review to exclude all of the evi*275dence that does not support its ruling. See, e.g., ante at 180, 699 A.2d at 626 (noting that “the record does not reveal whether dot-intensity analysis can be performed on a DQ Alpha test,” but ignoring evidence that it can be); ante at 181-182, 699 A.2d at 626-627 (acknowledging that the HBGG result obtained for the mixed sample is unexplainable by dot-intensity analysis, but refusing to explore inconsistency until post-conviction-relief proceedings); ante at 183-184, 699 A.2d at 628 (recognizing that dot-intensity analysis “provides an opening for cross-examination and contradictory expert testimony,” but refusing to remand the case for a hearing to develop such evidence); ante at 193, 699 A.2d at 633 (“It is too late in the proceeding for defendant to insist that the State should have presented statistical evidence that defendant now believes would have been helpful at trial.”); ante at 194, 699 A.2d at 633 (noting that “[o]n this record, the dissent’s arguments are particularly unpersuasive,” but ignoring the weight of contrary precedent); ante at 200, 699 A.2d at 636 (rejecting defendant’s claim in part because he missed his “opportunity to challenge the State’s statistical evidence, to present his own evidence, [or] to argue to the jury”). In fact, the majority even strikes from the record contrary evidence because it determines that when evaluating scientific evidence “[a]n appellate court ... generally confines itself to the record.” Ante at 201, 699 A.2d at 637; see also ante at 202, 699 A.2d at 637 (“The place to introduce expert testimony is at trial....”).
Accepting post-hearing evidence that only supports one side of the argument is indefensible. As Justice McMorrow noted,
[t]he majority cannot have it both ways. If trial court decisions concerning the general acceptance of novel scientific evidence cannot be reversed absent an abuse of discretion, then upon review, only material which was part of the trial record should be considered by this court. If, on the other hand, the majority believes that it is proper to rely on scientific articles and court cases which were not part of the trial record to determine whether a novel scientific technique has become generally accepted in the relevant scientific community, then the majority must acknowledge that the standard of review is not a simple abuse of discretion standard.
[Miller, supra, 219 Ill.Dec, at 60, 670 N.E.2d at 738 (McMorrow, J., specially concurring).]
*276Dot-intensity analysis is a separate scientific technique that must be generally accepted prior to the evidence being used against a defendant in any criminal trial, and most especially in a capital prosecution. From what little we know about the analysis, it is not generally accepted or reliable. The procedure enjoys no support aside from Dr. Word’s equivocal statements. In any event, the technique should have been excluded here because of the number of questionable and provably wrong assumptions that were made. Those assumptions render the evidence in this ease substantially more confusing than clarifying and more prejudicial than probative.
III
The trial court prevented defendant from effectively challenging the reliability of both the polymarker testing procedure and Dr. Word’s theory regarding dot-intensity analysis. Specifically, the court did not permit defendant to examine the scientific experts on Cellmark’s validation studies or to elicit testimony from defendant’s own expert on the F.B.I. study. Cellmark’s validation study showed that not even Cellmark, the laboratory responsible for performing the prosecution’s DNA testing, conducted tests on the reliability or accuracy of the dot-intensity analysis. See discussion supra at 161-168, 699 A.2d at 617. The F.B.I. study concluded that dot imbalances occurred when they were not supposed to occur and that, therefore, the F.B.I. would not perform or rely on dot-intensity analysis for casework. See discussion supra at 160-161, 699 A.2d at 616-617. The majority acknowledges the trial court’s error in limiting examination on the F.B.I. study, but finds the error to be harmless, and inexplicably finds no error in the limitation on cross examination regarding Cellmark’s study.
The majority’s reasoning is particularly troubling because of its previous decision to allow into evidence the “esoteric” dot-intensity analysis without much, if any, support from the scientific community. The majority reasons that all of defendant’s complaints *277about dot-intensity analysis “concern not the admissibility, but the weight of the evidence.” Ante at 178, 699 A.2d at 625. Despite that determination — that defendant’s complaints go only to the weight — the majority finds the limitations placed on the defense examination of the experts regarding Cellmark’s validation study to be proper because “[t]he issue of Cellmark’s validation studies on the polymarker [test] more properly concerns the reliability of the polymarker [test]. As such, the issue goes to the admissibility of the test and not its weight.” Ante at 188-189, 699 A.2d at 630 (emphasis added). The contradiction is overwhelming and confounding.
The majority’s decision to sanction the limitations placed on the examination into these subjects is not reasonable under any circumstances. The fact that experts dispute whether a particular scientific technique has gained general acceptance is relevant and material not only to the admission of the evidence, but also to the weight that the jury should attach to the evidence, even if the technique in question is found admissible.
Admissibility and weight of scientific evidence may overlap. This Court repeatedly has noted that general acceptance does not mean unanimous acceptance. E.g., Windmere, supra, 105 N.J. at 379, 522 A.2d 405; State v. Tate, 102 N.J. 64, 83, 505 A.2d 941 (1986); Kelly, supra, 97 N.J. at 178, 478 A.2d 364. To the extent that a scientific technique, although generally accepted, does not enjoy unanimous acceptance, the party attacking the evidence must be permitted to inform the jury that the technique does not enjoy full support. To hold otherwise, as the Court does today, would be to immunize scientific techniques from attack and to mislead the jury into believing the infallibility of the techniques.
The weight of the evidence is a matter for the jury to decide. Cavallo, supra, 88 N.J. at 520, 443 A.2d 1020. ‘“[A]n expert witness is always subject to searching cross-examination as to the basis of his opinion.’ ” State v. Martini, 131 N.J. 176, 264, 619 A.2d 1208 (1993) (Martini I) (quoting Glenpointe Assocs. v. Township of Teaneck, 241 N.J.Super. 37, 54, 574 A.2d 459 (App.*278Div.), certif. denied, 122 N.J. 391, 585 A.2d 392 (1990)), cert. denied, — U.S. -, 116 S.Ct. 203, 133 L.Ed.2d 137 (1995). Indeed, “[t]o determine the credibility, weight and probative value of an expert’s opinion, one must question the facts and reasoning on which it is based.” Ibid, (citing Johnson v. Salem Corp., 97 N.J. 78, 91, 477 A.2d 1246 (1984)) (emphasis added).
An expert’s credibility always can be attacked using treatises or other authorities on the subject. See Jacober v. St. Peter’s Med. Ctr., 128 N.J. 475, 486, 608 A.2d 304, modified in part on other grounds, 130 N.J. 586, 617 A.2d 1213 (1992). That is especially true when the expert has relied on the literature in forming his or her opinion. Id. at 494, 608 A.2d 304 (noting that “juries should not be insulated from the literature that experts use in formulating their opinions”). Indeed, without such evidence, the factfinder would be inhibited “from assessing the credibility of an expert’s opinion by reference to that opinion’s support in the relevant literature.” Ibid.; see also Blitz v. Hutchinson, 252 N.J.Super. 580, 594-95, 600 A.2d 485 (App.Div.1991) (finding error in the refusal of trial court to permit cross examination of expert about an article on which he had relied but which contradicted the party’s position). Under those well-established standards, the trial court’s limitations on defendant’s challenges to this evidence were erroneous.
The Court, as noted, does find error in the trial court’s refusal to permit questioning on the F.B.I. study. Yet it finds the error to be harmless. Ante at 189, 699 A.2d at 630. The majority comes to that conclusion while recognizing that the “polymarker evidence was an important link in tying defendant to the crime.” Ante at 155, 699 A.2d at 613-614. The dot-intensity analysis added the steel to the links that chained defendant to the crime. Without the dot-intensity analysis, the DNA evidence would have been ambiguous and possibly inadmissible under N.J.R.E. 403. Yet, the majority maintains that an error going to the very foundation of the dot-intensity analysis was harmless. Cf. Williams v. State, 342 Md. 724, 679 A.2d 1106, 1120-22 (1996) *279(holding that limitations on cross examination of state’s DNA expert was prejudicial error, and that challenges to “the reliability of the testing procedures used by Cellmark____could have been vital to the jury’s determination of how much weight to give to the PCR test results”).
The dot-intensity analysis was used by the State to narrow drastically the number of individuals who could have contributed to the blood mixture found at the scene. The technique permitted the State to argue that far less than one percent of the population had the types of DNA markers found at the scene. (Actually, as will become evident, the State erroneously argued that only one out of 1,400 persons had those markers. The correct figure, according to the State’s own expert, even using dot-intensity analysis, is far more inclusive.) Without dot-intensity analysis, at least ten percent of the African-American population could have contributed to the blood found at the scene.23
The majority finds the error harmless because “the jury knew that some scientists questioned the validity of dot-intensity analysis.” Ante at 189, 699 A.2d at 631. That assertion is simply wrong. The jury knew of only one person who challenged the method — defendant’s own witness, Dr. Shaler. The jury did not know that anyone not hired by the defense did not support the evidence. Indeed, the jury was led to believe that at least the other two scientists from whom the jury heard, Dr. Word and Ms. Cooper, as well as the laboratory performing the tests, Cellmark *280Diagnostics, accepted the technique.24 In reality, the F.B.I.’s study (undertaken by six experienced forensic scientists) lends no support to and flatly contradicts the assumptions of dot-intensity analysis. In fact, no scientist in any study has supported dot-intensity analysis. The jurors did not possess that information. Had they known, they may well have rejected as speculative, unsupported, and unreliable, the technique used by the State. Had the genetic evidence been rejected, a conviction would have been improbable.
Errors that are “clearly capable of producing an unjust result” demand reversal. R. 2:10-2. Where, as here, the party objects to the ruling and is precluded from introducing material evidence, the test is the sufficiency of the error “to raise a reasonable doubt as to whether the error led the jury to a result it otherwise might not have reached.” State v. Macon, 57 N.J. 325, 336, 273 A.2d 1 (1971); see also State v. Bey, 112 N.J. 45, 94, 548 A.2d 846 (1988) (Bey I) (“[T]he inquiry concerns whether the error contributed to the verdict or the sentence.”). Here, the evidence of defendant’s guilt was clearly not “overwhelming.” Cf. State v. Tirone, 64 N.J. 222, 227, 314 A.2d 601 (1974) (finding error not capable of having produced an unjust result because of the overwhelming evidence of the defendant’s guilt). Certainly, where, as here, the error infects the most substantial evidence admitted at trial and where the error has the effect of precluding a meritorious attack on the very foundation of that evidence, the error cannot be said to be harmless.25
*281IV
The State used the results of dot-intensity analysis to narrow the class of individuals who could have contributed to the blood found at the scene. The quantification and definition of that class is critical in forensic DNA testing. The persuasiveness of DNA testing is not simply to determine whether a defendant can be included or excluded from the class of persons having the same genetic markers as the DNA recovered from the scene, but also to determine the size of the class of individuals sharing those markers. From the percentage of the population having certain DNA markers, one can determine how probable or improbable it is that the defendant’s blood was the blood found at the scene. By creating a population-frequency database and then analyzing the results, DNA testing can reveal the percentage of individuals who share a particular genetic marker. The State, here, went even further. The State determined the percentage of individuals having each of the six markers analyzed in the polymarker and DQ-Alpha testing and then multiplied the results together ostensibly to reveal the percentage of the population sharing all six of the markers. The State asserted that only one in 1,400 African-Americans shared the genetic markers of the blood found at the scene. The State’s figure is wrong; the analysis is misleading, flawed, and unreliable. The entire statistical evidence was improperly admitted without foundation and without a hearing.
A.
Dr. Word, a genetic scientist, testified about frequency data and frequency calculations. Frequency data was obtained by testing certain population groups and determining how often the different genotypes for the DQ-Alpha and the five polymarker markers occur. Frequency calculations are then performed by multiplying *282the frequencies of the various genotypes together. So if a particular polymarker marker genotype is present in twenty percent of the population and a different polymarker marker genotype is present in fifty percent of the population, by multiplying twenty percent by fifty percent, Cellmark would conclude that only ten percent of the population shares those two genotype markers. This calculation is called the “product rule.”
In this case, Cellmark used data collected from its own laboratory and from Roche laboratory. Cellmark’s database came “from individuals in our laboratory and ... from paternity casework blood samples.” Cellmark compiled separate databases for Caucasians and for African-Americans. Cellmark’s database contained fifty African-Americans, Roche’s contained 100.
Applying the product rule to the types of markers defendant possessed (the five polymarker marker genotypes and the DQ-Alpha genotype), Dr. Word and Ms. Cooper concluded that defendant’s particular genotypes were shared by one in 1,400 African-Americans. Using Cellmark’s own data, but excluding the figure for the DQ-Alpha marker, the product rule reveals that approximately one in 170 African-Americans would share defendant’s genotypes for the five polymarker markers.
Without using the dot-intensity analysis to limit the possible genotypes, Dr. Blake, one of defendant’s experts, used the product rule to limit the African-American population sharing the suspect’s DNA characteristics to one in forty-seven. Dr. Word admitted that Dr. Blake’s analysis was correct if one “totally ignore[s] the [dot-]intensity differences.”
B.
The one-in-1,400 figure held up by the State is plainly and simply misleading and wrong. In summation, the State encouraged the jury to use the product rule and asserted that the jurors should conclude that only one in 1,400 African-Americans had the same DNA type as that found at the crime scene: “That is excluding one thousand three hundred and ninety-nine people *283from, that sample,[26] That’s excluded over ninety-nine point seven three ... of the entire population of the world.”27 (emphasis added).
Although the State was correct that defendant had that DNA type, the State is flat out wrong in asserting that the blood recovered from the scene revealed the “4,4” DQ-Alpha genotype. To get the one-in-1,400 figure, the State relied on dot-intensity analysis. Finding the DQ-Alpha “4,4” genotype to be present in 11.9% of African-Americans,28 the LDLR “B,B” to be present in *28456% of African-Americans, the GYPA “A,B” to be present in 50%, the HBGG “A,C” in 27%, the D7S8 “A,B” to be present in 45%, and the GC “A,B” genotype to be present in 17% of African-Americans, the State concluded that only one in 1,400 African-Americans shared those genotypes. The State’s calculations were correct (.119 x .56 x .50 x .27 x .45 x .17 = 0.0006882, which in turn equals 1/1,458). However, even assuming the figures were entirely accurate, the State’s calculations were flawed.
The State, relying on Cellmark’s figures, found that 11.9% of African-Americans have the “4,4” DQ-Alpha marker. Defendant has the “4,4” DQ-Alpha marker. However, the blood found at the scene could have come from someone with DQ-Alpha marker that was either “2,2” or “2,4” or “4,4.” (Indeed, as discussed earlier, see supra at 165, 699 A.2d at 619, using the dot-intensity analysis, one would expect the suspect to have the “2,4” DQ-Alpha genotype.) Even Cellmark recognized that the suspect contributing to the mixture must have a DQ-Alpha genotype of either a “2,2” or “2,4” or “4,4.” Letter from Dr. Word, Cellmark Diagnostics, to Mr. Corbin, Assistant Prosecutor, at 3 (Oct. 18,1994). Therefore, even Cellmark, although originally providing the one-in-1,400 figure, revised its numbers, in response to the criticisms by the two defense, experts. The revised numbers from Cellmark were that one in 170 African-Americans shared the blood type as found at the scene (assuming the blood type was a mixture from two individuals, one being the suspect and one being the victim). Id. at 5 29
*285The State was not content with just using the one-in-170 figure and improperly used the rejected and flawed one-in-1,400 figure. That figure is wrong because it was based on a DQ-Alpha genotype of “4,4” instead of being based on the three DQ-Alpha genotypes that Cellmark admitted could be the suspect’s. By using only the figure for the “4,4” genotype, all the State was calculating was the population that shared Harvey’s DNA markers. The State was not calculating the percentage of the population that shared the DNA markers of the blood found at the scene and identified as the suspect’s. It is this latter figure that is relevant.30
The majority finds the one-in-1,400 figure material because “Mo described, the one-in-1400 evidence was relevant to show that defendant’s relatively rare composite genotype could not be excluded by the PM/DQ Alpha tests as a contributor to the box-spring sample.” Ante at 193, 699 A.2d at 632. The majority’s reasoning is not sound. The fact that defendant had a “relatively rare composite genotype” — a fact not established anywhere in the record — does not inform the jury one bit. Under the majority’s reasoning, if an exceedingly rare genotype is found in only one person in the world and if the sample recovered from the crime scene was determined to contain alleles that could be found in virtually everyone in the world including that person, the State, according to the majority, would be able to highlight how rare that person’s blood is to show that it is somehow more likely that that *286person’s blood contributed to the common sample. The fact that the defendant had a “rare” genotype does not make it more likely that he, rather than someone with a common genotype who also fit the DNA profile, contributed to the sample. The National Resource Council has recognized the prevalence of this type of error and has emphasized that “[u]sually the subgroup to which the suspect belongs is irrelevant, since we want to calculate the probability of a match on the assumption that the suspect is innocent and the evidence DNA was left by someone else.” NRC Report, supra, at 29. “The proper question is,” according to the NRC, “What is the probability that a randomly chosen person, other than the suspect, has the genetic profile of the evidence DNA?” The probability is, by the State’s expert’s own calculations, no more than one in 170.
C.
Even had the State used the correct figures in calculating the number of people who could have contributed to the DNA material found at the scene, the use of the unmodified product rule in making those calculations was erroneous. The use of the unmodified product rule in calculating population frequencies from DNA loci is unsupported by the scientific and legal authorities.
While the product rule is clearly generally accepted as an accurate theoretical proposition, the issue is whether the product rule can be applied to individual DNA frequencies. See Cauthron, supra, 846 P.2d at 514-15 (holding in regard to admission of DNA statistical evidence and product rule calculations that “[t]he expert must show more than the theory. For the evidence to be admitted, the theory must be valid in its application.”); State v. Carter, 246 Neb. 953, 524 N.W.2d 763, 780 (1994) (adopting same approach). A famous example serves to demonstrate why the important issue before us is the product rule’s application to the given situation and not the rule’s theoretical soundness. In People v. Collins, 68 Cal.2d 319, 66 Cal.Rptr. 497, 438 P.2d 33 (1968), the prosecutor applied the product rule to the facts of that *287ease to argue that the odds that the eyewitnesses improperly identified the suspect were one in 12 million. The prosecutor noted that the probabilities of a man with a mustache was only twenty-five percent, the odds of a black man with a beard was only ten percent, the odds of a girl with a ponytail was only ten percent, the odds of a girl with blond hair was thirty-three percent, the odds of seeing a partly yellow automobile was ten percent, and the odds of seeing an interracial couple in a car was only .1 percent. Because the witnesses observed such occurrences and because the defendant was linked to such occurrences, the prosecutor argued that the odds that the defendant was not involved in such occurrences was one in 12 million. Id. at 501, 438 P.2d at 37, cited in Shea, 957 F.Supp. at 336 n. 12. The fallacy in the example is that the characteristics observed were not independent. For product rule calculations to be reliable, admissible, and relevant, the proponent of the evidence must establish the independence of the variables used in the calculation.
Because of the importance of establishing independence of the variables, virtually every one of the 100 or so published cases, including three cases from this State, that have discussed the rule’s application to DNA analysis have included a hearing to examine and test the independence of the variables. These cases are far from unanimous in their conclusions. The reason behind the differing results is simple: A debate has been long raging in the scientific and legal community over the application of the product rule to DNA analysis. See NRC Report, supra, at 25-36, 89-204.
In 1992, the National Research Council issued its first report of the evaluation of forensic DNA evidence. The NRC concluded that the variables — even if based on reliable, random, and sufficiently large samples — were not fully independent. Specifically, the NRC concluded that the variables were neither in “Hardy-Weinberg Equilibrium” nor in “Linkage Equilibrium.” Without getting overly detailed about the scientific principles behind Hardy-Weinberg and Linkage Equilibrium, for our purposes it is *288sufficient to say that because population mating is not completely random, because different subpopulations may have different frequencies of certain alleles, and because of different population structures, the unmodified product rule, which takes into account none of these phenomena, may not be reliable. This was recently made clear by one federal court:
The product rule can be applied reliably ... only if the estimate of allele frequencies is reasonably accurate and the conditions in the population approximate what are known as Hardy-Weinberg equilibrium and linkage equilibrium____ Hardy-Weinberg equilibrium and linkage equilibrium are rarely attained in real populations, most significantly because real populations are finite and contain subgroups that are perpetuated by non-random mating. Accordingly, debate about whether the product rule can be used reliably often focuses on the power of the statistical methods used to detect deviations from Hardy-Weinberg equilibrium and linkage equilibrium and the adequacy of the measures that are used to account for potential deviations.
[Shea, supra, 957 F.Supp. at 336-37.]
Thus, the debate has been over the method of determining the deviations and the size of the deviations from the product rule.
Finding that the two principles of equilibrium were violated and that many of the databases being used were not entirely accurate, the 1992 NRC report recommended the use of the “ceiling principle” or the “interim ceiling principle” in calculating frequencies to be used in courtrooms. These principles essentially provided a lower limit of the overall population frequencies even assuming the violations of the equilibrium rules. See NRC Report, supra, at 35-36, 156-59.
Although some courts came to accept the ceiling principles, the NRC rapidly found itself under attack for advocating their use. Scientists complained that the rules were overly conservative and recommended instead the use of different modified product rules that would not so dilute the power of the population statistics. By last year, the NRC had abandoned advocating for the ceiling principles. Id. at 35. The NRC, however, has continued to reject an unmodified product rule and has instead provided several alternative formulas all of which provide conservative estimates of profile frequencies. The recommended adjustment to the product rule that is relevant to the DNA analysis conducted here is the *289faetor-of-ten adjustment. Id. at 34 (‘We conclude that, when several loci are used, the probability of a coincidental match is very small and that properly calculated match probabilities are correct within a factor of about 10 either way.”); id. at 156 (“[T]he uncertainty of a value calculated from adequate databases (at least several hundred persons) by the product rule is within a factor of about 10 above and below the true value.”). Applying such a modified product rule to the data in this case reveals that as many as one in seventeen people may share the DNA characteristics of the blood found at the scene.
Because of the widespread acceptance of the deficiencies with the product rule in calculating genetic profile frequencies, the vast majority of courts have permitted only modified versions of the product rule to be admitted. See, e.g., Shea, supra, 957 F.Supp. at 341-43 (holding that product rule is only reliable and admissible under Daubert if the rule is modified to account for likely uncertainties caused by population substructuring and random error; allowing product rule when calculated using NRC’s recommended factor of ten estimated value); Lowe, supra, 954 F.Supp. at 407-08 (finding modified product rule, using the factor of ten calculations, to be admissible); State v. Johnson, 186 Ariz. 329, 922 P.2d 294 (1996) (finding modified ceiling principle to be admissible); State v. Bible, 175 Ariz. 549, 858 P.2d 1152, 1181 (1993) (holding product rule in admissible because prosecution failed to establish independence of alleles), cert. denied, 511 U.S. 1046, 114 S.Ct. 1578, 128 L.Ed.2d 221 (1994); State v. Sivri, 231 Conn. 115, 646 A.2d 169 (1994) (remanding for hearing on admissibility of modified product rule); United States v. Porter, 618 A.2d 629, 643 (D.C.App.1992) (finding interim ceiling principle to be admissible, but remanding for additional hearings); Brim, supra, 695 So.2d at 273 (finding ceiling principle, but not unmodified product rule, potentially to be admissible pending a hearing); Caldwell v. State, 260 Ga. 278, 393 S.E.2d 436, 443-44 (1990) (finding unmodified product rule in admissible based on evidence of departure from Hardy-Weinberg equilibrium); Commonwealth v. Lanigan, 413 Mass. 154, 596 N.E.2d 311, 314-16 (1992) (finding product-rule estimates by Cell*290mark to be in admissible; noting in dicta agreement with ceiling principles); Commonwealth v. Curnin, 409 Mass. 218, 565 N.E.2d 440 (1991) (rejecting Cellmark’s probability analysis as not being generally accepted at that time); State v. Bloom, 516 N.W.2d 159 (Minn.1994) (recognizing problems with unmodified product rule and finding product-rule calculation modified by ceiling principle to be admissible); Carter, supra, 524 N.W.2d at 776-83 (ruling that unmodified product rule is not generally accepted and therefore is in admissible based on violation of Hardy-Weinberg proportions); State v. Vandebogart, 139 N.H. 145, 652 A.2d 671 (1994) (finding produet-rule estimate using ceiling principle to be admissible after finding unmodified product-rule calculations to be in admissible); State v. Campbell, 691 A.2d 564 (R.I.1997) (finding that product rule calculated using ceiling principle to be admissible); State v. Morel, 676 A.2d 1347, 1356 (R.I.1996) (permitting modified ceiling principle); State v. Streich, 163 Vt. 331, 658 A.2d 38 (1995) (finding unmodified product rule to be in admissible, and insisting on use of ceiling principles before such evidence can be used); State v. Jones, 130 Wash.2d 302, 922 P.2d 806 (1996) (finding interim ceiling principle to be generally accepted); State v. Copeland, 130 Wash.2d 244, 922 P.2d 1304 (1996) (finding product rule admissible only if independence of alleles is established).
Because the State never established the independence of the loci used in its product rule calculation — indeed, because the State cannot possibly establish the total independence of the loci — the unmodified product rule calculations used here to convict defendant were improperly admitted.
D.
The State’s figures are misleading in another way. The population numbers are based on samples, not on the entire population. Therefore, the figures represent not the frequency with which the genotypes actually occur in the population, but the frequency that, based on the samples, we would expect with a reasonable degree *291of certainty the genotypes might occur. The samples, even by Cellmark’s calculations are not 100% accurate, yet the prosecutor presented them as such.
We do not even need a statistician to inform us that small samples do not represent to a certainty, the actual population. At best, assuming the database was sufficiently large and random, it may be an accurate sampling of the population, thus permitting us, with a reasonable degree of confidence, to predict the actual frequencies within the population as a whole. See State v. Marshall, 130 N.J. 109, 211-12, 613 A.2d 1059 (1992) (Marshall II) (discussing statistical sampling and noting that statisticians would expect the actual results to differ from the sample by standard deviations), cert. denied, 507 U.S. 929, 113 S.Ct. 1306, 122 L.Ed.2d 694 (1993); Harvey I, supra, 121 N.J. at 427, 581 A.2d 483 (recognizing that a confidence interval provides information on the accuracy of an expert’s opinion). • Although the NRC endorses the use of sampling, when match probabilities are estimated from such a database, “such calculations are subject to uncertainties.” NRC Report, supra, at 33. The NRC opined that “[t]he accuracy of the estimate will depend on the genetic model, the actual allele frequencies, and the size of the database.” Ibid.
The databases used here were neither random nor large enough to be an accurate sampling of the population, even to a reasonable degree of certainty. The National Resource Council, perhaps the most impressive advocate for the introduction of DNA population-frequency statistics in criminal cases, concluded that “an adequate database [has] at least several hundred persons.” NRC Report, supra, 33 (emphasis added); see also Lowe, supra, 954 F.Supp. at 409-10 & n. 10 (noting that even the government’s experts testified that an adequate database had at least several hundred persons). Cellmark’s African-American database for the poly-marker markers contained only fifty people and Roche’s just 100 people. Moreover, if the databases are not random, a larger sample size is usually required. We know nothing about Roche’s, but Cellmark’s was certainly not random. Dr. Word admitted *292that the database was based in part on “individuals in our laboratory” and paternity cases Cellmark already had in the laboratory. The group used was obviously “selected,” even though the criteria implicit in that selection are unknown. Although the selective or non-random aspects of a database might sometimes be excused, where as here, the genes measured are either functional or linked with functional genes, see NRC Report, supra, at 118-19, a confidence interval must be provided. It is essential that the population sample be truly random in order to negate the possibility that the results are idiosyncratic. The lack of randomness and the small size inhibits the reliability of the databases.31
Even in our original review of this case we recognized that the burden is on the State to show that the statistical figures it presents to tie defendant to the blood found at the scene must be “considered authoritative in the forensic-serology community.” Harvey I, supra, 121 N.J. at 431, 581 A.2d 483. In fact, the Court held that an “in-house study was an insufficient ground for [the State’s expert] to testify about the [percentages of matches].” Id. at 430, 581 A.2d 483. Despite our warning in this very case that such evidence was a prerequisite to admission, that testimony was clearly lacking.
E.
We can only speculate about exactly how defective the State’s analysis is here because the trial court failed to hold the required hearing on the admissibility of this evidence. The majority disregards our previous teachings that trial courts have an independent *293obligation to ensure the reliability of evidence and to hold a hearing on the admissibility of the evidence when necessary. Indeed, the Court today finds that it cannot review this issue because it strikes from the record what it considers to be the only evidence indicating the need to provide confidence intervals, ante at 200, 699 A.2d at 637, and because it refuses to remand the case for an evidentiary hearing on the matter. Cf. State v. Koedatich, 112 N.J. 225, 283, 548 A.2d 939 (1988) (Koedatich I) (requiring a searching and scrupulous review of the record in capital cases), cert. denied, 488 U.S. 1017, 109 S.Ct. 813, 102 L.Ed.2d 803 (1989); Bey I, supra, 112 N.J. at 93, 548 A.2d 846 (same).
The Court’s holding is this respect once again marks a broad departure from our precedent. For example, in State v. Spann, 130 N.J. 484, 617 A.2d 247 (1993), this Court held that a mathematical formulation, used to determine paternity from blood-grouping tests, must undergo a hearing as to its scientific validity and acceptance prior to the formula being admitted into evidence in a criminal proceeding. This was done even though the formula had gained acceptance in numerous prior civil cases in the State and had been written about extensively. Id. at 505-06, 617 A.2d 247. The Court found that “[w]hat is needed is what the trial court will have: examination and cross-examination on that issue.” Id. at 506, 617 A.2d 247. The trial court was instructed to hold a hearing, determine the expert’s qualifications, and rule on “the conditions attached to the admission of expert testimony, i.e., ‘general acceptance.’ ” Id. at 508, 617 A.2d 247. The testimony is only admissible if the hearing shows the evidence to be reliable and generally accepted and if the jury is told how the expert opinion might be affected by other variables in the case. Id. at 498-99, 617 A.2d 247. The Court concluded that “even if not objected to sufficiently by counsel, the expert’s opinion on probability of paternity did not satisfy the most fundamental requirement of expert testimony: its ability to aid the jury in its *294deliberations.” Id. at 498, 617 A.2d 247 (citing Kelly, supra, 97 N.J. at 209, 478 A.2d 364).
In Kelly, we also recognized that in certain circumstances the prudent course would be to remand the matter to the trial court to take additional testimony about the general acceptance of the scientific evidence. See Kelly, supra, 97 N.J. at 214, 478 A.2d 364 (opting to reverse conviction and order a new trial with a new hearing on the admissibility of the scientific evidence instead of ordering a limited remand, but recognizing that the Court retained the option of ordering “a limited remand ... to the trial court to exercise its discretion, a very broad discretion, on the issue of the expert’s qualifications and the reliability of the knowledge proffered”).
What is remarkable is not only the majority’s disregard for our precedent, but its refusal to recognize that the reasons for requiring an antecedent foundational hearing in this case are much more compelling than in those cases. For example, in Spann, the mathematical formulation had gained acceptance, even in courts in this State. Here, New Jersey courts are split on the validity of the product rule for analyzing DNA results. The product rule has even been rejected in the context of DNA evidence in a capital case in this State. See Cooper DNA Opinion, supra, at 23 (“I further find that the method of statistical calculation used, specifically the product rule method, has not been shown to be generally accepted in the scientific community. Furthermore, the results obtained through use of the product rule are not sufficiently reliable to admit into evidence for consideration by a jury.”). Subsequent to the admission of the evidence here, one New Jersey court has endorsed a modified product rule that takes into account some of the problems noted herein. Marcus, supra, 294 N.J.Super. at 284-87, 683 A.2d 221. More recently, another court, after “a lengthy Frye hearing to determine whether ... the statistical analysis and quantification of the test result was sufficiently reliable to be admitted at trial” found the product rule admissible. Dishon, supra, 297 N.J.Super. at 260, 281-85, 687 A.2d 1074. No court has admitted such evidence without a hearing.32
*295New Jersey has not been alone in requiring mathematical formulations to undergo examination prior to admissibility. Indeed, virtually every court that has permitted similar population database statistics and the product rule formula have required that they be tested for admissibility before the results are admitted into evidence. See cases cited supra at 193-194, 699 A.2d at 632-633. The majority ignores those cases and, in support of its decision to permit this evidence without a full exploration of the many issues surrounding the validity of the product rule in evaluating DNA match probabilities, it confidently asserts that “many other jurisdictions accept the product rule as scientifically reliable.” Ante at 196, 699 A.2d at 634. In fact, most jurisdictions permit only a modified product rule, and the authorities — even ones cited by the majority — are close to being unanimous in ruling that a Frye-type hearing must be held before reaching such a determination. Marcus, supra, 294 N.J.Super. at 288, 683 A.2d 221 (finding no scientific consensus on the use of the product rule and holding that the area “remains a legitimate subject for expert testimony”); Schweitzer, supra, 533 N.W.2d at 158-60; Fishback, supra, 851 P.2d at 893 & n. 18; see also cases cited supra at 194-195, 699 A.2d at 633-634.
Requiring the establishment of reliability prior to the admission of expert testimony is certainly not unique to DNA statistical evidence. In Landrigan v. Celotex Corp., 127 N.J. 404, 605 A.2d 1079 (1992), this Court noted that for any type of expert testimony to be admissible, “(1) the intended testimony must concern a *296subject matter that is beyond the ken of the average juror; (2) the field testified to must be at a state of the art. such that an expert’s testimony could be sufficiently reliable; and (3) the witness must have sufficient expertise to offer the intended testimony.” Id. at 413, 605 A.2d 1079. Under the relevant standards of this State, admission of the product rule and population-frequency data was clearly wrong.
As to the third element, sufficient expertise, this Court has noted that an expert must “be suitably qualified and possessed of sufficient specialized knowledge to be able to express [an expert opinion] and to explain the basis of that opinion.” State v. Moore, 122 N.J. 420, 458-59, 585 A.2d 864 (1991); see State v. Odom, 116 N.J. 65, 71, 560 A.2d 1198 (1989); Hake v. Manchester Tp., 98 N.J. 302, 314, 486 A.2d 836 (1985). When the subject matter of the expert testimony “falls distinctly within the province of a particular profession, the witness should generally be a licensed member of that profession.” State v. Frost, 242 N.J.Super. 601, 615, 577 A.2d 1282 (App.Div.1990), certif. denied, 127 N.J. 321, 604 A.2d 596 (1990). The expert must “possess a demonstrated professional capability to assess the scientific significance of the underlying data and information.” Rubanick, supra, 125 N.J. at 449, 593 A.2d 733. Therefore, in genetic testing where probability statistics are also offered, “[t]he expert should be qualified not only as a geneticist but also as a mathematician.” Spann, supra, 130 N.J. at 519, 617 A.2d 247; cf. Biro v. Prudential Ins. Co., 57 N.J. 204, 271 A.2d 1 (1970) (noting that the danger of wrongly permitting expert testimony is that jurors will give the evidence undue credence because the opinion is offered by an “expert”). In Odom, supra, 116 N.J. 65, 560 A.2d 1198, we emphasized that it was essential that the expertise of the witness coincide with the opinions the expert witness offers.
Dr. Word had no demonstrated ability to delineate the problems with the numbers, nor did she appear to comprehend the inherent unreliability in any statistical sample. Furthermore, she did not provide information from which one could conclude the databases *297she utilized were either random or substantial enough to support her assertions. Indeed, the State provided no information regarding the “total lack of ‘neutrality' on the part of the assumption[s]” the State asked the jurors to make. Spann, supra, 130 N.J. at 499, 617 A.2d 247. The Court ignores this precedent and rehashes the old standard that “the competency of a witness to testify as an expert is an issue remitted to the sound discretion of the trial court.” Ante at 201, 699 A.2d at 637. Sound discretion does not equate with uncritical acceptance.
The error is exacerbated here by the trial court’s rejection of the defense’s challenge to the expert. Defendant pointed out that “[o]ne of the cheeks that the Supreme Court has required [when admitting statistical evidence] is that a mathematician testify. The State has failed to meet that requirement.” Defendant also argued that, “[t]he population frequency statistics cited in the State’s reports are not sufficiently reliable so as to be generally accepted in the scientific community.” Indeed, defense counsel noted that, “[i]n this particular case, the State’s testing laboratory failed to perform the compilation and calculation of the population frequency statistics without error.” These meritorious arguments were rejected by the trial court and are virtually ignored by the majority today.
In permitting the product rule and the population databases to be used without a hearing on the admissibility of the evidence, the Court turns its back on the long line of precedent in this State and others. Worse, the Court chooses today to do so in a case in which the evidence in question is suspect and, worse yet, where a defendant’s life is at stake.
Part Two
I
The trial court erred, understandably, in failing to anticipate our decision in State v. Mejia, 141 N.J. 475, 662 A.2d 308 (1995). In Mejia, we held that capital juries must be instructed that they need not be unanimous in determining whether a defendant intended to kill or merely seriously to injure a victim. Id. at 481, *298662 A.2d 808. Only if the jury in this case unanimously found that defendant’s intent was to kill, could defendant be subject to the death penalty. Gerald, supra, 113 N.J. 40, 549 A.2d 792.33 The issue here is whether the failure to provide the nonunanimity instruction was harmless in that no rational basis existed in the evidence on which a juror could conclude that defendant intended only to commit serious bodily injury. See Harris, supra, 141 N.J. at 549, 662 A.2d 333.
In every other case in which this Court has found such errors to be harmless, see, e.g., Harris, supra, 141 N.J. at 548-51, 662 A.2d 333; State v. Bey, 129 N.J. 557, 576-80, 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. Biegenwald, 126 N.J. 1,16-19, 594 A.2d 172 (1991) (Biegenwald IV); State v. McDougald, 120 N.J. 523, 558-60, 577 A.2d 419 (1990); State v. Hightower, 120 N.J. 378, 412-14, 577 A.2d 99 (1990) (Hightower I); State v. Rose, 120 N.J. 61, 63-64, 576 A.2d 235 (1990) (Rose II), the defendant did not request the instruction that we now find it was error for the trial court not to have given. Unlike in any of those cases, here defendant did request that instruction. The trial court failed to deliver the instruction not because it felt the instruction was unwarranted but rather because it erred in its understanding of the law. In such circumstances, and where the jury is making a life-or-death decision, the plain error standard under which all previous cases have been adjudged is inapplicable. Rather, the Court simply should hold the error harmful.
Virtually dispositive of this issue is that we have already determined that the error committed here was actually harmful. In Harvey I, supra, 121 N.J. 407, 581 A.2d 483, we reversed defendant’s conviction and death sentence because “[t]he record provid*299ed ‘a rational basis for the jury to find that the defendant intended to cause only serious bodily injury.’ ” Id. at 414, 581 A.2d 483 (quoting State v. Coyle, 119 N.J. 194, 209, 574 A.2d 951 (1990)). That conclusion was based on five enumerated factors:
1. The State’s concession that Harvey’s initial intent was to commit burglary, not murder;
2. Harvey’s confession that he struck Schnaps only once and in response to being punched in the nose;
3. The State’s argument during penalty-phase that some of the fifteen blows inflicted upon Schnaps were intended to injure and cause pain, not to kill;
4. The jury’s freedom to reject the pathologist’s testimony when other evidence indicated a lack of murderous intent; and
5. The fact that the trial court opted to instruct the jury on lesser and included offenses which did not require murderous intent.
[Id. at 413-14, 581 A.2d 483.]
The Court unmistakably recognized and stated that defendant’s mental state was “clearly in issue.” Id. at 414, 581 A.2d 483. Remarkably, on evidence that is in all pertinent ways but one identical to what was adduced at the first trial, the Court reaches the extreme opposite conclusion today. Not only does the majority find that defendant’s mental state was no longer “clearly in issue,” but the majority holds that no reasonable juror could possibly conclude that his mental state was in the slightest debatable.
The only difference between the first trial and the retrial is that Harvey’s self-serving confession was not admitted at the retrial. The majority views this as “a critical distinction.” Ante at 150, 699 A.2d at 611. In Harvey I, that factor was not deemed critical. No reason is offered as to why it has become so critical now. In fact, the difference is not particularly important. All of the pertinent facts contained in Harvey’s confession were before the jury in the retrial. Indeed, in the confession, defendant did not assert that he had not intended to murder Irene Schnaps. The majority interprets the confession as containing one fact not before the jury on retrial — that Harvey claims he struck the victim only once. Ante at 150, 699 A.2d at 611. The confession *300does not support the majority’s assertion.34 Moreover, even if Harvey told the police that he hit the victim only once, such a claim was clearly implausible, suggesting, at best, that he did not recall distinctly hitting her more than once or, at worse, that he was lying. The medical examiner testified to numerous skull fractures, a iractured jaw, and a deep laceration on the victim’s skull. The medical examiner opined that Irene Schnaps had been assaulted at least fifteen times with a blunt object. Photos of the victim confirmed his assessment. In sum, the absence of the confession does not change the evidence regarding defendant’s intent to Mil versus his intent to cause serious bodily injury.
The “critical distinction” on wMch the Court now rests its opinion was only one of five pertinent factors the Court then noted. That factor, as just described, was not especially impressive. The other four factors are equally, if not more important. Defendant entered the victim’s apartment without any intent to Mil or even to injure her. Even the State concedes as much. The State theorizes, and the evidence supports such a position, that the victim awoke, perhaps because of the noise defendant was maMng, and found defendant in her bedroom. A struggle ensued. Irene Schnaps’s bedroom was left in disarray. During the struggle, the *301victim injured defendant enough to cause him to bleed. Defendant struck back, administered several blows to cause pain to the victim and presumably to limit her ability to hurt him. Even the State admits that defendant still did not have the intent to kill the victim. In fact, the State argued, and the jury found, that the first blows were intended not to kill, but to injure and cause the victim pain. The State theorizes, and the majority uncritically accepts, that only after the first blows did defendant form an intent to kill; in that instant between blows, he formed a new intent, and proceeded to kill Irene Schnaps. There is no question that the perpetrator killed Schnaps, but certainly a rational person could find that the State did not prove beyond a reasonable doubt that, in the brief moment between blows, defendant’s intent changed from causing pain and disabling injury to causing death.35 A reasonable juror might find that defendant’s intent remained only to cause severe bodily injury.
Certainly that evidence could leave sufficient doubt in one juror’s mind. The evidence was at least “minimally adequate to provide a reasonable basis for the jury to hold a reasonable doubt that the defendant intended to cause death.” Mejia, supra, 141 N.J. at 489, 662 A.2d 308; see also State v. Dixon, 125 N.J. 223, 254, 593 A.2d 266 (1991) (“The error was not harmless because there was evidence in this case that could have sustained an SBI ... verdict. We do not suggest that such a verdict was likely, but merely that if the jury returned that verdict, the court could not reject it.”); State v. Pennington, 119 N.J. 547, 561, 575 A.2d 816 (1990) (noting that this is a “low-threshold”); State v. Pitts, 116 N.J. 580, 615, 562 A.2d 1320 (1989) (same).
The evidence in this ease is not like the evidence in Harris and related cases in which the Court has held that the error was *302harmless. There is no evidence that defendant admitted or stated any intent to kill. Cf. McDougald, supra, 120 N.J. at 558-60, 577 A.2d 419 (noting evidence of defendant’s acknowledgement of intent to kill); State v. DiFrisco, 118 N.J. 253, 571 A.2d 914 (1990) (DiFrisco I) (same), cert. denied, — U.S. -, 116 S.Ct. 949, 133 L.Ed.2d 873 (1996); Pitts, supra, 116 N.J. at 614-20, 562 A.2d 1320 (same); State v. Hunt, 115 N.J. 330, 374-77, 558 A.2d 1259 (1989) (same). The weapon used by defendant, a hammer-like instrument, is not like a gun fired at close range into a person; the victim’s death is not so assured. Cf. Hightower I, supra, 120 N.J. at 412-14, 577 A.2d 99 (finding no basis for a conclusion that defendant only intended serious bodily injury because defendant shot the victim with a .32-caliber handgun from close range in the chest, neck, and head); Rose II, supra, 120 N.J. at 63-64, 576 A.2d 235 (finding no basis for a charge on serious bodily injury where defendant fired a sawed-off shotgun from inches away into victim’s chest). In addition, there is no evidence that defendant took further steps to assure Ms. Schnaps’s death from which one could conclude he possessed an intent to kill. Cf. Hightower I, supra, 120 N.J. at 413, 577 A.2d 99 (noting that defendant dragged victim into freezer); Pitts, supra, 116 N.J. at 618, 562 A.2d 1320 (noting that defendant took victim’s pulse); see also Dixon, supra, 125 N.J. at 254, 593 A.2d 266 (noting that defendant submerged body in creek, but finding that evidence could have sustained serious bodily injury verdict).
Indeed, there is more evidence here that defendant lacked murderous intent than there has been in many of the cases where this Court has found the error reversible. State v. Jackson, 118 N.J. 484, 572 A.2d 607 (1990), is instructive. In Jackson, the defendant stated that he did not enter the apartment of the victim intending to attack or kill her. Rather, on the night of the crime he reacted to a threat by the victim. Id. at 491, 572 A.2d 607. His reaction was repeatedly to stab the victim; he did so fifty-three times, including eighteen in the genital area. Id. at 486, 572 A.2d 607. The similarities with Harvey are striking. Like in Jackson, there was evidence that defendant entered the victim’s *303apartment without any intent to harm her. Like in Jackson, there was evidence that defendant reacted to a threat by the victim; the evidence indicated that Irene Schnaps was even able to draw defendant’s blood. Like Jackson, defendant then proceeded with a vicious and unrelenting attack that left the victim dead.
The majority points to only one bit of evidence in support of its conclusion that this record could not possibly support a reasonable doubt that defendant intended to kill. The majority points to the severity of the wounds. Ante at 151, 699 A.2d at 612. Standing alone, the severity of the wounds never has been sufficient to support that conclusion when the weapon was something other than a gun fired at close range. E.g., Jackson, supra, 118 N.J. at 491, 572 A.2d 607 (“Not every stabbing wound is intended to kill.”).
Another plausible set of circumstances that the majority does not even consider is that even if defendant intended to kill the victim, he formed the intent after he already had killed her. The State asserted that the first blows were meant only to cause pain. The medical examiner also asserted that the victim would have died almost immediately after the blows to the head. He could not say which blow or blows killed her, but he opined that a single blow might have been sufficient. It is logical and possible that defendant kept assaulting the victim even after she fell unconscious and died. In fact, the medical examiner testified that he believed that defendant did exactly that. The medical examiner stated that the victim was probably rendered unconscious, and perhaps died, from the first blow or two. If defendant formed his murderous intent after he had already murdered Irene Schnaps, he would not be eligible for the death penalty. The evidence certainly supports that possibility.
Not only does the majority disregard all evidence contradicting its conclusion, but it even ignores the trial court’s view of the evidence. The trial court obviously found this evidence to be ambiguous and inconclusive. In addition to providing a Gerald charge, as required by our previous decision, the court instructed *304the jurors on the lesser-included offense of aggravated and reckless manslaughter. Those latter offenses would only have been submitted to the jury if the trial court believed that the evidence could support a finding that the defendant had neither the intent to kill nor the intent to cause serious bodily injury. See Dixon, supra, 125 N.J. at 255-56, 593 A.2d 266 (noting that a lesser and included offense must be charged if rationally based on the evidence).
The trial court’s decision to provide a Gerald charge was clearly correct. The majority suggests that the Gerald charge was not required by our previous decision, because in the absence of the confession, no rational basis existed to support it. Ante at 151-152, 699 A.2d at 612 (“In [Harvey I,]... we did not predetermine the need for such a charge in a retrial in which the confession was excluded.”). The majority turns Harvey I on its head. There, we reversed defendant’s conviction and death sentence for failing to deliver the Gerald charge. 121 N.J. at 411, 581 A.2d 483 (noting in the very first paragraph of the opinion that “[bjecause the trial court’s jury instruction ... did not comply with ... Gerald ... we reverse the conviction....”). In dicta, we noted that the confession should not have been admitted. Id. at 425, 581 A.2d 483. Now the majority asserts the later (lesser) error, if corrected, renders the former error harmless. That result defies logic, common sense, and basic fairness.
The majority flagrantly disregards our prior holding in this case, our holdings in similar cases, the facts of this case, and our general requirement that death-penalty cases warrant special consideration. Despite the majority’s contentions, a reasonable juror certainly could have concluded that defendant lacked murderous intent; therefore, the trial court was required to instruct the jury that it need not be unanimous on this point, and the failure to so instruct was plain error.
II
Once again I feel compelled to dissent from the majority’s arbitrary and contradictory finding that the evidence adduced at *305trial could support the c(4)(f) aggravating factor. N.J.S.A. 2C:11-3c(4)(f) (permitting imposition of the death penalty if the murder was “committed for the purpose of escaping detection”); see State v. Loftin, 146 N.J. 295, 398-408, 680 A.2d 677 (1996) (Loftin I) (Handler, J., dissenting); State v. Hightower, 146 N.J. 239, 280-94, 680 A.2d 649 (1996) (Hightower II) (Handler, J., dissenting); Hightower I, supra, 120 N.J. at 436-38, 577 A.2d 99 (Handler, J., dissenting). With this decision, the majority confirms that literally any set of facts can support the c(4)(f) factor and, as construed and applied by the Court, this aggravating factor, rather than narrowing the class of defendants who may be subject to the death penalty, is virtually boundless. The dismissive manner in which this Court addresses the challenge to the factor demonstrates that the majority has become indifferent to the dangers of its expansive interpretation. Additionally, the Court does not attempt to cure the grave error made by the prosecutor when he argued a legally improper, and uncorrected, basis to support the c(4)(f) factor.
A.
The majority’s support for the c(4)(f) factor rests entirely on the fact that there were signs of a disturbance at the crime scene. The evidence on which the majority relies is, in its entirety, that:
Defendant entered Schnaps’s apartment by forcing open a glass patio door. Apparently Schnaps awoke and discovered defendant in her bedroom, which was the only room with signs of disturbance. The jury reasonably could have inferred that defendant decided to kill Schnaps to prevent her from alerting her neighbors, calling the police, and later identifying defendant as the person who intended to rob her. Indeed, when questioned by police, neighbors in the apartment complex denied hearing any noises from Schnaps’s apartment.
[Ante at 225-226, 699 A.2d at 649 (internal citations omitted).]
It is not comprehensible how any juror could conclude that such scant evidence proves, beyond a reasonable doubt, that defendant killed Irene Schnaps in order to avoid apprehension and future prosecution for the robbery. Silence in the evidence is never enlightening, no less dispositive. The majority converts silence, mute circumstances, into powerful and deadly proof.
*306The majority now holds that signs of a struggle support the c(4)(f) factor. Last year, the majority held that the absence of signs of a struggle established the c(4)(f) factor. Loftin I, supra, 146 N.J. at 378, 680 A.2d 677 (noting that with no signs of a struggle, no other motive could explain why the killing took place). Perhaps one should not be surprised by this latest flagrant contradiction; it is just another in a long string of inconsistent and conflicting rulings regarding the c(4)(f) factor. For example, in Loftin I, supra, 146 N.J. at 377-78, 680 A.2d 677, the Court held that the wearing of a mask is evidence supporting the c(4)(f) factor, and in Martini I, supra, 131 N.J. at 283, 619 A.2d 1208, the Court noted that disguising oneself is evidence supporting the c(4)(f) factor, but in Harris, supra, 141 N.J. at 535-36, 662 A.2d 333, the Court found that the wearing of a mask and then the removal of the mask is evidence supporting the e(4)(f) factor. In Loftin I, supra, 146 N.J. at 377, 680 A.2d 677, the Court held that planning a crime where defendant would not be known or recognized supports the c(4)(f) factor, but in Martini I, supra, 131 N.J. at 283, 619 A.2d 1208, the Court held that a victim’s knowledge of defendant supports the e(4)(f) factor. In Loftin I, supra, 146 N.J. at 377, 680 A.2d 677, the Court ruled that committing an offense miles from one’s home is evidence supporting the c(4)(f) factor, but in Hightower II, supra, 146 N.J. at 268, 680 A.2d 649, the Court found that committing an offense near one’s home is evidence supporting the e(4)(f) factor. Lastly, in Loftin I, supra, 146 N.J. at 318-19, 680 A.2d 677, the Court indicated that a robbery at nighttime is evidence supporting the c(4)(f) factor, yet in Hightower II, supra, 146 N.J. at 268, 680 A.2d 649, the Court noted that robbing in daylight is evidence supporting the c(4)(f) factor. It becomes impossible not to conclude under the Court’s analyses, that any murder — and every murder — is committed in order to avoid detection and apprehension.
That trail of contradictory opinions is marked not to underscore the Court’s aimlessness, but to show that the majority has imposed no limitations on the aggravating factor; its failure will permit prosecutors to charge the c(4)(f) factor based on any set of *307facts. That is limitless discretion, opening the door to arbitrary and capricious sentences, and it is unconstitutional. “To pass constitutional muster, a capital sentencing scheme must ‘genuinely narrow the class of persons eligible for the death penalty and must reasonably justify the imposition of a more severe sentence on the defendant compared to others found guilty of murder.’ ” Lowenfield v. Phelps, 484 U.S. 231, 244, 108 S.Ct. 546, 554, 98 L.Ed.2d 568, 581 (1988). Aggravating factors are the key element to narrowing the class of death-eligible defendants. See Gregg v. Georgia, 428 U.S. 153, 194-95, 96 S.Ct. 2909, 2935, 49 L.Ed.2d 859, 886-87 (1976) (“[Aggravating factors] provide guidance to the sentencing authority and thereby reduce the likelihood that it will impose a sentence that fairly can be called capricious or arbitrary. Where the sentencing authority is required to specify the factors it relied upon in reaching its decision, the further safeguard of meaningful appellate review is available to ensure that death sentences are not imposed capriciously or in a freakish manner.”); State v. Ramseur, 106 N.J. 123, 182-201, 524 A.2d 188 (1987). Aggravating factors that fail to narrow the class of death-eligible persons render death sentences under their provisions unconstitutional. Cf. Gregg, supra, 428 U.S. at 201, 96 S.Ct. at 2938, 49 L.Ed.2d at 890 (finding that although the “outrageously or wantonly vile” aggravating factor can be construed to serve no limiting purpose, the Georgia death-penalty scheme was not unconstitutional because no evidence existed that the state supreme court would not sufficiently limit the scope of the factor); Ramseur, supra, 106 N.J. at 197-201, 524 A.2d 188 (noting the Court’s “power and obligation to narrow imprecise statutory language in order to render [the wantonly vile, horrible, or inhuman aggravating factor] constitutional”). The Court has failed to abide by those vital mandates rendering “imprecise statutory language” even more imprecise and utterly without merit.
B.
In addition to the lack of probative evidence supporting the c(4)(f) factor proffered by the majority, the only argument ad*308vanced by the prosecutor during summation in support of the factor was improper. The evidence to which the prosecutor pointed did not support the theory that defendant was motivated in committing the murder by an intent to avoid apprehension. Despite the fact that the argument went uneorrected by the trial court, the majority finds the error to be harmless.
The prosecutor argued that the e(4)(f) factor was supported by the fact that defendant concealed and destroyed evidence after the commission of the murder:
Aggravating factor [c(4)(f) ] is [“]was this murder additionally committed in an attempt to avoid prosecution, apprehension?[”] Yes, it was. Items were removed not only to be stolen and kept by Mr. Harvey, but items were removed from this apartment to ... prevent detection. I suggest to you bedding. The body of Irene Schnappes [sic] was washed clean in an attempt to prevent the police from locating the individual that committed this offense, Mr. Harvey.
For evidence to support the c(4)(f) factor, it must indicate that the murder itself was “intended to eliminate a potential witness to the crime.” Martini I, supra, 131 N.J. at 281, 619 A.2d 1208. Each of the facts argued by the prosecutor related to defendant’s actions taken after the murder and do not support the conclusion that the murder itself was committed to conceal the robbery offense. New Jersey courts repeatedly have held that the actions of a defendant to conceal the murder are not admissible to support the c(4)(f) factor. Hightower I, supra, 120 N.J. at 422, 577 A.2d 99; State v. Monturi, 195 N.J.Super. 317, 326-27, 478 A.2d 1266 (Law Div.1984). The majority correctly notes, as it must, that the prosecutor’s argument was “improper.” Ante at 223, 699 A.2d at 648.
The majority, however, seriously misapprehends the record when it finds that the prosecutor’s impermissible representations were cured by the trial court’s instruction. The trial court provided no curative instruction. The trial court’s concluding instructions, although correctly stating the law, were not made in response to the prosecutor’s improper remarks and in no way neutralized or cancelled them: the court did not inform or even intimate to the jury that the prosecutor’s remarks were erroneous *309and were to be disregarded. Rather, as part of a lengthy charge, the court simply observed to the jurors that “[a]ny evidence of action taken by the defendant to conceal the murder itself cannot be used to prove this aggravating factor.”
Although in some cases an instruction by the trial court may cure the prejudice done by a prosecutor’s improper statement, see Koedatich I, supra, 112 N.J. at 320-23, 548 A.2d 939, a statement of the law, even if correct, made in the final instructions to the jury without any reference to or acknowledgement of the improper remarks is not sufficient. A curative charge should be delivered immediately after the objectionable action and should directly address the error. See State v. Zola, 112 N.J. 384, 426, 548 A.2d 1022 (1988), cert. denied, 489 U.S. 1022, 109 S.Ct. 1146, 103 L.Ed.2d 205 (1989); Koedatich I, supra, 112 N.J. at 324-25, 548 A.2d 939. The jurors were never told to disregard the improper remarks and the comments never were withdrawn. The Court now assumes that the jury fully understood that the prosecutor’s remarks were wrong and were not to be taken into account, when the trial court, defense counsel, and the prosecutor all failed to appreciate that fact. That is a drastic assumption in view of the stakes — a death sentence that turns solely on whether the jury would be persuaded by the prosecutor’s argument. In light of both the absence of substantive evidence supporting the c(4)(f) factor and the prosecutor’s improper and gravely misleading argument regarding evidence of the sole aggravating factor, this Court must infer that the jury was influenced by the prosecutor’s remarks, and is obligated to find the error harmful.
Ill
I also dissent from the Court’s ruling on several other points: that the incomprehensible felony-murder charge was harmless error, ante at 152-154, 699 A.2d at 612-613; that the erroneous jury charge permitting the penalty jury to consider all guilt phase evidence was harmless error, ante at 223-225, 699 A.2d at 647-648; that the Court’s finding that the prosecutor’s reference to *310“pubic hair” was harmless error, ante at 215-217, 699 A.2d at 643-645; that permitting the penalty jury to be misled into believing that defendant would be eligible for parole was proper, ante at 227-228, 699 A.2d at 649-650; and that the aggregation of so many “harmless errors” did not have the effect of denying defendant a fair trial, ante at 229, 699 A.2d at 650.
A.
The verdict sheet, which permitted the jury to consider the felony murder count only if it first found defendant guilty of death-eligible murder, was utterly incomprehensible and plainly and starkly wrong. The majority finds error, but holds the error harmless. I disagree; such an error can never be harmless.
In State v. Purnell, 126 N.J. 518, 601 A.2d 175 (1992), this Court vacated a death sentence because the jurors had not been given the opportunity to convict defendant of the noncapital offense of felony murder. The Court noted that:
We have consistently held that all forms of homicide rationally supported by the evidence, whether they be lesser-included or alternative offenses, should be placed before the jury. To truncate the definitions of the murder statute and thus deny a jury the mechanism to decide which of the forms of murder has been proven is unacceptable.
[Id. at 530-31, 601 A.2d 175.]
That principle has been repeatedly reiterated by this Court. See Mejia, supra, 141 N.J. at 484, 662 A.2d 308 (noting that a capital “jury must be given every opportunity to convict of the charge not carrying the death penalty” and disapproving of a sequential charge) (quoting John M. Cannel, New Jersey Criminal Code, Annotated, comment 13 to N.J.S.A. 2C:1-8(e) (1994)); Dixon, supra, 125 N.J. at 256, 593 A.2d 266 (holding that the felony-murder charge must be submitted to the jury in a capital case where the evidence clearly indicates the appropriateness of the charge).
The Court today finds that the error did not constitute plain error because “[defendant has not advanced any plausible version *311in which the jury could have convicted him of felony murder without also convicting him of purposeful-or-knowing murder.” Ante at 154, 699 A.2d at 613. The majority poses the incorrect question. The issue, as this Court underscored in Purnell, supra, 126 N.J. at 532, 601 A.2d 175, is whether the evidence could support a felony-murder charge. As this Court previously has recognized in the context of a murder committed in the course of a robbery, it clearly could:
By relying on the robbery as an aggravating factor, the State necessarily affirmed that proofs existed that provided a rational basis for the jury to choose the death-ineligible option of finding defendant guilty of felony murder. To deprive a capital defendant of a lesser-included alternative murder charge, which arguably would have affected the deliberation of a death sentence, is not constitutionally permissible.
[Ibid.]
The Court acknowledges that “the combination of the jury charge and the verdict sheet led the jury away from rendering a noncapital verdict of felony murder,” ante at 154, 699 A.2d at 613. To conclude that the error was harmless is insupportable.
The Court’s conclusion is even more indefensible in light of the facts of Purnell itself. In that case, the defendant stabbed the victim fifteen times in the neck, chest, and abdomen. Purnell, supra, 126 N.J. at 528, 601 A.2d 175. Despite the extreme nature of those facts — significantly more extreme and indicative of intent to kill than the facts here — and despite the implausibility of an acquittal of purposeful-or-knowing murder and a conviction of felony murder, we reversed the murder conviction because of the failure to charge felony murder. How the same error that we deemed reversible error in Purnell is harmless here defies explanation. The Court should stand by its holding in Purnell and not implicitly overrule it by denying that it controls in the circumstances of this case and establishes plain error.
B.
The trial court incorrectly told the penalty-phase jurors that they could consider “all the evidence presented” in both the *312penalty and guilt phases of the trial in determining whether the State had proven the aggravating factors. The majority recognizes that charge was erroneous, but it finds the error to be harmless. I disagree. The charge cannot be considered harmless at least in respect of the c(4)(f) aggravating factor.
Although this Court has long held that trial courts must inform jurors how and to what extent they may use evidence in their penalty deliberations, see State v. Erazo, 126 N.J. 112, 133, 594 A.2d 232 (1991); State v. Bey, 112 N.J. 123, 183, 548 A.2d 887 (1988) (Bey II); see also Bench Manual for Capital Causes, at 231-32 (Nov. 1, 1996), the trial court here failed to fulfill that obligation. Rather, the court instructed the jurors that:
[T]he evidence to be considered by you includes that material presented at both phases of the trial, all the witnesses and all the physical exhibits.
The court then reiterated that instruction:
I want to repeat that. The evidence to be considered by you includes the material presented at both phases of the trial, all the witnesses and all physical exhibits.
The majority acknowledges that those instructions were incorrect. Ante at 224, 699 A.2d at 648. Yet again, though the majority finds the error harmless, this time because, the majority notes, the prosecutor properly directed the jurors attention to the relevant evidence.
Certainly, this error cannot be viewed as harmless in relation to the c(4)(f) factor. As previously pointed out, the prosecutor argued the wrong evidence. See discussion supra at 210-212, 699 A.2d at 641-642. In dismissing the error, the majority attempts to rely on the trial court’s instruction to the jury that “[a]ny evidence of actions taken by the defendant to conceal the murder itself cannot be used to prove this aggravating factor.” Ante at 226, 699 A.2d at 649. That comment in no way rectified the error. Nor can the majority point to where the court directed the jurors to what evidence they could properly consider. In fact, the only evidence the jurors were told they could consider in support of the e(4)(f) factor was that noted by the prosecutor. That evidence, concerning defendant’s attempts to cover up the murder, however, *313could not be used to support the factor. Supra at 210-212, 699 A.2d at 641-642. Despite that fact, the majority holds that “[t]he court’s instructions, when viewed against the background of the prosecutor’s summation, sufficiently informed the jury of the evidence concerning the aggravating factors.” Ante at 229, 699 A.2d at 651. When the improper charge is “viewed against the background of the prosecutor’s [improper and misleading] summation,” one cannot rationally conclude that the error in the charge was harmless.
C.
In Harvey I, supra, 121 N.J. at 407, 581 A.2d 483, this Court took pains to advise the prosecutor that any reference at retrial to the hair found at the crime scene should not include the information that the hair was a pubic hair. The Court observed that “[w]hether it might have come from his head, his chest, or his pubis is irrelevant in the absence of allegations of a sexual encounter. On retrial the prosecution should refrain from referring to the hair as a ‘pubic hair.’ ” Ibid. Despite that warning, the prosecutor elicited comments from his investigator that “a pubic hair control” was taken from the victim, and the investigator testified that he had also obtained a “[p]ubie hair sample from the sus — ’’before he was interrupted. Although the Court claims that “[t]he State never characterized the hair found in Schnaps’s apartment as a public hair,” ante at 220, 699 A.2d at 646, I credit the jury with being able to draw the logical inference that if the State obtained public hair samples from the victim and the suspect, then the State intended to use those samples in comparison with a pubic hair found at or near the scene. In addition, the investigator noted, and the prosecutor highlighted that probability in a diagram, that the police recovered as evidence from the scene a “pair of light gray panties inside out with stains.” Lastly, the investigator testified that “oral, vaginal, and anal swabs [were] taken from the victim at the time of the autopsy.”
*314Those innuendos of sexual assault had no place in this trial. There was no allegation of sexual assault and certainly no such charge against defendant. The hints of sexual assault that the prosecutor purposefully injected into this capital-murder trial were not only erroneous, but because of our prior ruling, bordered on contemptuous. The prosecutor’s intent can be gleaned from the diagram he prepared and used at trial. The diagram included only eight of more than forty items seized as evidence. One of the eight items highlighted was the panties. The other items — the empty watch box, the open pocketbook, the empty jewelry case— had clear relevance, while the panties were at best marginally relevant and substantially more inflammatory.
Although, normally, I would be content to find that such an error was harmless, especially because of the trial court’s instructions and the dearth of objections from defense counsel, I cannot find so here because we already have warned the prosecutor in this very same case about this same evidence.
D.
I continue to maintain that capital juries should be informed of the actual period of parole ineligibility that defendants face if they are sentenced to life imprisonment instead of death. See Loftin I, supra, 146 N.J. at 420-31, 680 A.2d 677 (Handler, J., dissenting); see Simmons v. South Carolina, 512 U.S. 154, 114 S.Ct. 2187, 129 L.Ed.2d 133 (1994). The jury here was erroneously informed that defendant would be eligible for parole in 2014. The prosecutor even asked the jurors whether they felt that was “sufficient punishment.”
Prior to the imposition of the sentence here, defendant had accumulated an aggregate sentence of sixty-five years with a thirty-two-and-a-half year parole bar for several unrelated offense. Furthermore, defendant’s prior sentence would be extended because of his violation of parole. There was no way that defendant was going to be paroled in 2014 or anytime soon thereafter.
*315The majority finds that there was no error in falsely informing the jury that defendant could be paroled in 2014. The majority reasons that the prosecutor was unable to discuss defendant’s true prospects for parole because he could not inform the jury of defendant’s prior convictions. Ante at 231, 699 A.2d at 652. Although it would be impermissible to inform the jury of defendant’s prior convictions, informing the jury that defendant would die in prison long before he even had the opportunity for parole would not tell the jury that defendant had those convictions.
Our capital sentencing scheme is premised on the principle that juries must be informed of the legal effect of their findings. Mejia, supra, 141 N.J. at 485, 662 A.2d 308 (citing Bey II, supra, 112 N.J. at 164-435, 548 A.2d 887). Misleading the jury on material points in a way detrimental to the defendant cannot satisfy the constitutional demands of a valid capital-punishment scheme. The jury should have been told that defendant would never be eligible for parole.
Conclusion
I have little doubt that when the time comes, this case will eventually be reversed either by this Court or a federal court. The errors are so grave and the efforts of defense counsel so futile that an ineffective assistance of counsel claim must líe. I have reluctantly highlighted only some of the errors here. I fear that someday some court will use this dissent as evidence that those issues were raised either below or before us and were addressed by the Court. The Court’s consideration and treatment of the issues, however, is too lacking and meager to justify that conclusion. Because of the ineffective efforts of both trial and appellate defense counsel, most of the issues were not in fact dealt with. However, a scrupulous review of the record may reveal a few of them, even if our discussion is impaired by the failure of the adversarial process. Fortunately, the majority at least recognizes the numerous deficiencies in the record and questions raised by this DNA and statistical evidence and it preserves those issue for post-conviction relief.
*316The Court, however, is wrong to brush aside the errors and to postpone the inevitable. Nathaniel Harvey cannot be executed on the basis of this record. Despite the length of the majority’s opinion, its analysis is unbalanced and uncritical. The Court appears caring and conscientious because it expends so much energy in its appellate review, but, in actuality, that effort is misspent to confirm an insupportable death sentence. In result, the opinion is not principled and should not be the means to seal the death of this defendant. I dissent.
Justice O’HERN, concurs in the opinion and judgment of the Court except with respect to Part II thereof. He would therefore affirm the convictions except insofar as the conviction of murder establishes death eligibility. He joins Part Two, Section I of Justice Handler’s opinion on the issue of a non-unanimous verdict.
For affirmance — Justices POLLOCK, GARIBALDI, STEIN and COLEMAN — 4.
For affirmance in part; for reversal in part — Justice O’HERN — 1.
For reversal — Justice HANDLER — 1.
The six types of DQ-alpha alleles recognized at the time of this trial are known as: 1.1, 1.2, 1.3, 2, 3, and 4. Thus, the 21 genotypes include: (a) 1.1 paired with 1.1; (b) 1.1 with 1.2; (c) 1.1 with 1.3; (d) 1.1 with 2; (e) 1.1 with 3; (f) 1.1 with 4; (g) 1.2 with 1.2; (h) 1.2 with 1.3; (i) 1.2 with 2; etc.
The polymarker alleles are denoted by letters. For example, the LDLR polymarker, has two possible alleles (A and B), and three possible genotypes (A,A; A,B; and B,B). The HBGG polymarker has three alleles (A, B, and C), and thus has six possible genotypes (A,A; A,B; A,C; B,B; B,C; and C,C).
The State's expert called this technique "association of alleles.” In the scientific literature, however, association of alleles is a phrase used to describe something else, namely, the interaction among or dependence between different alleles. See, e.g., Bruce Budowle, Jenifer Lindsey, Jacqueline DeCou, Barbara Koons, Alan Giusti, & Catherine Comey, Validation and Population Studies of the Loci LDLR, GYPA, HBGG, D7S8, and Gc (PM loci), and HLA-DQa Using a Multiplex Amplification and Typing Procedure, 40 J. Forensic Sci. 45, 53 (Jan. 1995) ("F.B.I. Study ”). The more accepted terminology is "dot-intensity analysis.”
A fourth sample was taken from a piece of bloodied cardboard found at the scene, but this last sample was too degraded to be analyzed and therefore was not introduced in evidence.
The column labeled "Suspect's Expected” refers to the genotype or genotypes one would expect the suspect to have if we assume the mixed sample contains both (and only) Ms. Schnaps’s blood and the suspect's blood. The column labeled "Susp.’s w/ Dot I.” refers to the genotype one would expect the suspect to have if we make the same assumptions (and some others) and if we use dot-intensity analysis. The column labelled "Dot Intensity" refers to the comparisons that Cellmark made and the results that Cellmark obtained from the testing strips in comparing the relative intensities.
The Lowe court did not decide “who wins th[e] semantic debate" over whether the change in technique for reviewing autorads must independently be found generally accepted, but instead ruled that an examination of the Daubert [v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993)] factors to ensure reliability was necessary. Lowe, 954 F.Supp. at 411.
The Fishback court went so far as to demand that the statistical frequency analysis performed in the case be subject to the test of general acceptance for admissibility. Id. at 893 & n. 18. Even that was not done here. See discussion infra part one, section IV.
Although the Court cites Schweitzer, it refuses to measure the statistical evidence at issue here against the same standard the court applied in Schweitzer. See discussion infra at part one, section IV.C. Specifically, in Schweitzer, the court recognized that even a statistical expert's opinion must be found to be reliable, relevant, and based on scientifically valid principles before such evidence could be admitted. 533 N.W.2d at 159.
Although the "general acceptance" test still governs the admission of scientific evidence in this situation, I am not convinced that such a high standard would justify the exclusion of similar evidence proffered by a defendant in a capital prosecution.
The notion of different standards for admission by the prosecution and the defense is common in criminal law in general and death-penalty jurisprudence in particular. For example, the State has the burden of proving guilt beyond a reasonable doubt, and yet no burden can be shifted to the defense. In capital cases, the State must prove each aggravating factor beyond a reasonable doubt to the satisfaction of each juror before any one juror can consider the factor. In contrast, mitigating factors need only be found by one juror. Similarly, a capital defendant may offer "without regard to the rules governing the admission of evidence at criminal trials,” evidence on any mitigating factor, N.J.S.A. 2C:11-3c(2)(b), yet the State is more restricted in what evidence it may introduce. See *254State v. Gerald, 113 N.J. 40, 103, 549 A.2d 792 (1988); State v. Davis, 96 N.J. 611, 621-23, 477 A.2d 308 (1984).
Nor is a lower hurdle for admission of scientific evidence that exculpates a defendant a novel idea. This Court previously has suggested such a possibility. In Windmere, supra, 105 N.J. 373, 522 A.2d 405, the Court noted that voiceprint identification, while not admissible when offered by the State in a criminal trial as evidence of guilt, "is at least a reliable method to eliminate a person as the unidentified speaker." Id. at 383, 522 A.2d 405; see also id. at 383 n. 5, 522 A.2d 405 ("Similarly, in State v. Prudden, 212 N.J.Super. 608, 617, 515 A.2d 1260 (App.Div.1986), the court, while rejecting the use of footprint techniques to identify a defendant, recognized that the techniques were nevertheless sufficiently reliable to exclude a person as the guilty party."). See generally Davis, supra, 96 N.J. at 621, 477 A.2d 308 ("We have recognized that standards of proof may vary depending upon the litigational context.") (citing Romano, supra, 96 N.J. 66, 474 A.2d 1; In re Polk License Revocation, 90 N.J. 550, 449 A.2d 7 (1982)); State v. Millett, 272 N.J.Super. 68, 98-99, 639 A.2d 352 (App.Div.1994) (recognizing that a criminal defendant may present evidence "if the proof offered has a rational tendency to engender a reasonable doubt with respect to an essential feature of the State's case") (citing State v. Sturdivant, 31 N.J. 165, 179, 155 A.2d 771 (1959), cert. denied, 362 U.S. 956, 80 S.Ct. 873, 4 L.Ed.2d 873 (1960)); David McCord, But Perry Mason Made It Look So Easy!: The Admissibility of Evidence Offered by a Criminal Defendant to Suggest that Someone Else Is Guilty, 63 Tenn. L.Rev. 917 (1996) (advocating lower standard of admissibility in such circumstances).
An additional problem arises as a result of the subjectivity of the interpretations. Because the dot intensity must be viewed immediately after the completion of the testing and while the test strips are still wet and because the strips deteriorate quickly, the results cannot be preserved or even fully captured by photos, thus effectively rendering the subjective interpretations immune from critical evaluation.
In a similar vein, the first New Jersey case admitting PCR testing relied on the testimony, expertise, and testing procedures of Dr. Edward Blake. Williams, supra, 252 N.J.Super. at 381, 599 A.2d 960; see also Dishon, supra, 297 N.J.Super. at 278, 687 A.2d 1074 (discussing PCR testing). Dr. Blake was a consultant for the defense in this case. He resoundingly criticized the State's analysis.
The F.B.I. was unable to determine the percentage of test runs afdicted with unbalanced dots "because the determination of unbalanced intensity is somewhat subjective.” Ibid.
The six possible combinations that would involve all three found alleles are: (1) "AA, BC”; (2) "AB, BC”; (3) "AB, AC”; (4) "AB, CC”; (5) "AC, BC”; (6) "AC, BB.” Genotype pairs (1) and (3) have more A alleles than B or C alleles. Genotype pairs (2) and (6) have more B alleles than A or C alleles. Genotype pairs (4) and (5) have more C alleles than A or B alleles. In none of the possible combinations are the alleles present in roughly equal amounts.
The fact that Dr. Word testified that "the possible combinations could be only AA and BC, AB and CC, or AC and BB,” ante at 166; 699 A.2d at 619, raises serious questions abut the accuracy of her testimony. No one can seriously doubt that there exist six — not three — possible combinations.
Instead, the Court reserves discussion of the issue for post-conviction relief proceedings. Ante at 180-182, 699 A.2d at 626-627.
Although completely outside the record in this case, it is worth noting that in a different capital trial in this State, a trial court refused to admit evidence of a polymarker test because the defendant had shown that variances occur not only at the GYPA loci, but also at the GC loci. See State v. David Cooper, No. 93-10-01627 (trial court's unpublished opinion captioned "Admissibility of DNA Evidence”), at 19-20. No such evidence was presented here.
The Court, although recognizing that defendant argued that "variances ... destroyed the integrity of the dot-intensity analysis,” ante at 178, 699 A.2d at 625, never addresses the argument. Indeed, the Court merely reaffirms that variances, not laboratory errors, account for the anomalous dot intensities. Ante at 179, 699 A.2d at 626. Precisely because the strange results cannot be accounted for by laboratory error is why they call into doubt the reliability of dot-intensity analysis.
Each PM testing strip has a control dot that helps the tester to tell if the PCR enhancing and the PM test itself were performed correctly. For readings to be valid, the dot intensity of the alleles present must be brighter than the dot intensity of the control dot.
Dot imbalances cannot be compared when the genotype being examined has two of the same alleles (i.e., "homozygous," for example, the "B,B” genotype) because only one dot on the strip will turn blue. Therefore, out of the five PM markers that both defendant and the victim had (for a total of ten markers), four were homozygous and could not be compared.
We can check for improper imbalances only in the samples containing a single person's DNA. In the mixed sample, the State assumes that all imbalances are due not to variances and the like, but to different contributors to the mixture. As previously noted, one of these, the HBGG marker, can be examined, because logic tells us that if two pairs of alleles make up the mixture, and three 1ypes of alleles are found present, the three types of alleles present cannot be in equal proportion. One of the three types of alleles must be present in twice the strength as the other two 1ypes of alleles. The results, however, showed that the dot intensities for all of the alleles were balanced. This is yet another unexplainable result.
No evidence was introduced about the relative decomposition rates for the different allele dots. While different degeneration rates may explain the contradictory readings, they would render dot-intensity analysis even more unreliable.
The subjective nature of dot-intensity analysis is particularly troubling here, where the laboratory performing the tests knew what results it wanted to obtain from this novel procedure prior to its testing. Cellmark, even before the performance of dot-intensity analysis, knew defendant’s genotype and set out to *270establish that his blood was a contributor to the sample. Cellmark's partiality can be seen in its original calculation of the odds that a person would have the same markers as those recovered from the sample. Cellmark used defendant’s markers in its original calculations even though its own testing did not establish that those particular markers (specifically, the DQ-Alpha one) were the only ones that could have contributed to the sample. A far better procedure would have been to conduct dot-intensity analysis first or to use different testers for determining defendant's genotypes and for performing dot-intensity analysis.
The "only two contributors" assumption is also cast in doubt by the findings of the State regarding faint dots on the test strips. Those faint dots, according to the State's expert, indicate other contributors. To the extent that the faint dots are erroneous readings due either to manufacturing defects in the testing kits or performance errors by the technicians, the faint dots cast doubt on the validity of the dot-intensity analysis.
Without dot-intensity analysis, the contributor to the blood mixture could have had any LDLR genotype (100% of the population); two of three GYPA genotypes (approximately 77% of African Americans); the "A,C” genotype for the HBGG marker (approximately 27% to 32% of African Americans); two of three possible D7S8 genotypes (approximately 66% of African Americans); and three possible GC genotypes (approximately 89% of African Americans). F.B.I. Study, supra, at 50 (table 3). Using the product rale (and assuming the accuracy of the rule, assuming the accuracy of the numbers, and assuming no margin for error — all flawed assumptions, see discussion infra part one, section IV), 15% of African Americans share those markers.
In fact, Cellmark did not stand by the technique with any sort of validation study and Dr. Word, the State's own expert, testified pretrial that the technique is only "sometimes possible" even when there are only two contributors to the mixture and is impossible when there are more than two contributors. She neglected to mention in what circumstances such analysis is possible. See discussion supra at 161-163, 699 A.2d at 617.
Additionally, this error might have prevented the penalty-phase jury's formation of "residual doubt" and thus influenced the death sentence meted out. See State v. Marshall, 148 N.J. 89, 172, 690 A.2d 1 (1997) {Marshall III) (recognizing concept of "residual doubt" and its potential impact on penalty-phase delibera*281tions); State v. Harris, 141 N.J. 525, 580, 662 A.2d 333 (1995) (Handler, J., concurring in part and dissenting in part) (noting jurisdictions that recognize "residual doubt").
26 The prosecutor had previously explained this sample as having been the mixture taken from the boxspring at the crime scene. The prosecutor's error in calculation relates back to his earlier error in stating that Cellmark's testing of the boxspring sample revealed that the suspect had a "4,4” DQ Alpha genotype:
"Tested the boxspring matress [sic] and came up with DNA that [Ms. Cooper at Cellmark Diagnostics] believed was a mixture from two different sources.
Initially they do — I put the chart up here Ms. Cooper used, DQ Alpha strip and come back with a type of four, four and I marked that up here, four, four.
She indicated the percentage of people in the population, the entire population, that have that four, four, is 17 percent.
So you can exclude 83 percent from the entire population of the world just by that DQ Alpha.”
In fact, the testing revealed that the suspect had either a "4,4" genotype, a "2,4" genotype, or a "2,2” genotype. One could not "exclude 83 percent” of the population from having contributed to the sample.
The majority incorrectly states that "[t]he prosecution did not present the one-in-1400 figure to prove the percentage of African Americans whose genetic compositions could be comparable with the PM/DQ Alpha profile of the box-spring sample.” Ante at 192, 699 A.2d at 632. Even if the majority were correct, which the above-quoted statement by the prosecutor flatly shows not to be the case, the one in 1,400 figure would be irrelevant. See infra at 285-286, 699 A.2d at 679.
At trial, Cellmark actually used the figure of 17%. That figure represents the presence of this genotype within the entire population. Remarkably, the State never mentioned the 11.9% figure at trial except during its calculation of the one-in-1,400 number. Arguably, using the lower percentage was incorrect. See Cooper DNA Opinion, supra, at 6-8 (recognizing danger of subgrouping and requiring use of "ceiling principle" under which frequency of alleles is capped at *284maximum rate of allele's presence in any single subgroup); see also NRC Report, supra, at 35-36 (discussing ceiling principles).
The majority suggests that the correct figure is somewhat above the approximately one in 666 that the defendant calculated could be the appropriate figure instead of the one in 1,400. Ante at 192, 699 A.2d at 632. The majority fails to realize that the defense was not advocating the one-in-666 figure but was rather using it to demonstrate the fallacy of the State's figure. In any event, if the one-in-666 figure is misleading in any way, it is misleading in a manner that helps the State. The defense's calculation did not even include the percentage of individuals who share the "2,2" DQ-Alpha genotype — a percentage that, even by *285the majority's analysis, should have been included. No evidence was adduced at trial concerning what that figure was. The Court's conclusion that "[i]t is too late in the proceeding for defendant to insist that the State should have presented statistical evidence that defendant now believes would have been helpful at trial," ante at 193, 699 A.2d at 633, is not tenable. Defendant, being sentenced to death, has every right to demand that the State's statistics be reliable.
Excluding the DQ-Alpha population frequency from the calculations, one in 173 African-Americans would share the other five DNA markers with defendant. More importantly, only one in 173 African-Americans (defendant being one of them) would share the DNA markers of the blood found at the scene.
The Court’s citations to two publications, which reported statistical results from samples of only ninety-four and 116 persons, ante at 193, 699 A.2d at 632, is amiss. There is no suggestion that samples of such small size have been found sufficiently reliable to be used in a criminal trial. Cf. Government of Virgin Islands v. Byers, 941 F.Supp. 513, 520 (D.V.I.1996) (noting that a DNA profile database "of only 750 Caucasians” is "quite small”). In any event, those studies at least included individuals chosen at random.
Some people have asserted that the product rule was endorsed in New Jersey by the trial court in Williams, supra, 252 N.J.Super. 369, 599 A.2d 960. See NRC Report, supra, at 207. However, a careful reading of the opinion in that case reveals that it never addressed whether population statistics are scientifically reliable. The opinion has, therefore, fallen under some criticism. Carter, supra, 524 N.W.2d at 783 ("An examination of [Williams] reveals that the issues of population substructure and the reliability of the population data bases were not addressed; therefore, [the] case[][is] not persuasive nor relevant as to [the] validity of ... statistical methodology.”).
That decision subsequently has been abrogated by constitutional and statutory amendments. N.J. Const., art. I, ¶ 12; L. 1993, c. 111 (signed May 5, 1993). The abrogation affects only offenses committed since the amendment’s enactment. Mejia, supra, 141 N.J. at 482, 662 A.2d 308. Irene Schnaps was murdered in 1985.
Harvey's confession was never written down or tape recorded by the police officers. At the first trial, Sgt. Rizzo recounted Harvey’s confession:
Mr. Harvey stated that he ... was in the area of Hunter's Glen apartments and he was trying sliding glass doors, patio doors seeing which was open and which wasn’t. He found one, which happened to be the apartment of Mrs. Schnaps. He went in, he made his [way] back to the bedroom, the master bedroom. He saw an object lying on the bed, he assumed it was a man at first. The person appeared to be sleeping. He had gone through the dresser drawers or the bureau and he had taken some jewelry, including the watch.
At that time Mrs. Schnaps woke up and she went to strike Mr. Harvey. I believe she struck him in the nose causing his nose to bleed. He then responded by striking her with a hammer or what he described as an object like a hammer. He struck her in the head.
From there she fell to an area in the bedroom near a fan. She laid there for some time and then she got up and fell closer to the doorway of the bedroom where she lay.
The Court responds that whether or not that change in mens rea occurred in a brief moment "is irrelevant." Ante at 152, 699 A.2d at 612. If defendant's intent did not change in that brief moment, as is certainly possible — indeed, probable — then the factor is inapplicable. The probability that defendant's assaultive intent did not change to homicidal intent is highly relevant.