delivered the opinion of the court:
Plaintiff, Bonnie Franson, on behalf of her minor child, Elizabeth Franson, filed a paternity suit against defendant, Philip Micelli. Following trial, the jury returned a verdict of paternity. On appeal, defendant contends that he is entitled to a new trial in light of the improper admittance of DNA probability evidence because: (1) the methodology by which DNA probability evidence is derived has not been generally accepted in the scientific community; (2) plaintiff failed to meet the burden of demonstrating the adequacy of the data base where the evidence did not show that the sampling was truly random, sufficiently large, and inclusive of individuals from defendant’s ethnic and geographic background; and (3) the statistical probability evidence was not a proper subject of expert opinion because it relieved the jury of its duty to find the ultimate facts and unduly prejudiced defendant.
Plaintiff testified that defendant was the father of her child. Defendant denied the allegation. Defendant filed a motion in limine to bar DNA evidence from being admitted at trial. The trial court held an evidentiary hearing and concluded that the DNA evidence satisfied the Frye test (Frye v. United States (D.C. Cir. 1923), 293 F. 1013) governing the admissibility of novel scientific evidence and allowed plaintiff to present expert testimony at trial on the issue. The DNA identification process consists of the following components:
(1) Extraction: DNA is removed from the specimen and "washed” with an organic solvent.
*22(2) Fragmentation: The extracted DNA chain is then cut into fragments at specific sites by mixing it with a restriction enzyme.
(3) Gel Electrophoresis: The DNA is placed in a gel to which an electrical current is applied, causing separation of the fragments into bands according to their length.
(4) Southern Blotting: The DNA bands are transferred to a nylon membrane while retaining the same positions they previously occupied on the gel. The double-stranded bands are then treated with a chemical that causes them to separate into single strands.
(5) Hybridization: Genetic probes (DNA clones) are applied, which bind to a specific, complementary DNA sequence on the membrane; the excess probe is then washed off.
(6) Autoradiography: The membrane is exposed to an X-ray film and developed so that the DNA banding patterns and their lengths can be visualized. Finally, the autoradiography is interpreted by comparing the DNA print to another DNA sample to determine if they match the band length.
DNA strands are double stranded and every person receives 50% of his or her genes from each parent. Therefore, one strand represents the DNA fragment which the person inherited from the mother and the other strand represents the DNA fragment inherited from the father. In paternity cases, DNA specimens from the child, mother and alleged father are placed next to each other. The child’s maternal DNA strand is first identified by matching it with the mother’s DNA. If the alleged father’s sample does not contain a strand which matches the child’s DNA, then the alleged father is not the father of the child. If a strand from the alleged father’s DNA does match, he is not excluded.
If DNA strands match, then a paternity index is calculated for the particular locus (i.e., each match). The paternity index is a function of comparing the likelihood of obtaining the child’s genotype from the mother and the alleged father to the likelihood of obtaining a child of the same genotype from mating the mother with a random man in the appropriate population. This calculation is determined for each probe. A cumulative paternity index is then calculated by multiplying the results of each individual calculation for each probe. This is called the "product rule.” This cumulative paternity index allows the tester to determine the statistical probability that a randomly chosen man from the appropriate population contributed the DNA.
In this case, the tests were conducted by Rick Staub, the associate director of paternity analysis, at Genetic Design. Genetic Design maintains DNA frequency data bases composed solely of males tested *23by Genetic Design and which it uses as representative samples of the representative populations. Genetic Design’s gene frequency tables are divided by race. The Caucasian male data base is composed of about 12,000. It is not further divided by ethnicity.
Staub testified that the cumulative paternity index in this case was 29,217,637 to 1. According to Staub, this meant that the probability that defendant was the father was 99.99%. Staub admitted that differences in gene frequencies among ethnic groups do exist although he did not believe the differences were "all that great.” He also admitted that gene frequency estimates are more accurate if larger data bases are employed.
Dr. Pravatchai Boonlayangoor testified for the defendant and pointed out flaws in the method of determining statistical probability calculations. First,.he noted that if subpopulations exist within the Caucasian race, then calculations based upon the general "Caucasian” data base would be inaccurate. He also asserted that a statement that it was 99.99% probable that defendant was the father is deceptive because there is very little difference between a combined cumulative index of 1/200 and 1/29,000,000. Additionally, he pointed out that in most cases one to three probes are performed; nonetheless, in this case, five probes were conducted. He asserted that to include additional probes dramatically increases the cumulative paternity index and never lowers it. Therefore, according to him, the probability statistic was misleading. Moreover, he testified that the "Caucasian” data base used by Genetic Design was too small and that an appropriate random sampling should include at least 100,000 subjects. He did admit, however, that some statisticians believe that 5,000 is a valid starting point. Boonlayangoor also found Genetic Design’s data bases to be inappropriate because they were divided by racial typing. Thus, if the probability statistic is determined by using the "Caucasian” data base and the putative father had mixed blood, the cumulative paternity index would be wrong. He concluded that it is more likely than not that defendant is not the father of plaintiff’s child.
Defendant’s first contention is that DNA statistical probability evidence should not have been admitted because the methodology by which such evidence was derived has not been generally accepted in the relevant scientific community.
Both parties concede that the standard in Illinois for determining the admissibility of novel scientific evidence is the Frye test. (Frye v. United States (D.C. Cir. 1923), 293 F. 1013; People v. Eyler (1989), 133 Ill. 2d 173, 549 N.E.2d 268, cert. denied (1990), 498 U.S. 881, 112 L. Ed. 2d 174, Ill S. Ct. 215.) The Frye test requires that, before *24expert testimony on a new scientific principle will be admitted at trial, "the thing from which the deduction is made must be sufficiently established to have gained general acceptance in the particular field in which it belongs.” Frye, 293 F. at 1014; Eyler, 133 Ill. 2d at 211, 549 N.E.2d at 285.
In support of his argument that DNA statistical probability evidence has not been generally accepted in the relevant scientific community, defendant relies upon the recent appellate court decision in People v. Watson (1994), 257 Ill. App. 3d 915, 629 N.E.2d 634, and the cases, articles, and studies mentioned therein. Plaintiff, on the other hand, asserts that the Frye test has been satisfied and cites a number of Illinois Appellate Court cases which have held that DNA evidence has been generally accepted. (People v. Johnson (1994), 262 Ill. App. 3d 565, 634 N.E.2d 1285; People v. Stremmel (1994), 258 Ill. App. 3d 93, 630 N.E.2d 1301; People v. Wardell (1992), 230 Ill. App. 3d 1093, 595 N.E.2d 1148; People v. Miles (1991), 217 Ill. App. 3d 393, 577 N.E.2d 477, appeal denied (1991), 142 Ill. 2d 661, 584 N.E.2d 136; People v. Lipscomb (1991), 215 Ill. App. 3d 413, 574 N.E.2d 1345, appeal denied (1991), 141 Ill. 2d 553, 580 N.E.2d 127.) In order to arrive at a conclusion, it is necessary to analyze the development of this line of cases dealing with DNA evidence.
Illinois courts recognize People v. Lipscomb (1991), 215 Ill. App. 3d 413, 574 N.E.2d 1345, and People v. Miles (1991), 217 Ill. App. 3d 393, 577 N.E.2d 477, both fourth district cases, as the first significant cases to consider the admissibility of DNA evidence. In Lipscomb, the defendant was convicted of four counts of aggravated criminal sexual assault. Based on the testimony of several experts, the trial court found that both the restriction fragment length polymorphism (RFLP) analysis used for DNA identification and the frequency procedures for determining the statistical probability of certain random matches were scientifically acceptable under Frye.
At trial, the results of the DNA tests were the primary evidence against the defendant. The expert testified that she conducted five "probes” of defendant’s blood and the semen sample recovered from the scene. She stated that for each probe the DNA of the semen sample matched the DNA of defendant’s blood. She determined the statistical frequency of this result occurring randomly in the general population by calculating the probability of each individual match and then multiplying the frequencies of the respective probes together, e.g., the "product rule.” She concluded that the statistical probability of a second match on all five probes in the black population was 1 in 6.8 billion. It was also explained to the jury that, since only 5 billion people live in the world, the likelihood of a second match obviously was extremely remote.
*25The Lipscomb court first rejected defendant’s contention that the trial court improperly applied the Frye test to DNA testing generally and not to the specific procedures used such as statistical frequencing. The court reasoned that Frye applies to new scientific principles and that the principle involved "is the general DNA forensic analysis” and not the procedures used within that framework. The court noted that different companies use different procedures when performing DNA analyses. It then determined that "[a]ny question concerning the specific procedures used by the company or expert goes to the reliability of the evidence and is properly considered by the jury in determining what weight to give to this evidence.” (Lipscomb, 215 Ill. App. 3d at 432, 574 N.E.2d at 1357.) The court did state, however, that "[i]f it is shown that the procedures used give an unreliable result, then the court may find it necessary to exclude th[e] evidence entirely.” Lipscomb, 215 Ill. App. 3d at 432, 574 N.E.2d at 1357.
The Lipscomb court then rejected defendant’s assertion that the testimony of statistical probability prejudiced him by usurping the jury’s role as fact finder and placing too much emphasis on numerical conclusions. The court reasoned that the evidence was admissible as relevant to identification and that any challenge to the reliability of the statistics went only to its weight.
In Miles, defendant was convicted of numerous criminal offenses and, on appeal, challenged the admissibility of the DNA evidence against him on the ground that the probability statistics were inadmissible. (Miles, 217 Ill. App. 3d at 402, 577 N.E.2d at 483.) According to expert testimony at trial, the probability that another person of defendant’s race would have randomly matched the DNA sample taken from the scene of the crime was 1 in 300,000. This frequency was arrived at by applying the "product rule.”
The Miles court expressly relied upon Lipscomb in arriving at its conclusion that probability statistics were admissible under Frye. The Miles court reasoned: "Implicitly, the [Lipscomb] court held the process of generating probability statistics is an integral part of the DNA identification process. Because the DNA identification process meets the Frye test and is admissible, probability statistics operated thereby are admissible.” Miles, 217 Ill. App. 3d at 405, 577 N.E.2d at 485.
The next case dealing with the admissibility of DNA evidence is People v. Wardell (1992), 230 Ill. App. 3d 1093, 595 N.E.2d 1148, a first district case. In Wardell, the defendant contended that the trial court erred in precluding DNA testing which he alone requested. The Wardell court held that the trial court did not abuse its discretion in *26precluding the DNA tests because at the time of trial DNA methodology was in its early stages of development and had not yet been generally accepted in the scientific community. The court then cited Lipscomb and Miles, however, and stated that "Illinois cases have recently held that the DNA identification or fingerprinting procedure is generally accepted within particular scientific fields and is admissible.” (Wardell, 230 Ill. App. 3d at 1100, 595 N.E.2d at 1153.) Inexplicably, and in direct contradiction with its statement of general acceptability, the Wardell court then recognized in a footnote that scientific experts have recognized that DNA techniques are not infallible and that "[t]he admissibility and reliability of DNA testing continues today to be the subject of debate.” Wardell, 230 Ill. App. 3d at 1101 n.2, 595 N.E.2d at 1154 n.2.
In People v. Stremmel (1994), 258 Ill. App. 3d 93, 630 N.E.2d 1301, a second district case, defendant contended that the trial court erred in admitting DNA probability evidence. Defendant conceded, however, that the HELP DNA procedures and the concomitant probability estimates met the Frye standard. Defendant argued that the Frye test was insufficient to determine the admissibility of such evidence due to its potential to be extremely prejudicial and asserted that, in addition to satisfying Frye, the reliability of the evidence should also be established "by showing that the DNA tester in each case actually performed accepted scientific procedures in analyzing the forensic samples in question.” (Stremmel, 258 Ill. App. 3d at 105, 630 N.E.2d at 1309.) Relying on Lipscomb, the Stremmel court rejected defendant’s proposed new standard and stated that the adequacy of the tester’s procedures was a matter for the jury to determine in its role as fact finder.
Recently, in People v. Watson (1994), 257 Ill. App. 3d 915, 629 N.E.2d 634, the first district pulled away from the Lipscomb /Miles line of cases. The court did so because it appeared that in those cases "there was a complete absence of opposing scientific perspectives challenging the DNA evidence.” (Watson, 257 Ill. App. 3d at 928, 629 N.E.2d at 643.) In Watson, the State appealed the trial court’s exclusion of DNA evidence on the ground that it did not satisfy the Frye test. The Watson court first concluded that the theory underlying DNA profiling is generally accepted. The court did not stop here, however, as did the Lipscomb court and its progeny and conclude, therefore, that the procedures used when performing DNA analyses were admissible. The Watson court went on to analyze each step in the DNA methodology under the Frye test. The court then concluded that the RFLP technique used for matching DNA was also generally accepted in the relevant scientific community.
*27Finally, the Watson court addressed whether the manner of determining a match’s statistical significance was generally accepted. In concluding that the manner of determining the statistical significance of a DNA match was not generally accepted, the court cited extensively to the California Appellate Court case of People v. Barney (1992), 8 Cal. App. 4th 798, 10 Cal. Rptr. 2d 731, and considered expert testimony and scientific literature. The Barney court documented articles and studies which demonstrated " 'a fundamental disagreement among population geneticists concerning the determination of the statistical significance of a match of DNA patterns.’ ” (Watson, 257 Ill. App. 3d at 930, 629 N.E.2d at 644, quoting Barney, 8 Cal. App. 4th at 814, 10 Cal. Rptr. 2d at 740.) The Barney court cited to an article written by Richard C. Lewontin of Harvard University and Daniel L. Hartl of Washington University which was published in a leading scientific journal. Lewontin and Hartl first attacked the "product rule” of determining statistical probabilities in DNA testing because its efficacy is dependent upon a truly random population. In other words, when applying the "product rule” within the "Caucasian” data base, its validity is dependent upon the assumption that all Caucasians mate within their groups randomly and without regard to religion, ethnicity, or geography. Lewontin and Hartl contend, however, that contrary to the assumption of random mating, ethnic subgroups tend to mate with persons within that same subgroup. In other words, people mate with others of the same religion, ethnicity and geographic region. Consequently, the existence of a DNA fragment may be relatively common in certain subgroups but not in the broader data base. Thus, the frequency estimate calculated by the "product rule” can result in an extremely inaccurate statistical probability or a random match in the defendant’s subgroup. (See also People v. Pizarro (1992), 10 Cal. App. 4th 57, 12 Cal. Rptr. 2d 436 (extensively discussing debate on data bases, statistical probabilities and, particularly, substructures).) Additionally, Lewontin and Hartl point out that the "product rule” also assumes that DNA fragments occur independently. They assert that it has not been sufficiently established that all fragments occur independently. The "product rule” does not take into account that two fragments may commonly occur together.
The Barney court also acknowledged an article by Ranajit Chak-roborty of the University of Texas and Kenneth K. Kidd of Yale University in which Chakroborty and Kidd assert that Lewontin and Hartl "exaggerate” both the extent of endogamy in the population and the effect of substructuring on the reliability of DNA statistical analysis. The court noted that the preface introducing the Lewontin-*28Hartl and Chakroborty-Kidd articles described the debate on DNA statistical analysis as "bitter” and "raging” and stated that " 'tempers are flaring, charges and countercharges are flying.’ ” (Watson, 257 Ill. App. 3d at 932, 629 N.E.2d at 645, quoting Barney, 8 Cal. App. 4th at 816, 10 Cal. Rptr. 2d at 741.) The Watson court also considered a report recently published by the National Research Counsel titled "DNA Technology in Forensic Science” (NRC Report) which concluded that "a substantial controversy” exists concerning the present method of determining frequency probabilities.
The Watson court concluded that since there clearly was no consensus as to the accuracy of the Federal Bureau of Investigation’s method of determining statistical probabilities and because the probability of a coincidental match is an essential part of the DNA evidence, the manner of determining statistical probabilities was not generally accepted under Frye. The court expressly rejected the argument that any objections to the calculations go only to its weight. Finally, the court held that since "a probability assessment is essential in order to give meaning to a 'match,’ and that absent such an assessment, the fact of the match, standing alone, is inadmissible at trial.” Watson, 257 Ill. App. 3d at 930, 629 N.E.2d at 644.
The court did not, however, set down a flat rule. It held that if the proponent of the DNA evidence could demonstrate a consensus within the appropriate scientific community of a "more conservative” approach to determining statistical probabilities, the evidence may be admissible. The court remanded the case because the NRC Report was not available to the trial court at the time of the trial for a determination as to whether the NRC suggested alternative methods satisfied Frye and, therefore, whether the statistics would be admissible.
In People v. Johnson (1994), 262 Ill. App. 3d 565, 634 N.E.2d 1285, the fourth district rejected the Watson decision. In ruling that DNA evidence was properly admitted, the Johnson court stated:
"We note that in Lipscomb, the trial court conducted an exhaustive, lengthy pretrial hearing on DNA identification testimony at which the State’s proffered evidence was vigorously and thoroughly challenged. With all due respect to the reservations expressed by the Watson court, we stand by and reaffirm our holdings in Lipscomb and Miles.” Johnson, 262 Ill. App. 3d at 570, 634 N.E.2d at 1289.
The fifth district partially rejected Watson in People v. Heaton (1994), 266 Ill. App. 3d 469, ruling that the trial court did not abuse its discretion in admitting DNA identification evidence based on the record before it. The issue before the court was whether the method *29used to calculate probability of random match, e.g., the "product rule,” met Frye. Defendant argued that it did not, relying solely on the NEC Report. While acknowledging a debate over the "product rule,” the court held that because the NEC Report was never submitted to or brought to the attention of the trial court, the appellate court would not now rely on it. The trial court had only the rulings of Lipscomb and Miles before it at the time of trial. In following the law available to it, the trial court concurred with those decisions and held that the "product rule” satisfied Frye. The fifth district rejected Watson’s "broad review” analysis in favor of looking only to the evidence contained in the record on review.
After reviewing the above cases, we accept the reasoning of Watson that the manner of determining DNA statistical probability evidence is not yet generally accepted in the relevant scientific community and reject the Lipscomb/Miles line of cases as well as the Heaton rationale.
The first reason we reject the Lipscomb /Miles cases is because these courts improperly applied the Frye test. The Lipscomb court reasoned that Frye applies to new scientific principles and the principle involved is the general DNA forensic analysis and not the procedures used within that framework. Thus, the court asserted that the fact that different companies may use different procedures in arriving at their results only goes to the weight of the evidence. The Miles court then interpreted the Lipscomb court’s reasoning to mean that, since the DNA identification process meets the Frye standard, probability statistics based upon the DNA identification are also admissible because "generating probability statistics is an integral part of the DNA identification process.” (Miles, 217 Ill. App. 3d at 405, 577 N.E.2d at 485.) The subsequent cases which followed Lipscomb and Miles simply adopted this reasoning.
The first flaw in this reasoning is that, although "general DNA forensic analysis” may be the new scientific principle, Frye requires that "the thing from which the deduction is made” must be generally accepted. It cannot be disputed that the theory behind DNA testing is well accepted. This does not mean, however, that simply because the concept is recognized the procedures used are not subject to a Frye analysis and any challenge to their validity only goes to the weight of the evidence. Under Frye, if the procedure or "thing” upon which the DNA result was determined is not generally accepted, then the result is inadmissible. For instance, DNA testing to determine a "match” may be well recognized, but the "thing” upon which the result is based is the procedure used to arrive at the determination of whether there is a "match.” Thus, if the procedures are *30not generally accepted, then the result is inadmissible under Frye. Similarly, although it may be generally accepted that statistical probabilities can be calculated based upon the "matching” results, if the method used to calculate the statistical probabilities is not generally accepted as valid in the relevant scientific community, then the statistics should be inadmissible.
The second flaw in the reasoning of these courts is that "[b]ecause the DNA identification process meets the Frye test and is admissible, probability statistics operated thereby are admissible.” (Miles, 217 Ill. App. 3d at 405, 577 N.E.2d at 485.) This is an illogical jump in reasoning. The procedures used to "match” or "identify” DNA strands are completely separate and result in different findings than the procedures underlying probability statistics. The result of DNA identification procedures is the determination of whether the DNA strands "match.” If a "match” occurs, the result of statistical probability procedures is the determination of the frequency of such a "match” in the relevant population. Thus, the fact that DNA identification procedures may be generally accepted and thus the results admissible does not mean that subsequent statistical probabilities should then be admissible if the statistical probability procedures are not generally accepted.
The second reason we follow Watson’s rationale is because it is clear that there is a fundamental disagreement within the scientific community concerning the manner of determining the statistical significance of a match of DNA patterns. We cannot close our eyes to this lack of general acceptance simply because previous Illinois appellate courts have found general DNA testing to satisfy the Frye test. The manner of determining statistical probabilities must be generally accepted within the relevant scientific community and not the Illinois Appellate Court. Clearly, these courts were not presented with the detailed evidence that we have before us documenting the controversy presently raging in the scientific community concerning this procedure.
The NEC Report recommends the use of alternative methods of calculating statistical probabilities. As in Watson, we remand to the trial court for a determination as to whether either of these methods is generally accepted in the relevant scientific field and, therefore, satisfies Frye. If the trial court determines that either of the methods satisfies Frye, it must then ascertain the appropriate probability estimate to be admitted at a new trial. If the court concludes that neither of the methods meets the Frye test as generally accepted in the scientific field, it must exclude the DNA evidence at the new trial.
*31Based on resolution of defendant’s first contention, we need not address defendant’s other contentions.
For the foregoing reasons, the judgment of the circuit court of Cook County is reversed and remanded for a new trial in conformance with this decision.
Reversed and remanded with directions.
CAMPBELL, P.J., and O’CONNOR,1 J., concur.