Summary
ruling that admission of novel DNA evidence "is a question of law" subject to de novo review
Summary of this case from State v. HarveyOpinion
No. 93SC167
Decided March 6, 1995. Rehearing Denied April 3, 1995.
Certiorari to the Colorado Court of Appeals.
JUDGMENT AFFIRMED.
David F. Vela, Colorado State Public Defender, Martin Gerra, Deputy State Public Defender, Denver, Colorado, Attorneys for Petitioner.
Gale A. Norton, Attorney General, Stephen K. ErkenBrack, Chief Deputy Attorney General, Timothy M. Tymkovich, Solicitor General, Deborah Isenberg Pratt, Assistant Attorney General, Criminal Enforcement Section Denver, Colorado, Attorneys for Respondent.
We granted certiorari in Lindsey v. People, 868 P.2d 1085 (Colo.App. 1993), to consider whether, in view of this court's decision in Fishback v. People, 851 P.2d 884 (Colo. 1993), the court of appeals erred in holding that the techniques employed to calculate the statistical frequency of a declared match in a Deoxyribonucleic Acid (DNA) typing case related to the weight of the evidence and not to its admissibility. Petitioner advances three reasons for reversal: (1) that at the time the DNA evidence was offered there was significant scientific disagreement concerning the validity of the statistical techniques employed; (2) that the district court heard expert testimony regarding the disagreement; and finally (3) that the disagreement has not been resolved but has become more pronounced. We hold that the evidence was properly admitted and affirm the court of appeals.
I. FACTUAL AND PROCEDURAL BACKGROUND
The defendant, Gregory Lindsey (Lindsey), was convicted of first degree sexual assault, second degree burglary, and four habitual criminal counts. At trial, the court allowed the People to introduce evidence that the defendant's DNA matched DNA samples taken from seminal fluid obtained by medical examination of the victim following the assault. The court also allowed expert testimony regarding the probability that the defendant's DNA profile would match a DNA profile obtained from a randomly selected African American individual. Lindsey argues that the trial court improperly admitted the DNA evidence because the procedure used to compute the statistical probability of a random DNA profile match was not generally accepted by the scientific community when the trial court ruled that the evidence was admissible.
A. Facts
On February 16, 1988, a man wearing a mask broke into a woman's Colorado Springs home and sexually assaulted her. She did not look at her attacker who identified himself as a "black man". She told police she believed the man who attacked her lived next door because the two men living next door were tall and powerfully built, similar to her attacker. Lindsey, in fact, lived next door to the woman, but was not a suspect until after he was arrested in connection with sexual assault on a second woman.
The second woman was sexually assaulted in January 1988, at which time police recovered evidence from the crime scene including seminal fluid obtained from a medical examination of the woman, and from stains on her bedclothes. In May 1988, an intruder again broke into the second woman's home, and attempted to assault her, but fled when her neighbor telephoned. Lindsey was arrested after the May assault, at which time police obtained samples of his blood. In June 1988, police furnished Lindsey's blood samples together with evidence obtained at both crime scenes to Cellmark Diagnostic Corporation (Cellmark) for DNA analysis. Cellmark's report declared a match between the DNA extracted from Lindsey's blood sample and the DNA recovered from both crime scenes.
Lindsey was originally charged in a single information with sexual assault on both victims. After the court granted Lindsey's motion to sever the charges, the January 1988 assault case was tried first. At the trial of the February 1988 assault case the People's experts testified regarding the probability that Lindsey's DNA profile would match the profile of a randomly selected African American individual. The estimates ranged from one in 340 billion down to one in 21 million using more conservative frequency calculations.
Upon conclusion of the first trial Lindsey was convicted of second degree burglary and first degree attempted sexual assault and habitual criminal counts. The decision was affirmed on appeal, People v. Lindsey, 89CA340 (Colo.App. Sept. 27, 1990) (not selected for official publication), cert. denied, (Colo. Feb. 11, 1991). No DNA issues were raised on appeal.
B. Pretrial proceedings
Prior to admitting the DNA evidence, the trial court conducted proceedings to determine its admissibility under Frye v. United States, 293 F. 1013 (D.C. Cir. 1923).
The first Frye hearing was held on September 16, 1988, before the charges against Lindsey were severed. The People presented two expert witnesses who testified regarding the general acceptability of the theories and techniques used to conduct DNA typing. Lindsey's counsel advised the court that no funds were available to retain experts to refute the People's evidence. The court continued the Frye hearing at Lindsey's request. On January 10, 1989, after the charges were severed, the parties reconvened to decide the admissibility of the DNA evidence while jury selection in the January 1988 assault case was underway (January 1989 hearing). At this hearing the People recalled Drs. Housman and Garner and presented testimony from an additional expert who testified regarding the general acceptance of the methods used to calculate random match probabilities. Again, Lindsey presented no contrary evidence. Following the hearing the trial court ruled that the DNA typing evidence completed by Cellmark was admissible under the Frye test.
Dr. David Housman, a molecular biologist from Massachusetts Institute of Technology, testified regarding the underlying scientific theories and techniques used in DNA typing. Dr. David Garner, the director of Cellmark laboratories testified regarding the acceptability of Cellmark's DNA typing protocol.
Dr. David Goldman, an expert in population genetics, testified regarding the reliability of the statistical methods used to calculate the random match probabilities.
The trial court concluded:
[T]he standard in Fry[e] is for the court to decide whether or not the technique being presented to the court is generally accepted in the relevant scientific community . . . in this case including both the scientific community studying the DNA and the scientific community studying biological statistics, whether the application of that technique and the particular procedure is valid and then whether the ultimate result would be more helpful to the finder of fact than prejudice [sic] to the defendant.
Upon conclusion of the January 1988 assault case, and prior to commencing the February 1988 assault case, Lindsey requested a new Frye hearing, and advised the court the defense was now ready to challenge the admissibility of the DNA evidence. In response, the People requested the court to rule that the prior hearings were dispositive with respect to admissibility of DNA evidence. The People referred the court to several recent Colorado cases where trial courts had admitted DNA evidence, together with decisions from Maryland, Florida and New York where courts had allowed DNA evidence. The court denied Lindsey's motion for a new Frye hearing and ruled that the Frye test had been met with respect to both the scientific techniques of analysis and the statistical methods of calculating random match probabilities. Lindsey objected, arguing Cellmark's DNA typing procedures, particularly the genetic probes relied upon to isolate polymorphic DNA segments, had not been accepted by the scientific community. In response, the court granted Lindsey permission to file a motion in limine to address his concerns regarding Cellmark's DNA testing protocols.
Lindsey's counsel explained that the DNA evidence was not attacked at the first trial because evidence of Lindsey's guilt apart from the DNA evidence existed.
Andrews v. State, 533 So.2d 841 (Fla.Dist.Ct.App. 1988) (DNA typing admissible under relevancy test); Cobey v. State, 559 A.2d 391 (Md. Ct. Spec. App. 1989) (DNA typing admissible under Frye test); People v. Wesley, 533 N.Y.S.2d 643 (N.Y.Co.Ct. 1988) (finding DNA typing generally acceptable and allowing extraction of blood samples to conduct DNA testing), aff'd 633 N.W.2d 451 (N.Y. 1994).
The trial court explained its ruling:
The Court has considered the motion presented to it, and the court in determining whether or not a Frye hearing should be held in this case must address the issue of a Frye hearing, which is not whether a particular test was conducted in a manner accepted, nor is proper in the community, nor whether a particular vendor did the job correctly or is personally qualified or individually qualified to perform certain tests, but whether or not the scientific basis upon which the evidence being presented to the Court is valid and timely accepted in the scientific community, which uses those techniques.
Based upon the precedent in the State of Colorado, not only ruling in this trial court, but other trial courts, based upon this Court's consideration of the evidence presented at its own Frye hearing previously in a case against this same defendant, and the written decisions of the Court in New York and the Appellate decisions that have come out of Florida and Maryland, the Court is going to deny a Frye hearing.
The court finds that the scientific basis upon which both the DNA testing occurs and the scientific basis for analyzing the probability of a particular characteristics [sic] through the use of statistics is commonly accepted and [are] commonly accepted tools used within the scientific community.
Lindsey filed such a motion, and requested that the court suppress the DNA evidence. At the same time he requested reconsideration of the denial of a Frye hearing. The court heard argument on the motion in limine in February 1990. Although Lindsey originally objected to Cellmark's genetic probes, at the hearing his argument focused on the statistical procedures Cellmark used to calculate random match probabilities. Lindsey claimed Cellmark's database was in disequilibrium causing the random match calculations to be unreliable. Because the court had already ruled that the DNA evidence was admissible under the Frye test, the court ruled on Lindsey's motion in limine by analyzing Lindsey's objections under CRE 702 and 403. The court summarized the testimony:
At the hearing, defense counsel called Dr. Seymour Geisser, an expert statistician, who testified regarding flaws in Cellmark's database and calculation of statistical probabilities. Dr. Paul Hagerman, an expert in biogenetics and population genetics, testified that forensic application of Restricted Fragment Length Polymorphism (RFLP) analysis was not generally accepted. Dr. Lawrence Mueller, an expert in population genetics, testified that Cellmark's database was deficient for the calculation of allele frequencies. Dr. Randell Libby, an expert in molecular biology and forensic testing, testified that DNA typing in the forensic setting had been subject to criticism, identified potential problems with Cellmark's protocols and matching criteria and opined that Cellmark's statistical analysis was not generally accepted. Dr. Moses Samuel Schanfield, an expert in human genetics, questioned Cellmark's DNA testing protocols, matching criteria and database used to calculate match frequencies. In rebuttal, the People called Dr. Kenneth Kidd, an expert in genetics, psychiatry and biology, who testified that Cellmark's database was sufficient to calculate random match probabilities, and provided an estimate based on more conservative frequency calculations of one in 21 million. In addition, the People relied on the record of the testimony presented at the September and January hearings.
The Defendant's witnesses unanimously agreed that no reliable statistical prediction of probability of a match between two unrelated individuals could be established upon the evidence or the system utilized by Cellmark. Witnesses for the prosecution have recognized that further study and refining of the techniques should continue, but testify that the technique was reliable at this point to establish the identity of persons to a reasonable degree of scientific probability by utilization in comparing DNA patterns. Applying Rule 702 of the Colorado Rules of Evidence, the Court finds that the witnesses produced by both the People and the Defendant are qualified experts and, therefore, qualified to testify. . . . The existence of significant disagreement among experts to the validity of the opinions expressed is not a reason to exclude this testimony from the finder of fact.
The court went on to rule that the "probative value of the evidence is not outweighed by the danger of unfair prejudice or the jury being misled, therefore, the Court, having designated the motion not as a Frye hearing but as a motion in limine, denies the Defendant's motion in limine."
The trial court found that the controversy regarding Cellmark's statistical analysis went to the weight rather than the admissibility of the evidence. Lindsey argues the DNA evidence should have been excluded because the methodology used to determine the statistical significance of a DNA match was not accepted by the relevant scientific community at the time of trial, and therefore was not admissible.
C. Court of Appeals' analysis
In People v. Lindsey, 868 P.2d 1085 (Colo.App. 1993), the court of appeals affirmed, applying a three step analysis adopted in People v. Castro, 545 N.Y.S.2d 985 (N.Y. Sup. Ct. 1989). The court first explained Lindsey's challenge to forensic application of the statistical portion of the evidence required review under the Frye test. Lindsey, 868 P.2d at 1089. Whether Cellmark properly implemented generally accepted techniques required review under a relevancy standard. Id. at 1090. Based on these parameters that court held that the DNA evidence was admissible under both the Frye test and CRE 702. Id. at 1091. The court found that the in limine hearing served the same purpose as Prong III of the Castro test; that is, to decide whether the testing laboratory in this case applied accepted scientific techniques in its forensic DNA analysis. Id.
In Castro, the court identified three prongs to the Frye test applicable to DNA typing:
Prong I.
Is there a theory, which is generally accepted in the scientific community, which supports the conclusion that DNA forensic testing can produce reliable results?
Prong II.
Are there techniques or experiments that currently exist that are capable of producing reliable results in DNA identification and which are generally accepted in the scientific community?
Prong III.
Did the testing laboratory perform the accepted scientific techniques in analyzing the forensic samples in this particular case?
People v. Lindsey, 868 P.2d at 1088 (Colo.App. 1993) (quoting People v. Castro, 545 N.Y.S.2d at 987).
The court went on to explain the first two prongs deal exclusively with admissibility under the Frye test. It held that the third prong could be handled at a motion in limine to determine whether the "testing laboratory performed the accepted scientific techniques in analyzing the forensic samples in this particular case." Castro, 545 N.Y.S.2d at 995.
II. SCIENTIFIC BACKGROUND [17] A. Generating DNA profiles
In Fishback, we explained the procedures used to extract and isolate DNA fragments in order to construct a DNA profile using Restriction Fragment Length Polymorphism (RFLP) analysis. Lindsey does not question the scientific theories underlying RFLP analysis. Lindsey nevertheless argues that the evidence should have been excluded based on scientific disagreement over the statistical analysis used to provide a context for the DNA profile match.
The debate regarding the theory and techniques applied in RFLP analysis has been resolved to allow DNA profile evidence under Frye. See Fishback v. People, 851 P.2d 884, 893 (Colo. 1993) (holding that trial courts may take judicial notice of the acceptability of the techniques employed in RFLP analysis); see also Thomas J. Fleming, Annotation, Admissibility of DNA Identification Evidence, 84 A.L.R.4th 313 § 4 (1991 Supp. 1994).
In order to understand Lindsey's argument we must briefly revisit the scientific processes involved in RFLP analysis explained in Fishback, 851 P.2d at 883-84. RFLP analysis involves extracting and isolating small portions of the DNA molecule to examine sites on the DNA that exhibit highly variable characteristics. Scientists refer to the sites of examination as loci. Each locus consists of variant forms of genes known as alleles. Two allele forms occupy each locus, one on a chromosome inherited from the mother, and one on a chromosome inherited from the father. When the same form of allele occupies both loci on the chromosomes the individual is homozygous for that allele, and when different alleles occupy the loci, the individual is heterozygous for both alleles.
Alleles are any one of a series of two or more different genes that may occupy the same position or locus on a specific chromosome. Stedman's Medical Dictionary 45 (25th ed. 1990).
The alleles which comprise the loci consist of multiple pairs of the nucleotide bases Adenine (A), Cytosine (C), Guanine (G) and Thymine (T) which bond according to the base pair rule — A with T and G with C. In order to generate a profile that is unique Cellmark studies four highly variable sites referred to as polymorphic loci. Polymorphic sites contains repeat units of identifiable base pair sequences which distinguish one person from another, referred to as Variable Number Tandem Repeat (VNTR) loci. The sequences of nucleotide bases may repeat hundreds of times at each VNTR loci, depending on the individual. See B.S. Weir B.S. Gaut, Matching and Binning DNA Fragments in Forensic Science, 34 Jurimetrics J. 9, 10 (1993).
Genetic probes play a central role in identifying polymorphic regions of the DNA molecule. The probes act as magnets to lock onto DNA segments known to occur in, or adjacent to, these highly variable regions of the DNA chain. See William C. Thompson Simon Ford, DNA Typing, 75 Va. L. Rev. 45, 74-76 (1989). The probes are treated with a radioactive solution before they are combined with the DNA. The radioactivity enables scientists, after completing a series of complex scientific procedures, to record the probe's location on x-ray film. Thus, a picture of the individual's DNA is created, called an autoradiogram (autorad) which resembles an electronic bar code. Cellmark's four probes create an eight band autorad if the individual is heterozygous for each allele. Fewer bands appear when the individual is homozygous for any particular allele.
Scientists have long used genetic probes in a variety of medical settings unrelated to forensic DNA typing including identification of genes for inherited diseases and inherited genetic disorders; as a tool for gene splicing; to detect genetically engineered microorganisms in the environment and as a tool for quality control in the food and drug industry. Thompson Ford, DNA Typing at 72 n. 7.
The processes involved in creating an autorad are discussed at length in Fishback. Fishback, 851 P.2d at 886-88.
The number of bands in the autorad will depend on the form of alleles at each locus. Heterozygous alleles contribute two bands to the autorad, and homozygous alleles contribute a single band. See I.W. Evett, DNA Statistics: Putting the Problems into Perspective, 33 Jurimetrics J. 139, 140 n. 2 (1992).
B. Frequency calculations
After autorads are generated for both the known DNA sample and the evidence sample, scientists compare them to decide whether the bands match. Once a match is declared scientists put the data in context by explaining the relative rarity of such an occurrence. Thus, the significance of the match is demonstrated by calculating the likelihood that a matching autorad profile would appear in a randomly selected individual.
The first step in generating the statistics for comparison involves the creation of an autorad database. When Lindsey was tried scientists typically compared the known sample with a random sample of autorads from individuals who shared the same broad racial characteristics. Here, Cellmark's database was comprised of DNA autorads from African American individuals who donated blood at a Detroit blood bank.
When the suspect was Caucasian, the sample population would consist of Caucasians, if the suspect was Hispanic, the sample population would be Hispanic. Theoretically, racially similar databases were used to create a greater likelihood of a DNA profile match which safeguards against understating the probability of a random match. The use of broad racial categories for classification purposes underlies one of Lindsey's objections to the DNA evidence. See discussion infra pp. 16-17.
Once a database is constructed, autorad band frequencies are calculated by comparing the known bands with the database bands to determine how frequently bands match. For example, if an individual band appears in three out of one hundred individuals its frequency is .03. After individual band frequencies are calculated, the likelihood that the complete autorad would be duplicated in a randomly selected individual is calculated using the product rule. The product rule is simply the mathematical formula used to determine the probability that two independent events would occur simultaneously, calculated by multiplying the probability of each event.
Scientists use either fixed or floating bin analysis to calculate match frequencies. Binning helps to account for variables in recording autorads, and provides confidence limits on frequency estimates. In fixed bin analysis autorads from the database are assigned to certain bins which have been previously measured and assigned a frequency. A tolerance is then built around the known or evidence band after which the band is assigned to a bin. If the band falls within two bins the band is assigned the highest bin frequency for probability calculations.
Floating bin analysis focuses on the autorads obtained from the evidence sample, around which individual bins are constructed. Typically, floating bins are constructed with a resolution tolerance of a certain number of standard deviations centered around the evidence band. Database bands which fall within the bin created for the evidence band are then assigned to the evidence band's floating bin to calculate band frequencies. Here, Cellmark used a floating bin based on Lindsey's known sample to calculate autorad band frequencies.
In both cases binning typically results in higher probable match frequencies. The higher match frequencies weigh in a suspect's favor because the probability calculations will yield a greater chance of a random match. See B.S. Weir B.S. Gaut, Matching and Binning DNA Fragments in Forensic Science, 34 Jurimetrics J. 9, 12-14 (1993).
Two product rule calculations are required to determine the likelihood of a random profile match. First, the frequency of a match at each locus is calculated. Because the bands generated at each locus depend on whether the individual is homozygous or heterozygous for those particular alleles, the formula varies depending on the allele form. The frequency for a homozygous allele is computed using the formula p(al)2; the frequency for a heterozygous allele is computed using 2(p(al1)p(al2)); where p denotes probability, and al represents each allele. After the frequency for each locus is calculated, the frequency of the complete genotype is calculated by multiplying together the four loci frequencies.
For example, one variant of a heterozygous allele may occur one time in a group of one hundred individuals for a frequency of .01, and the corresponding variant of the allele may appear in one out of every ten individuals for a frequency of .1. Scientists estimate the probability that these alleles will occur in any individual at the locus identified by the corresponding probe using the formula 2(p(al 1)p(al 2)) where pal1 is the likelihood of variant one occurring, and pal2 is the likelihood of variant two occurring. Thus, the probability of the random occurrence of both bands is 2((.01)(.1)) = .002, or one in five hundred.
For example, if a random sample shows the frequencies of alleles Al1, Al2, Al3 and Al4 are .04, .06, .4 and .01 respectively, the estimated frequency for the genotype is the product of all four individual allele frequencies (.04)(.06)(.4)(.01)=.0000096; or one in approximately 104,170. See National Research Council Publication DNA Technology in Forensic Science (NRC Report) at s-10.
Product rule probabilities are only accurate estimates if the events underlying the calculations are truly independent and random. Here, independence means the probability of finding one allele is not affected by having found any other allele. For each locus, event independence occurs when there is no correlation between the allele inherited from an individual's mother and the allele inherited from that individual's father. When no correlation between the two parental alleles exists the population sample is considered in Hardy-Weinberg equilibrium. Event independence for step two of the calculation, the frequency of the multi-locus genotype, occurs when no there is no statistical correlation between alleles within a target population. If no correlation between the alleles identified at different loci exists, the population sample is in linkage equilibrium.
Hardy-Weinberg principles derive from an algebraic equation that describes the genetic equilibrium within a population. First discovered in 1908 by G.H. Hardy and W. Weinberg, the principle states that gene frequencies will remain constant from generation to generation within a population unless outside forces act to change it, provided mating remains random. See People v. Wesley, 533 N.Y.S.2d 643 (N.Y.Co.Ct. 1988).
Lindsey argues that Cellmark's population database was out of both Hardy-Weinberg and linkage equilibrium. Consequently, he contends it was not possible to calculate a meaningful random match probability. The scientific debate Lindsey relies upon to claim statistical probability calculations were not generally accepted within the relevant scientific community relates to the possibility of subgrouping among populations. Subgrouping is based upon the premise that census populations designated "Caucasian," "Black," or "Hispanic" actually consist of multiple genetically diverse subpopulations. See R.C. Lewontin Daniel L. Hartl, Population Genetics in Forensic DNA Typing, 24 Science 1745 (Dec. 21, 1991). Scientists have speculated that such subgrouping may cause both Hardy-Weinberg and linkage disequilibrium within the population sample used to calculate random match probabilities, affecting the validity of product rule calculations.
III. STANDARD OF ADMISSIBILITY
Initially, we recognize that neither the trial court nor the court of appeals had the benefit of our view on the admissibility of DNA evidence explained in Fishback v. People 851 P.2d 884 (Colo. 1993). In Fishback we affirmed our prior rulings that Frye v. United States, 293 F. 1013 (D.C. Cir. 1923) sets forth the applicable standard for determining the admissibility of novel scientific evidence. Consequently, we review the trial court's rulings with reference to the Frye test, and the parameters outlined in Fishback for the application of Frye to DNA typing.
We apply the Frye test fully aware that since our decision in Fishback the United States Supreme Court has rejected Frye as the standard for admissibility of scientific evidence in favor of what has been characterized as a more lenient relevancy standard under FRE 702. Daubert v. Merrill Dow Pharmaceutical, Inc., 113 S. Ct. 2786 (1993). Here, however, we construe the Colorado Rules of Evidence and are not bound by a non-constitutional construction of the Federal Rules of Evidence. We do not consider the relative merits of the Frye test or our corollary state rules of evidence for the simple reason the issue is not now before us, and has not been fully briefed or argued.
In light of Daubert many state courts are reconsidering the viability of the Frye test with mixed results. Compare People v. Leahy, 882 P.2d 321 (Cal. 1994) (after analysis of the merits of the more conservative Frye approach, the California Supreme Court concluded "Daubert, which avoided the issue of Frye's `merits' presents no justification for reconsideration of that aspect of our holding in Kelly [adopting Frye]") with State v. Alberico, 861 P.2d 192, 203 (N.M. 1993) (explaining Frye test is not a legitimate means for determining what is and what is not scientific knowledge, "`[a]ccordingly, we hold that a particular degree of acceptance of a scientific technique within the scientific community is neither a necessary nor a sufficient condition for admissibility; it is, however, one factor that a district court normally should consider in deciding whether to admit evidence based upon the technique.'" (quoting United States v. Downing, 753 F.2d 1224, 1227 (3d Cir. 1985)); Fishback v. People, 851 P.2d 884, 898-99 (Colo. 1993) (Mullarkey, J., concurring in result); see also State v. Bible, 858 P.2d 1152 (Ariz. 1993), cert. denied 114 S. Ct. 1578 (1994) (concluding "[t]he field of DNA testing is probably the worst subject to use to decide whether or how to refine, replace or abolish Frye. . . . leav[ing] Daubert for another day").
In Fishback, we explained that before novel scientific evidence is admissible both the "theory and techniques underlying novel scientific evidence must be generally accepted under Frye." There, we adopted a two-pronged test for admissibility: (1) general acceptance in the relevant scientific community of the underlying theory or principle, and (2) general acceptance in the relevant scientific community of the techniques used to apply that theory or principle. Fishback, 851 P.2d at 891. Once the Frye standard has been met, any challenge to the implementation and execution of these techniques "goes to the weight to be accorded such evidence." Id.
Before we can properly consider the trial court's ruling we must resolve the two threshold questions underlying the first prong of our Frye test; (1) who comprises the relevant scientific community; and (2) when a finding of general acceptance is appropriate. In Fishback we explained the relevant scientific community for DNA typing included several scientific disciplines. Fishback, 851 P.2d at 892. Here, the attack on the statistical significance of a declared match involves the disciplines of population genetics, human genetics and demographics. Id.
The more difficult inquiry involves the determination of when a novel scientific theory is "generally accepted" within the relevant scientific community. Frye does not clarify general acceptance. In Fishback we explained a "party need not prove the absolute validity of the techniques used in producing novel scientific evidence before it can be admitted." Id. at 884. Scientific unanimity is neither required nor, in some instances, desirable. Recently, a California court of appeal rejected the argument that the existence of "two sides" in a scientific discussion requires total exclusion of the evidence, explaining if admissibility required the absence of debate "even Copernicus's theory of a sun-centered solar system could not be mentioned in a court of law. . . . The flat earth society would carry the day." People v. Soto, 35 Cal.Rptr.2d 846, 856 (Cal.Ct.App. 1994). See also People v. Leahy, 882 P.2d at 329 (Cal. 1994) ("First, we should make clear `general acceptance' does not require unanimity, a consensus of opinion or even majority support by the scientific community."); People v. Wesley, 633 N.E.2d 451, 454 (N.Y. 1994)("This Court has noted that the particular procedure need not be `unanimously endorsed' by the scientific community but must be `generally acceptable as reliable.'").
This court has adhered to the "generally accepted" standard. We have not implicitly created a more demanding standard by changing the meaning of the adjective "generally." See, e.g., Commonwealth v. Devlin, 310 N.E.2d 353, 356 n. 3 (Mass. 1974) (requiring "substantial authority for" acceptance within the scientific community). Nor are we of the view that general acceptance requires the trial court to "count scientific noses," an often voiced criticism of Frye. See United States v. Downing, 753 F.2d 1224, 1238 (3d Cir. 1985); but see People v. Perryman, 859 P.2d 263, 268 (Colo.App. 1993) (explaining reliance on the testimony of a single witness is appropriate when "the witness is qualified through knowledge, skill, training, education or experience to render an opinion as to the general acceptance of the techniques . . . .")
"Generally" accepted means accepted "in a reasonably inclusive manner." Webster's 3d New International Dictionary 954 (1986). Under Frye, this requires the trial court to give qualitative consideration to the evidence presented in court, the state of science ascertainable from scientific commentary and journals, and rulings from other jurisdictions which consider the same admissibility questions. See Paul C. Giannelli, The Admissibility of Novel Scientific Evidence: Frye v. United States, a Half-Century Later, 80 Colum. L. Rev. 1197, 1215-19 (1980).
Despite the criticisms leveled at Frye, this standard is not far removed from evaluation required under FRE 702. Under FRE 702 the court must still screen the evidence to ensure "any and all scientific testimony or evidence is not only relevant, but reliable." Daubert, 113 S. Ct. at 2795 (emphasis supplied). An inquiry into the reliability of the science assures the evidence is indeed "scientific, technical or specialized knowledge." See State v. Alberico, 861 P.2d 192, 203 (N.M. 1993) (Several factors could be considered by a trial court in assessing the validity of a particular technique to determine if it is `scientific knowledge' under Rule 702 including general acceptance (citing Daubert, 113 S. Ct. at 2790)); see also Thomas J. Mack, Scientific Testimony after Daubert: Some Early Returns from Lower Courts, Trial 23, 24 (Aug. 1994)("The science that judges previously considered generally accepted is now scientifically valid and reliable. The science that previously was found not to be generally accepted now appears to be lacking in scientific validity." (citations omitted)). Nor does the Frye test exclude evidence from a relevancy weighing. CRE 401 and 403 apply equally to scientific evidence deemed admissible under Frye which may require exclusion if its probative value is substantially outweighed by the danger of unfair prejudice. See People v. Anderson, 637 P.2d 354, 361 (Colo. 1981); see also Kenneth S. Brown, et al., 1 McCormick on Evidence § 703, at 868 (4th ed. 1992).
IV. LEGAL ANALYSIS
"[I]t is the task of an appellate court reviewing a Frye determination to assess whether novel scientific evidence was generally accepted in the relevant scientific communities at the time it was offered into evidence at trial." Fishback, 851 P.2d at 891. While this question involves factual considerations, the determination of whether general acceptance existed to admit the evidence is a question of law. Accordingly, we review the trial court's application of the Frye test de novo. See Giannelli, The Admissibility of Novel Scientific Evidence, supra at 1222-23 (explaining that while the qualification of an expert is within the sound discretion of the trial judge, the qualification of the "process" is better viewed as a matter of law); see also People v. Watson, 629 N.E.2d 634, 640 (Ill.Ct.App.) (finding of scientific acceptance a matter of law for the appellate court), appeal denied, 642 N.E.2d 1299 (Ill. 1994).
A. Scientific theory and techniques under consideration
Central to the application of the Frye test is the identification of the scientific theory and techniques under scrutiny. Were we to broadly define the scientific theory to relate to RFLP analysis our inquiry would end. Similarly, if we focus too narrowly and consider only the mathematical bases for the statistical calculations, further examination is unnecessary. Indeed, the People raise the question whether the probability calculations based on undisputed statistical theories rise to the level of a novel scientific technique to require analysis under the Frye test. While we recognize that the statistical formulae applied to calculate allele frequencies are neither novel nor the type of physical science typically associated with Frye analysis, such a narrow view of the frequency calculations ignores the inescapable intertwining of statistics with RFLP analysis. In our view the basic statistical formulae are subsumed within the novelty of DNA typing, and involve the type of manipulation of physical evidence that requires evaluation under Frye. See Campbell v. People, 814 P.2d 1, 8 (Colo. 1991). This view was made clear in Fishback where we left "to the trial courts the initial determination of whether, in light of events which have occurred subsequently to the trial in this case, the method for calculating the statistical frequency of a declared match remains generally accepted." Fishback, 851 P.2d at 895.
Here, the underlying theory in question is whether scientists can quantify the probability that two individuals will have DNA profiles with the same characteristics at highly polymorphic loci using statistical calculations based on known genetic principles. We take care to avoid the misconception that the theory somehow relates to the probability that a defendant committed the crime. See Jonathon J. Keohler, Error and Exaggeration in the Presentation of DNA Evidence at Trial, 34 Jurimetrics J. 21, 31-32 (1993).
The analogous theory in RFLP analysis is that individuals can be identified based on their unique DNA profiles.
The underlying technique used to test the theory requires creation of a database adequate to yield statistically significant allele frequencies for use in probability calculations. If the trial court concludes the theory and techniques are generally accepted, any questions regarding the implementation of the specific techniques go to the weight of the evidence.
B. State of scientific evidence at trial
Like Fishback, this case raises the question of when during the proceedings the Frye analysis must be applied. In Fishback we assessed the acceptance of the evidence with reference to the time the trial court considered its admission. Fishback, 851 P.2d at 891 n. 17. Because the trial court's rulings on admissibility spanned the period of time from late 1988 until early 1990 we consider the status of DNA typing throughout the relevant time period.
We recognize that our view of the relevant timeframe for a Frye analysis has been subject to criticism. State v. Bible, 858 P.2d 1152, 1189 n. 33 (Ariz. 1993). In Bible, the Arizona Supreme court elected to review admissibility of DNA evidence based on the views of the scientific community at the time of appellate review, comparing its approach to a motion picture that captures the evolving events of science, rather than a snapshot at the time of trial. In our view the Arizona court's approach is no less time specific than ours, rather the court elects to take its snapshot at a later point in time. In Fishback we explained Frye requires nothing more than general acceptance at the time the evidence is admitted. Fishback, 851 P.2d at 891. To require otherwise would compel trial courts to speculate regarding evidence that might be introduced if the court was privy to future knowledge. See also People v. Wesley, 633 N.E.2d 451, 453 (N.Y. 1994)("It should be emphasized that the inquiry here is into the reliability of the DNA evidence at the time of the proceedings in this case in 1988 and 1989.").
The Fishback evidentiary hearings were conducted in October, 1989. There, we explained that by mid-1989 courts typically allowed DNA typing evidence absent a showing of gross procedural error. See id. at 890 n. 14. Though we noted that the statistical frequency calculations were not effectively challenged in Fishback, on appeal we were aware of the debate unfolding in this arena. Id. at 894. We nonetheless held that at the time of trial the techniques for calculating statistical frequencies accompanying a declared match were generally accepted within the scientific community. Also, the vast majority of courts ruling on the matter accepted statistical probability calculations and there existed a relative lack of scientific objection. Id.
Our ruling here is influenced by the substantial similarity between the Fishback evidentiary timeframe and the evidentiary rulings in this case. Here, the Frye hearings were conducted in September, 1988, and January 1989, and the motion in limine hearing was conducted in February 1990. With respect to the trial court's rulings on admissibility during the Frye hearings we find Fishback dispositive. Lindsey made a tactical decision not to attack the DNA evidence presented at these hearings. Scientists were generally endorsing the use of DNA typing in a forensic setting and the debate regarding statistical evidence had not yet surfaced. Finally, courts around the country were admitting DNA evidence under the Frye standard.
We now turn to the trial court's evidentiary findings made at the motion in limine hearing. Contrary to Fishback, where evidence of the acceptability of statistical frequency calculations went uncontradicted, see Fishback, 851 P.2d at 893-94, Lindsey's attack on DNA typing was substantial, and followed two basic themes. The first related to whether ethnically diverse databases could properly be used to calculate genotype frequencies. The second challenged the reliability of Cellmark's DNA typing protocols.
We agree with Lindsey that portions of the testimony and the trial court's ruling related to the theory and techniques underlying statistical probability analysis and required evaluation under Frye. While we understand the trial court's reluctance to hold another Frye hearing on issues it had previously ruled on, as we explained in Fishback, reconsideration may be necessary if there is a change in the attitude of the scientific community. Id. at 891 n. 16.
Our acknowledgement that Frye questions were raised at the motion in limine hearing does not, however, render the trial court's ruling invalid. The trial court's previous ruling that DNA typing and statistical probability analysis were generally accepted within the relevant scientific community served as a backdrop for the hearings on the motion. Thus, Lindsey needed to show that a shift in the scientific community's views required the court to reverse its prior ruling.
No doubt exists that scientific debate regarding statistical analysis was emerging during this time period. Indeed, we noted this fact in Fishback. Id. at 894. The emergence of scientific debate does not require exclusion of evidence under Frye. Moreover, the debate did not crystallize until late 1991, over one year after the evidentiary rulings in this case when Science magazine published articles on the potential effects of subgrouping. Compare R.C. Lewontin and Daniel L. Hartl, Population Genetics in Forensic DNA Typing, 254 Science 1745 (Dec. 20, 1991) with Ranajit Chakroborty and Kenneth K. Kidd, The Utility of DNA Typing in Forensic Work, 254 Science 1735 (Dec. 20, 1991). Questions regarding subgrouping were again addressed in the National Research Council Publication DNA Technology in Forensic Science (NRC Report) first circulated in pre-publication copies made available in early 1992, two years after trial. Thus, despite the later criticisms leveled at product rule analysis, when the court ruled on the admissibility of the DNA evidence at the motion in limine hearing the controversy had not attained the dimension recently seen within the scientific community.
The NRC Report also detailed an alternate method of calculating random match probabilities using "ceiling frequencies." Under the NRC approach the product rule is applied by using the maximum occurrence of any allele within all populations without considering ethnic origins. Application of ceiling principles results in the maximum frequency for individual alleles, and theoretically a corresponding increase in the probability of duplication of the entire genotype.
Review of the record also supports the trial court's ruling to admit the evidence. Lindsey's experts raised questions about subgrouping, but failed to present any empirical evidence that subgrouping affected the independence of Cellmark's VNTR loci. Support existed for Lindsey's proposition that individuals marry others with similar ethnic or racial backgrounds, however, Lindsey offered only speculation that endogamous marriage caused Hardy-Weinberg or linkage disequilibrium in Cellmark's database. The People effectively countered that mating decisions based on these loci would be comparable to mating decisions based on bloodtype. Lindsey's experts also hypothesized that the large frequency of homozygous alleles appearing in Cellmark's database was evidence of subgrouping. The People answered with equally reliable experts who testified that this occurrence resulted from shorter frequency alleles running of the agarose gel during electrophoresis.
Lindsey labors under the misconception that the existence of a scientific debate translated into a lack of general acceptance. As previously explained debate alone is insufficient. Because the trial court had previously ruled that the Frye standard was met, and Lindsey did not show a shift in the views of the scientific community we hold that the DNA typing and statistical evidence was properly admitted at the trial level under the Frye test.
C. Allegations of trial court error in Frye analysis
Lindsey argues that errors in the trial court and the court of appeals' legal analyses requires exclusion of the DNA evidence and reversal of his conviction. He specifically questions the court's ruling to admit the statistical probability evidence under the "more relaxed" standard of CRE 702.
Contrary to Lindsey's allegations, the court of appeals recognized that "forensic application of the statistical portion of the evidence" required analysis under Frye. People v. Lindsey, 868 P.2d at 1089. The court concluded, however, that generally accepted techniques for the calculation of random match probabilities existed at the time the evidence was introduced at trial. Id. at 1089-90. We agree with the court of appeals conclusion that the district court proceedings satisfied the standard under both Frye and CRE 702.
Assuming arguendo that the trial court and court of appeals' analysis differed from the parameters for admissibility set forth in Fishback, reversal of the conviction is not required. See People v. Quintana, No. 93SC428 (Colo. Oct. 17, 1994). In Quintana, we explained when a trial court correctly admits evidence albeit under an incorrect rationale we are not required to reverse a conviction. See also People v. Mathes, 703 P.2d 608, 610 (Colo.App.) (admissible evidence does not become inadmissible because a trial court relied on an inappropriate rule of evidence), cert. denied (Colo. 1985); People v. Jenkins, 768 P.2d 727, 730 (Colo.App. 1988) (A defendant's conviction will not be reversed if a trial court reaches the correct result although by an incorrect analysis), cert. denied (Colo. 1989). Here, the trial court's analysis of complex subject matter was not far removed from the approach we provided nearly two years later. Because we conclude that application of Fishback would not warrant a contrary result, we reject Lindsey's argument.
Once evidence is properly admitted disputes over its reliability go to its weight. See Fishback 851 P.2d at 892. In Fishback, we explained challenges to the implementation and execution of accepted techniques go to the weight of the evidence. Id. at 893. Here, much of Lindsey's attack focused on Cellmark's procedures, including the size of Cellmark's database and whether the sampling of a Detroit blood bank was truly random. Lindsey also challenged Cellmark's match criteria, alleged problems existed with its electrophoresis processes and questioned the quality of Cellmark's database autorads. In our view these questions go to Cellmark's implementation of DNA frequency analysis, properly considered by the jury.
D. Current status of DNA typing evidence
Finally, Lindsey argues that the error in admitting the DNA evidence might have been harmless if "in the intervening period, those techniques have become generally accepted." Lindsey relies on the articles listed above and cases in which courts have rejected probability evidence. See, e.g., State v. Bible, 858 P.2d 1152 (Ariz. 1993) (en banc), cert. denied 114 S. Ct. 1578 (1994); People v. Barney, 10 Cal.Rptr.2d 721 (Cal.Ct.App. 1992); United States v. Porter, 618 A.2d 629 (D.C.App. 1992).
As Lindsey suggests, the debate over the effects of subgrouping briefly enveloped the scientific community. A trilogy of decisions from California courts of appeal aptly illustrates the evolution of the statistical analysis debate. In People v. Axell, 1 Cal.Rptr.2d 411 (Cal.Ct.App. 1991), a California court of appeal held both RFLP comparison and frequency calculations using the product rule met Kelly [California's analogue to Frye] requirements despite the then ongoing debate regarding Cellmark's probability analysis. Axell, 1 Cal.Rptr.2d at 431 ("[T]he prosecution showed that the method used by Cellmark in this case to arrive at its data base and statistical probabilities was generally accepted in the scientific community. Any question or criticism of the size of the data base or the ratio pertains to weight of the evidence and not to its admissibility."). The following year, in People v. Barney, 10 Cal.Rptr.2d 731 (Cal.Ct.App. 1992), the court concluded that the debate over the reliability of frequency data had "progressed to a point on the continuum of scientific debate which neither the Axell court nor the two trial courts in the present cases could have anticipated." Barney, 10 Cal.Rptr.2d at 744. The Barney court focused on the Lewontin-Hartle, Chakroborty-Kidd debate featured in Science magazine in late 1991. Id. at 740. As a result the statistical probability evidence was excluded. Id. Most recently, in People v. Soto, 35 Cal.Rptr.2d 846 (Cal.Ct.App. 1994), the court changed direction and concluded testimony regarding random match probabilities met the Kelly requirements. In Soto, the court rejected the Barney analysis explaining, "[w]e now have data showing that population substructuring is not `forensically significant' in estimating the random likelihood of a particular DNA profile." (citing FBI report, Vol. IA, p. 2). As evidenced by these cases the debate seems to have turned full circle as scientific questions are resolved and techniques are refined. See also People v. Wilds, 37 Cal.Rptr.2d 351 (Cal.Ct.App. 1995) (concluding that the "blight" in the scientific landscape over the general acceptance of genetic profiling evidence has passed); but see People v. Venegas, 36 Cal.Rptr.2d 856, 867 (Cal.Ct.App. 1995) (agreeing that genetic profiling evidence and statistical probability evidence are generally admissible, but finding that the FBI failed to perform analysis in accordance with accepted methodology).
Moreover, the calming of the DNA waters is evident in the scientific community. Recently, Eric Lander, a longstanding critic of the use of probability analysis in forensic DNA typing joined with Bruce Budowle to declare the debate over. See Eric S. Lander Bruce Budowle, DNA Fingerprinting Dispute Laid to Rest, 371 Nature 735 (Oct. 27, 1994). The article was written to explain that "the academic debate that continues to swirl about population genetic issues is rooted in a misunderstanding of the NRC report and is, in any case, of no practical consequence to the courts." Id. Lander and Budowle agree the "harmless pastime" of rehashing old debates within the academic community has not benefitted the legal system.
Lindsey relied on Lander's articles DNA Fingerprint on Trial, 339 Nature 501 (June 15, 1989) and Population Genetic Considerations in the Forensic Use of DNA typing, Banbury Report 32: DNA Technology and Forensic Science 143 (1989), in support of the proposition that statistical match calculations are not generally accepted.
Two significant developments influenced Lander's retreat from his opposition to the use of DNA typing in forensic settings. First is the NRC recommendation that forensic scientists use the alternate ceiling method to calculate random match probabilities. He observed that the use of much more conservative match parameters did not create a corresponding reduction in random match probability calculations, explaining "[w]hereas the product rule typically gives four-locus genotype frequencies of about 10 -8 or 10 -9, the ceiling principle pares them back to about 10 -6 or 10 -7. That extreme assumptions [under ceiling principles] have so little effect only underscores the power of DNA typing and the wisdom of taking a conservative approach."
Lander and Budowle focused on six principal objections raised to the use of the ceiling principle:
1. The ceiling principle is premised on the flawed analysis of Lewontin and Hartl that there is a significant population substructure.
2. The ceiling principle is scientifically flawed because it is not used in populations genetics. Moreover, the plan to sample 10-15 representations is unsound.
3. The ceiling principle makes ludicrous assumptions about the possible substructure of a population.
4. The ceiling principle is so conservative that it hampers the courtroom application of DNA fingerprinting.
5. The ceiling principle is not actually guaranteed to be conservative.
6. The NRC report is causing DNA fingerprinting cases to be thrown out of court.
The article answers each of these objections concluding that debate over use of the ceiling principle is purely academic. "The most extreme positions range over a mere two orders of magnitude: whether the population frequency of a typical four-locus genotype should be stated, for example as 10 -5 or 10 -7. The distinction is irrelevant for courtroom use." Id. at 738.
The second major event is the Federal Bureau of Investigation's (FBI) completion of the population surveys recommended by the NRC report. The FBI has studied 25 distinct subpopulations, as well as 50 separate samples from the U.S. Population. See United States Dept. of Justice, Federal Bureau of Investigation, VNTR Population Data: A Worldwide Study, (1993). The study, Lander and Budowle explain, observed only modest variation in loci used in forensic analysis among ethnic groups, and indicated empirically that random matches rarely occur. The study concluded:
(1) that there are sufficient population data available to determine whether or not forensically significant differences might occur when using different population data bases;
(2) that subdivision, either by ethnic group or by U.S. geographic region, within a major population group does not substantially affect forensic estimates of the likelihood of occurrence of a DNA profile;
(3) that estimates of the likelihood of occurrence of a DNA profile using major population group databases (e.g., Caucasian, Black, and Hispanic) provide a greater range of frequencies than would estimates for subgroups of a major population category; therefore, the estimate of the likelihood of occurrence of a DNA profile derived by the current practice of employing the multiplication rule and using general population databases for allele frequencies is reliable, valid, and meaningful, without forensically significant consequences; and
(4) that the data do not support the need for alternate procedures, such as the ceiling principle approach (NRC Report 1992), for deriving statistical estimates of DNA profile frequencies (Budowle et al. 1993a and 1993b, submitted)."
United States Dept. of Justice, Federal Bureau of Investigation, I-A VNTR Population Data: A Worldwide Study 2 (1993).
Thus, we reject Lindsey's contention that the debate continues to "rage". We are mindful that DNA typing will continue to evolve and the techniques will be refined. The refinement of the scientific processes involved in DNA typing does not require us to rule that prior scientific analyses are invalid. To do so would place courts at the caprice of scientific progression and preclude the admissibility of valuable evidence based on those theories and techniques generally accepted at the time of trial.
V.
In summary, we find that the trial court properly analyzed the DNA typing evidence under Frye v. United States and that application of Fishback v. People, 851 P.2d 884 (Colo. 1993), would not produce a different result. Further, we find that the existence of scientific debate does not, by itself, require exclusion of evidence under Frye or Fishback. We hold that DNA statistical frequency analysis was generally accepted by the scientific community at all times relevant to the trial court's decision to admit the DNA typing evidence. The judgment of the court of appeals is affirmed.
JUSTICE ERICKSON concurs in the result only, and JUSTICE LOHR joins in the concurrence.
JUSTICE MULLARKEY concurs in the result only.
. . .
The court with regard to the statistical evaluation finds from all the evidence presented that there is a valid statistical technique to be applied and in the opinion of that expert, this is a procedure which is followed throughout the biological community in analyzing biological statistics and that the population utilized by the expert in this case is sufficient to bring it within the statistical liable — statistical reliability of those using biological statistics in many areas and, therefore, find that [it] is acceptable.
Therefore, the Court finds from the evidence presented that — there is sufficient reliability and general accepted or general acceptance of the techniques involved in both again DNA analysis and the statistical analysis applied to that information to render them helpful to the finder of fact and to far outweigh any potential prejudice to the Defendant.