Brim v. State

Full title: Robert James BRIM, Appellant, v. STATE of Florida, Appellee

Court: District Court of Appeal of Florida, Second District

Date published: Oct 11, 2000

779 So. 2d 427 (Fla. Dist. Ct. App. 2000)

Opinion

No. 2D97-4559.

October 11, 2000.

Appeal from the Circuit Court, Hillsborough County, Debra K. Behnke, J.

James Marion Moorman, Public Defender, and Carol J.Y. Wilson, Assistant Public Defender, Bartow, for Appellant.

Robert A. Butterworth, Attorney General, Tallahassee, and Dale E. Tarpley, Assistant Attorney General, Tampa, for Appellee.

ORDER RELINQUISHING JURISDICTION TO THE TRIAL COURT TO CONDUCT ADDITIONAL EVIDENTIARY HEARING

ALTENBERND, Acting Chief Judge.

Robert James Brim appeals the trial court's order denying his motion for new trial and' declining to set aside his judgments and sentences for two criminal episodes involving sexual battery. This order was entered on remand from the Supreme Court of Florida after its decision in Brim v. State, 695 So.2d 268 (Fla. 1997). On remand, the trial court conducted an evidentiary hearing, relying partially on telephonic testimony, and determined that certain DNA statistical evidence was properly admitted at Mr. Brim's trial in early 1993.

The supreme court has instructed this court to conduct a de novo review of the trial court's order, describing such a Frye ¹ "determination" as a question of law. See Brim, 695 So.2d at 274. In this de novo review, we are to examine "expert testimony, scientific and legal writings, and judicial opinions" to decide whether the scientific principles and procedures relied upon to create such evidence are generally accepted by a relevant scientific community both at the time of trial and today. See Hadden v. State, 690 So.2d 573, 578 (Fla. 1997).

1 See Frye v. United States, 293 F. 1013 (D.C. Cir. 1923).

After considerable research and deliberation, this court concludes that the hearing on remand was insufficient. Moreover, because we are not adopting a rule of evidence in a rule-making proceeding, but are making a case-specific determination affecting Mr. Brim's liberty interests, we further conclude that due process requires specific constraints upon this unusual de novo review. Although the supreme court described this de novo review as addressing an issue of law, our experience with this review convinces us that this "determination" is actually a mixed question of fact and law. See Jackson v. State, 17 S.W.3d 664 (Tex.Crim.App. 2000). See also State v. Baity, 140 Wn.2d 1, 991 P.2d 1151 (2000); State v. Copeland, 130 Wn.2d 244, 922 P.2d 1304 (1996) ( en banc).

While this case has been pending on appeal, we have reviewed many published legal writings and judicial opinions. This is a process that we are trained to perform and with which we have great experience.² This traditional research establishes that DNA evidence is now routinely admitted in trial courts in many states.³ As a result of this research, we are inclined to believe that the statistical evidence in this case probably meets the Frye standard and was properly admitted.⁴ Therefore, our hesitancy to conclude this case today should not be construed as a signal that DNA testing is not admissible in Florida trial courts.

2 We would point out that this body of legal writing is large. More than 300 cases have discussed this specific type of DNA testing. Since September 1, 1999, when this case was orally argued, more than twenty-eight opinions have been written in this country making reference to this specific type of DNA testing. In addition, since Mr. Brim's trial in 1993, more than 120 journal or law review articles have been written on the subject of DNA. Depending upon the level of de novo review required by the supreme court, this case could be an overwhelming task for a panel of appellate judges.

3 . See, e.g., State v. Ferguson, 20 S.W.3d 485 (Mo. 2000); Roberson v. State, 16 S.W.3d 156 (Tex.App. 2000); Bluain v. State, 242 Ga. App. 125, 529 S.E.2d 155 (2000); Johnson v. Commonwealth, 12 S.W.3d 258 (Ky. 1999); State v. Edwards, 750 So.2d 893 (La. 1999); People v. Soto, 21 Cal.4th 512, 88 Cal.Rptr.2d 34, 981 P.2d 958 (1999); People v. Oliver, 306 Ill. App.3d 59, 239 Ill. Dec. 196, 713 N.E.2d 727 (1999); State v. Van Adams, 194 Ariz. 408, 984 P.2d 16 (1999); Alverson v. State, 983 P.2d 498 (Okla. Cr. 1999); Genry v. State, 735 So.2d 186 (1999); People v. Hamilton, 255 A.D.2d 693, 681 N.Y.S.2d 117 (N.Y.App. Div. 1998); Commonwealth v. Blasiola, 552 Pa. 149, 713 A.2d 1117 (1998); Bolin v. State, 114 Nev. 503, 960 P.2d 784 (1998); Commonwealth v. Rosier, 425 Mass. 807, 685 N.E.2d 739 (1997). These cases address a variety of DNA testing techniques and procedures, and all conclude the DNA evidence was admissible either under Frye or Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993) (requiring trial judge to determine if scientific evidence is relevant and reliable by examining four factors: (1) whether underlying scientific theory can or has been tested; (2) whether theory or technique has been subject to peer review and publication; (3) whether technique has known or knowable rate of error; and (4) whether theory or technique is generally accepted in relevant scientific community).

4 . Cases specifically finding testimony concerning statistical probabilities based upon DNA evidence admissible include: State v. Garcia, 3 P.3d 999 (Ariz.Ct.App. 1999); People v. Soto, 21 Cal.4th 512, 88 Cal.Rptr.2d 34, 981 P.2d 958 (1999); State v. Edwards, 750 So.2d 893 (La. 1999); Crawford v. State, 716 So.2d 1028 (Miss. 1998); People v. Hickey, 178 Ill.2d 256, 227 Ill. Dec. 428, 687 N.E.2d 910 (1997); State v. Harvey, 151 N.J. 117, 699 A.2d 596 (1997). See also United States v. Shea, 957 F. Supp. 331 (D.N.H. 1997), aff'd, 159 F.3d 37 (1st Cir. 1998) (admitting probability estimate on DNA evidence, even though database of profiles contained only 148 Caucasians, 145 African-Americans, 94 Southeastern Hispanics, and 96 Southwestern Hispanics); State v. Barone, 328 Or. 68, 969 P.2d 1013 (1998) (affirming admission of statistical evidence, even though state's statistical expert testified margin of error was plus or minus 50% due to small number of comparison samples in database, DNA sample had been degraded, and DNA database did not account for person of mixed race like defendant). But see State v. Schwartz, 447 N.W.2d 422 (Minn. 1989) (limiting admissibility of statistical DNA evidence and expressing skepticism that juries will not place undue weight upon it).

On the other hand, we have experienced extreme difficulty with the process of reviewing scientific literature. Without expert assistance, we question our competence even to locate, much less to comprehend fully, the current technical, peer-reviewed literature relevant to this de novo review. The parties did not include in the record any of the existing scientific documents needed for this review, nor has either party filed any recent scientific literature as supplemental authority. We have obtained and read scientific literature outside our legal record. We confess that we are not entirely clear whether the supreme court intends this court to read this factual literature (1) to assess the scientific content as science, or (2) merely to obtain a general understanding of the science and then measure levels of acceptance or dissonance within the relevant scientific community. We have concluded that the latter approach is correct under a Frye standard.

Although other courts have conducted such independent, fact-intensive, extra-record reviews of technical literature beyond the trained expertise of the judges,⁵ we are deeply concerned that this method does not permit full supreme court review of a district court's work and would not provide the parties with notice of the fact information relied upon by this court to decide their case. We have concluded that it is inappropriate for this court to evaluate or determine the scientific acceptability of such principles and procedures by examining extra-record, nonlegal materials. As a result of this decision, our record is currently inadequate for this court to complete its task.

5 . See, e.g. State v. Tankersley, 191 Ariz. 359, 956 P.2d 486 (1998) (testimony regarding PCR DNA evidence); Black v. Food Lion, Inc., 171 F.3d 308 (5th Cir. 1999) (testimony regarding whether trauma causes fibromyalgia); State v. Coon, 974 P.2d 386 (Alaska 1999) (testimony regarding voiceprint identification); Reed v. State, 283 Md. 374, 391 A.2d 364 (1978) (testimony regarding voiceprint identification); State v. Cauthron, 120 Wn.2d 879, 846 P.2d 502 (1993) (testimony regarding RFLP testing of DNA and statistical analysis of results).

Accordingly, we relinquish jurisdiction to the trial court for an additional evidentiary hearing and for the preparation of a revised order. At the conclusion of this painfully long order, we provide specific guidance to the trial court concerning the issues that should be addressed and the materials that should be included in the record at the evidentiary hearing. Upon entry of the trial court's revised order, this court will conduct an expedited review and conclude this appeal.

I. A Brief Procedural History of this Complex Case.

In 1990, several women were sexually attacked in Tampa, Florida. The police believed that one man was responsible for these attacks. Mr. Brim was eventually arrested and charged with three of these crimes. The cases were designated as circuit court case numbers 91-14399, 92-799, and 92-875.

Only case number 92-875 was actually tried. DNA evidence became the central focus of this case. Mr. Brim sought to exclude the DNA evidence on grounds that this new scientific method was not generally accepted. The trial court conducted a hearing on Mr. Brim's motion in limine to determine whether this evidence was admissible under the Frye test. The trial court denied the motion. The State then presented DNA evidence as a critical portion of its case. The jury found Mr. Brim guilty of burglary of a dwelling with an assault or battery, robbery, and two counts of sexual battery with slight force. The trial court sentenced Mr. Brim on February 18, 1993, to multiple concurrent sentences, the longest of which is life imprisonment as a habitual felony offender.

At sentencing in case number 92-875, Mr. Brim also entered open nolo contendere pleas in case numbers 91-14399 and 92-799⁶ In case number 91-14399, Mr. Brim did not reserve his right to appeal any issue. He received a fifteen-year sentence of imprisonment as a habitual offender, followed by fifteen years' probation. During our deliberations in this case, we severed the appeal in case number 91-14399 and affirmed that sentence. See Brim v. State, 754 So.2d 823 (Fla. 2d DCA 2000). Thus, Mr. Brim is currently serving his sentence in that case while these appellate proceedings remain pending.

6 . Our record contains no documents from these other cases except for the sentencing hearing and the motions in limine that were jointly filed in both cases.

In case number 92-799, Mr. Brim entered a plea reserving the right to appeal the denial of his motion in limine, which was identical to the motion in case number 92-875.⁷ Thus, the outcome of the appeal concerning case number 92-799 will be determined by the outcome of the appeal concerning case number 92-875.

7 . The record does not contain a specific finding that the DNA evidence would have been dispositive in case number 92-875. Nevertheless, the State has stipulated that the trial court made such a finding and that this issue is preserved for review.

In 1993, Mr. Brim appealed all of his judgments and sentences to this court.⁸ In 1995, we affirmed the admission of the DNA evidence, essentially ruling that the results of this specific type of DNA testing were admissible under a Frye analysis and that the statistical presentation of these results did not require any separate Frye analysis. See Brim v. State, 654 So.2d 184 (Fla. 2d DCA 1995)⁹ Thereafter, the supreme court reversed our decision and held that the statistical presentation of the evidence did require a separate Frye analysis. See Brim v. State, 695 So.2d 268, 275 (Fla. 1997).

8 . This court docketed these cases as three separate appeals. Trial court case number 92-799 became appellate case number 93-860. Trial court case number 92-875 became appellate case number 93-863. Trial court case number 91-14399 became appellate case number 93-864. We then issued a consolidated opinion, which traveled to the supreme court as a single case. Thus, this pending appeal is a consolidated appeal that relates back to our earlier cases, 93-860 and 93-863.

9 . We reversed certain sentencing issues in all three cases. It is not clear from our record whether the trial court has ever corrected the errors in trial court case numbers 92-875 and 92-799. During the relinquishment required by this order, the trial court should determine whether those corrections have been made. If not, we expressly grant the trial court jurisdiction to correct those errors.

The supreme court did not, however, reverse Mr. Brim's judgments and sentences when it disagreed with this court. Instead, it ordered that the case be remanded for another Frye hearing to reconsider the admissibility of the DNA statistical evidence. Upon receipt of the supreme court's mandate this court remanded the case to the trial court. The trial court conducted a limited evidentiary hearing on the admissibility of the DNA statistical evidence and determined that the evidence had been properly admitted in case number 92-875.

At that point, Mr. Brim's counsel attempted to appeal the trial court's ruling directly to the supreme court. Because the trial court's new order had been entered on instructions from the supreme court and that court had not fully resolved the issues on appeal, that decision was not entirely illogical. The supreme court, however, had not merely relinquished jurisdiction; it had issued its mandate. Thus, it transferred the case to this court as a new appeal in which this court must decide whether the trial court satisfied the supreme court's mandate. Today, we rule that the trial court must conduct an additional hearing because it did not make a sufficient record of the "exact methods used by the State in calculating its population frequency statistics" as required by the supreme court. See Brim, 695 So.2d at 275. Indeed, the record currently provides little to assist this appellate court with the unusual de novo review of this issue as mandated by the supreme court. This is not intended as criticism of the trial court, which clearly has attempted in good faith to accomplish an unusual and difficult task.

II. The Importance of DNA Evidence in this Case.

At the outset of our analysis, it is worthwhile to emphasize the critical role DNA played in this case. During the trial in case number 92-875, no one disputed that an African-American male, probably in his late twenties or earlier thirties, who was of average height and weight, had entered the victim's home during the night. He then raped and robbed her. The victim had the presence of mind not to disturb or destroy important physical evidence. A small sample of semen was discovered on her house dress.

The victim assisted in the preparation of a composite drawing of the perpetrator and reviewed photo packs including Mr. Brim's photograph. Although she believed that Mr. Brim had physical characteristics similar to those of the perpetrator, she was never willing to positively identify him as the rapist The State did not attempt to use any Williams ¹⁰ rule evidence to support its case against Mr. Brim. He made no incriminating statements. His fingerprints were not discovered at the scene of the crime, nor was any additional physical evidence connecting him to this crime ever located. Thus, but for the DNA evidence, the State may not have had enough evidence to convict Mr. Brim of this crime.

10 . See Williams v. State, 110 So.2d 654 (Fla. 1959).

III. The DNA Evidence at Trial.

In this case, a sample of semen from the crime scene was compared to a sample of blood obtained from Mr. Brim. The State's laboratory used a method known as restriction fragment length polymorphism (RFLP) to test the samples. Details of this method will be discussed later in this order. The tests did not exclude Mr. Brim as a possible perpetrator. Instead, the tests located five different segments of Mr. Brim's DNA that appeared identical to the DNA from the crime scene. At Mr. Brim's trial, the statistical analysis of the experts varied, but all provided statistical testimony. The State's experts opined that only 1 in 3.5 billion humans of African descent, randomly selected, would be predicted to have the same DNA as the crime scene sample at all five of these locations.¹¹

11 . In this court's first opinion and in the supreme court's opinion, this statistic is stated as 1 in 2.5 million. See Brim v. State, 654 So.2d 184, 185 (Fla. 2d DCA 1995); Brim v. State, 695 So.2d 268, 271 (Fla. 1997). Although the smaller statistic does appear on one occasion in the record, we are inclined to believe that it was either a slip of the tongue by a witness or a mistake by the court reporter because the estimate is elsewhere repeatedly stated as 1 in 3.5 billion. During the hearing on remand from the supreme court, the testimony established that the correct statistic in 1993 was 1 in 3.5 billion.

In his case, Mr. Brim provided testimony from a well-qualified expert who disagreed with the statistical methodology employed by the State's experts. He maintained that this very new technology was m such an early stage and that the scientific community had tested so few samples that the relevant scientific community could not extrapolate statistical estimates beyond the known population of test samples. Using this very conservative "counting method," he could state with certainty only that the likelihood of finding another person with DNA that appeared to be identical to Mr. Brim's at the five locations was at least 1 in 960. Even using a less conservative "modified ceiling principle" that did make predictions beyond the known samples — but adjusted those predictions downward to avoid risks that the five locations on the DNA were not truly independent of one another — he still concluded that the probability of finding DNA that matched Mr. Brim's was no higher than about 1 in 9,000. Obviously, when a jury must decide a question beyond a reasonable doubt, there is a huge difference in the weight of the evidence depending upon which of these calculations is admitted into evidence by the trial judge or, in the alternative, depending upon which calculation the jury accepts if it is permitted to choose among alternative statistical approaches. In Mr. Brim's trial, the jury was presented with statistics by both the State and Mr. Brim and thus was permitted to choose between them. Apparently, the jury was persuaded by the State's presentation.

IV. Conducting a Frye Hearing in a Trial Court after the Supreme Court's Decision in Brim.

Before discussing the evidence presented at the hearing on remand, we conclude that it is useful to briefly discuss the general nature of such a hearing and the difficult standard of review to which it is subjected in the appellate court. We have discovered that the de novo standard of review requires the litigants and the trial court to give careful attention to both the scope of the hearing and the content of the record generated by the hearing.

When this case returned to the trial court in 1997, the primary unresolved questions were (1) whether the State's statistical estimate had been based on scientific principles and procedures that rendered the estimate admissible under the Frye test at the time of trial, and (2) whether that statistical estimate remained admissible at the time of the hearing on remand.¹²

12 . Footnote 5 in the supreme court's opinion suggests that the alternative statistical approaches of the defense expert at trial were properly admitted. See Brim, 695 So.2d at 273 n. 8; cf. Clark u State, 679 So.2d 321 (Fla. 3d DCA 1996) (barring use of ceiling principle). Nevertheless, the admissibility of those statistical methods was not an issue on appeal, and the trial court properly did not consider further evidence about those methods on remand.

In its decision, the supreme court explained the Frye test in the following manner:

We start by emphasizing again that the Frye test is utilized in Florida to guarantee the reliability of new or novel scientific evidence. E.g., Stokes v. State, 548 So.2d 188 (Fla. 1989). Despite the federal adoption of a more lenient standard in Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), we have maintained the higher standard of reliability as dictated by Frye. E.g., Ramirez v. State, 651 So.2d 1164 (Fla. 1995). This standard requires a determination, by the judge, that the basic underlying principles of scientific evidence have been sufficiently tested and accepted by the relevant scientific community. To that end, we have expressly held that the trial judge must treat new or novel scientific evidence as a matter of admissibility (for the judge) rather than a matter of weight (for the jury). In Ramirez, we wrote:

In utilizing the Frye test, the burden is on the proponent of the evidence to prove the general acceptance of both the underlying scientific principle and the testing procedures used to apply that principle to the facts at hand. The trial judge has the sole responsibility to determine this question. The general acceptance under the Frye test must be established by a preponderance of the evidence.

Id. at 1168.

Brim, 695 So.2d at 271-72 (footnote omitted).

The supreme court further stated that "two conflicting principles or theories cannot simultaneously satisfy the Frye test," but it recognized that courts may allow "multiple reasonable deductions when all are based on generally accepted principles of population genetics and statistics." Brim, 695 So.2d at 272-73. In other words, the trial court cannot present the jury with two competing scientific theories that are irreconcilable with one another, but it can admit several approaches to a scientific problem if all approaches are generally accepted within the relevant scientific community and are based upon some accepted core scientific principle.¹³

13 . By analogy, when explaining a student's score on an examination, a teacher might compare that student's score to a "typical" student's score. To accomplish this, the teacher might rely upon the classroom's average score, the mean score, or the mode. Each method might produce a slightly different answer, but all three methods would be generally accepted statistical approaches used to measure central tendency. See, e.g., John Monahan Laurens Walker, Social Science in the Law: Cases and Materials 66-67 (3d ed. 1994).

The court explained that general acceptance is a finding that the trial court makes by evaluating both the quality and quantity of the evidence supporting the technique:

It is clear that scientific unanimity is not a precondition to a finding of general acceptance in the scientific community. People v. Dalcollo, 282 Ill. App.3d 944, 218 Ill. Dec. 435, 445, 669 N.E.2d 378, 387 (1996). Instead, general acceptance in the scientific community can be established "if use of the technique is supported by a clear majority of the members of that community." People v. Guerra, 37 Cal.3d 385, 208 Cal.Rptr. 162, 183, 690 P.2d 635, 656 (1984). "Of course, the trial courts, in determining the general acceptance issue, must consider the quality, as well as quantity, of the evidence supporting or opposing a new scientific technique. Mere numerical majority support or opposition by persons minimally qualified to state an authoritative opinion is of little value. . . ." People v. Leahy, 8 Cal.4th 587, 34 Cal.Rptr.2d 663, 678, 882 P.2d 321, 336-37 (1994). Therefore, while a "nose count" is not alone sufficient to establish general acceptance in the scientific community, such acceptance likewise need not be predicated upon a unanimous view.

Brim, 695 So.2d at 272.

Finally, the court ruled that a Frye "determination" is a "question of law" that must be decided by the trial court "based on that method's general acceptance within the relevant scientific community at the time of the hearing." Id. at 274-75.

It appears to this court that a trial judge involved in a Frye hearing must listen to the scientific evidence and resolve any disputed questions of fact using the same methods employed in any other non-jury hearing. Then the trial court makes a "decision of law" that the particular scientific principle or the testing procedure at issue is generally accepted by the relevant scientific community. In other words, reaching conclusions about the definition of the relevant scientific community and its general acceptance. of a principle or procedure is a judicial function rather than a scientific question. Although the supreme court describes this final step as a question of law, the final decision in a Frye hearing is similar to the final decision in a Fourth Amendment suppression hearing, which has been described as a mixed question of fact and law. See Ornelas v. United States, 517 U.S. 690, 116 S.Ct. 1657, 134 L.Ed.2d 911 (1996) (holding determinations of reasonable suspicion to conduct stop and probable cause to perform search should be reviewed de novo as mixed question of law and fact).

Although the Frye standard may be designed to "guarantee the reliability" of new scientific evidence, the trial judge is not actually called upon to determine whether various principles and procedures are "reliable" from a scientific perspective.¹⁴ A scientist would claim that a scientific procedure was "reliable" if its results could be consistently replicated.¹⁵ The procedure or principle would be "valid" if it actually measured what it claimed to measure. Unlike the Daubert standard,¹⁶ the Frye standard does not call upon the trial judge to read the scientific literature to understand or assess the scientific reliability or validity of a principle or procedure. The trial judge reviews this literature merely to determine whether there is quantitative and qualitative acceptance of the science. See State v. Garcia, 3 P.3d 999 (Ariz.Ct.App. 1999); Copeland, 922 P.2d at 1312. The trial judge is determining legal reliability, as a threshold test of legal relevance, by judging — as an objective outsider — the level of acceptance that a principle or procedure has achieved within a scientific community. See United States v. Yee, 134 F.R.D. 161, 196 (N.D.Ohio 1991). Thus, although the judge must develop at least a solid, layperson's understanding of the science, she is not expected to perform the superhuman task of transforming herself into an expert within the community. She is merely to position herself to determine the level of agreement or dissension within it.

14 . See United States v. Yee, 134 F.R.D. 161 (N.D.Ohio 1991).

15 . See Social Science in Law: Cases and Materials § 2, at 42.

16 . See Daubert, 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469. For articles comparing the Daubert approach to the Frye approach, see Adina Schwartz, A "Dogma of Empiricism" Revisited: Daubed v. Merrell Dow Pharmaceuticals, Inc. and the Need to Resurrect the Philosophical Insight of Frye v. United States, 10 Harv. J.L. Tech. 149 (Winter 1997); Thomas Lyons, Frye, Daubert and Where Do We Go From Here? 45. R.I.B.J. 5 (Jan. 1997); Comment, Lawrence B. Ebert, Frye after Daubert: The Role of Scientists in Admissibility Issues as Seen Through Analysis of the DNA Profiling Cases, 1993 U. Chi. L. Sch. Roundtable 219 (1993). For a compilation revealing which courts applied each standard as of 1995, see Joseph R. Meaney, From Frye to Daubert: Is a Pattern Unfolding?, 35 Jurimetrics J. 191 (Winter 1995).

We emphasize that the trial court should base its assessment of legal reliability at a Frye hearing upon the scientific evidence in its record. We do not believe that the supreme court's opinion in Brim authorizes a trial court to embark on a personal scientific exploration outside the record to reach its decision. In addition to the due process concerns inherent in such an endeavor, we would have no reliable record to examine on review if the trial court utilized such a method. The trial court in this case properly did not venture outside the confines of the record presented to it.¹⁷

17 . One of the parties gave the trial court a packet of scientific literature to review. Unfortunately, this was never placed in our record on appeal.

V. Appellate Review of a Frye Determination After Brim.

The supreme court's opinion in this case requires that we conduct a de novo review of the trial court's decision. The supreme court rejected an abuse-of-discretion review, primarily because that "would prohibit an appellate court from considering any scientific material that was not part of the trial record in its determination of whether there was general acceptance within the relevant scientific community." Brim, 695 So.2d at 274. Thus, the supreme court itself reviewed and relied upon the 1996 report by the National Research Council (NRC),¹⁸ to determine that "ceiling principles" in the calculation of population frequency statistics are no longer necessary. See Brim, 695 So.2d at 273.¹⁹ This report was not in existence at the time of Mr. Brim's trial, nor was it in the record even when the case came to this court after the remand from the supreme court.

18 . The full designation for this report is The Evaluation of Forensic DNA Evidence, prepared by the National Research Council's Committee on DNA Forensic Science in 1996. Because this publication is an update of a similar report prepared in 1992, it is hereinafter referred to as the "NRC II" report. The 1992 report of the same name will be hereinafter referred to as the "NRC I" report.

19 . We note that the Third District, prior to the supreme court's decision in Brim and without much analysis, held that testimony based on a ceiling principle was inadmissible. See Clark v. State, 679 So.2d 321 (Fla. 3d DCA 1996).

As we explained at the beginning of this order, we have struggled for nearly a year with our authority and competence to make a de novo "determination" regarding the general acceptance of a very technical, complex scientific procedure within some unspecified scientific community. We do not quarrel with the need for a healthy and thorough independent review by an appellate court of a Frye determination made by a trial court. However, both due process and the limited technical competence of the judiciary require that this review take place with certain safeguards that we have not yet provided. Our relinquishment in this case is necessitated in significant part, not by an error on the part of the trial court, but by our decision that we cannot conduct an independent and undisclosed investigation to determine what some scientific community may or may not have decided about the calculation techniques used in determining and reporting DNA population frequencies.²⁰

20 . We note that the supreme court has rulemaking authority and is the final arbiter on issues of Florida law. To the extent that a Frye determination establishes a new rule of evidence of general applicability, it may be that the supreme court can and should look beyond its record. The same cannot be argued for a district court. This may be a reason to consider certifying major Frye issues to the supreme court pursuant to Florida Rule of Appellate Procedure 9.125.

The Frye standard is not a direct measure of scientific trustworthiness. See Yee, 134 F.R.D. 161. Instead, it is based on the assumption that the science will be trustworthy if scientists worthy of trust have published articles and made public statements in support of the scientific principle or procedure. Frye does not contemplate review of the scientific literature to determine whether the science in this case is "good" science or is reliable from a scientific perspective. Such a review would simply be a Daubert analysis at the appellate level.²¹

21 . See Daubert, 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469.

Thus, when this case returns to us following the relinquishment, the parties may supplement the record with updated scientific literature, but only for the purpose of measuring levels of acceptance or disagreement within the relevant scientific community.²² We will then conduct a new examination of the same legal issues presented to the trial court at the Frye hearing, restrained only by our absence from the trial courtroom when the live testimony was presented.

22 . There are only two significant cases conducting a de novo review of a Frye determination since the supreme court's decision in Brim. See E.L DuPont De Nemours Co. v. Castillo, 748 So.2d 1108 (Fla. 3d DCA 2000), cert. granted, No. SCOO-490, 770 So.2d 156 (Fla. Aug. 31, 2000); Berry v. CSX Transp., Inc., 709 So.2d 552 (Fla. 1st DCA 1998). Both of these cases were civil appeals in which the parties apparently provided the district court with all of the necessary scientific literature. Thus the due process concerns raised by this type of review were not as pronounced in those cases. But see State v. Sybers, 743 So.2d 619 (Fla. 1st DCA 1999).

VI. The Hearing on Remand.

When this case returned to the trial court, the supreme court had already taken "judicial notice that DNA test results are generally accepted as reliable in the scientific community, provided that the laboratory has followed accepted testing procedures." Hayes v. State, 660 So.2d 257, 264 (Fla. 1995). The two prior opinions in this case further clarified that the basic principles and procedures in RFLP testing are also generally accepted methods²³ See Brim, 695 So.2d at 271. The supreme court had also decided that some method or methods of statistical analysis are appropriate and generally accepted as a means to derive admissible "deductions" about the information provided by RFLP testing. Id. at 272-73. What was unresolved in this case when the matter returned to the trial court was whether the methods the State's experts utilized in calculating their population frequency statistics fell within a generally accepted method in the relevant scientific community. Although this required, at a minimum, a determination by the trial court that certain population frequency tables were generally accepted, such a determination is not as straightforward as it might seem.

23 . In Hayes v. State, 660 So.2d 257, 264 (Fla. 1995), as in this case, the relevant DNA testing procedure was RFLP. There are other methods of DNA testing not yet addressed by judicial opinions in Florida, such as polymerase chain reaction (PCR). One court has represented that PCR is less reliable in its conclusions, but allows the extraction of DNA from a much smaller sample. See Campbell v. State, 910 S.W.2d 475, 478 n. 6 (Tex.Crim.App. 1995) ( en banc) (affirming admission of PCR DNA testing results). A trial court faced with a method of DNA testing other than RFLP could not rely upon the supreme court's opinions in Hayes or Brim to determine the admissibility of the DNA evidence. See Murray v. State, 692 So.2d 157 (Fla. 1997) (reversing admission of PCR DNA testing results). Indeed, as this area of science progresses, newer methods may cause the relevant scientific community to reassess its acceptance of the RFLP testing method. See generally Christopher H. Asplen, "Integrating DNA Technology into the Criminal Justice System," vol. 83, no. 3, Judicature 144, 148 (1999).

On remand, the trial court held a ninety-minute hearing on August 4, 1997, at which one of the two key State witnesses testified telephonically.²⁴ The witnesses verbally explained how they calculated the probabilities in this case and how certain probability tables or population frequency tables were established in Orlando for use with the Caucasian population. These Orlando tables, however, were of only marginal relevance to the statistical calculation applied to Mr. Brim, an African-American. No testimony was provided concerning how the FBI created the population frequency tables used to calculate the African-American statistical evidence specifically applicable to Mr. Brim. The basic formula that the experts described verbally was never introduced as an exhibit; thus this court has been unable to examine the formula. The State did not introduce into evidence the population frequency tables used to insert probabilities into the basic formula. Thus, this court lacks the basic information needed to understand how the experts reached the 1 in 3.5 billion statistic.

24 . The only defense expert also appeared telephonically.

One State expert testified that the procedures accepted by the relevant scientific community had changed since 1993 and, if the trial had occurred in 1997, the testimony provided under the 1996 NRC guidelines would have been 1 in 2.7 billion African-Americans, rather than 1 in 3.5 billion. See National Research Council, Committee on DNA Forensic Science, The Evaluation of Forensic DNA Evidence (National Academy Press 1996) (hereinafter "NRC II"). The trial judge was understandably uncertain about how such a change should affect her analysis.

At the hearing, the trial judge repeatedly asked the parties whether the testimony they were providing was what the supreme court expected the trial court to consider. It is clear that the trial judge had doubts that this testimony provided the "exact methods used" by the State in calculating its statistics. The assistant state attorneys convinced her that the additional testimony was sufficient We conclude that it was not.

Before we can explain what the trial court should consider during the relinquishment, a simplified explanation of the two prongs of RFLP testing and statistical analysis is necessary.

VII. Prong One: Creating the Data for Statistical Analysis.

In our first opinion, this court attempted to explain the restriction fragment length polymorphism method of DNA testing (RFLP) by dividing the process into three parts. See Brim, 654 So.2d at 185. The supreme court summarized the process by dividing it into two parts. See Brim, 695 So.2d at 269. It explained that the processing of DNA evidence-when used to present a probability that the defendant is the perpetrator²⁵ — can be roughly divided into one prong that creates data in the laboratory and a second statistical prong that analyzes the data.

25 . The statistical prong is not involved when the steps taken in the laboratory exclude the suspect as a possible perpetrator. Exclusion of a suspect involves a laboratory determination that the two samples of DNA are distinctively different. Barring an error in the laboratory results, such a determination is not a probability or prediction, but a categorical statement requiring no statistical evaluation.

Oh remand, both the experts and the trial court had difficulty dividing this process into two distinct parts. It is clear that the experts in this field divide the overall process into a lengthy protocol with numerous steps consuming weeks of time. Cautious scientists trained to question all assumptions prefer to render opinions identifying specific scientific principles and procedures that are integral to this DNA predictive process and are generally accepted within the relevant scientific community. They do not wish to lump them all together to make a blanket statement of general acceptance.

Likewise, we have come to realize that the general acceptance of the statistics cannot be assessed without considering the acceptance of certain specific methods and procedures employed in the laboratory, especially at the point when the FBI created the population frequency tables. We have discovered that some methods used during the laboratory prong, while not affecting the general acceptance of the laboratory work or DNA as a method of exclusion, are crucial to the accurate creation of the data used to establish the population frequency tables. Those tables in turn are used to calculate the probabilities in this and other cases. Thus, we cannot fulfill the supreme court's mandate without reviewing some of the laboratory methods.

A Frye analysis does not require a judge to examine each and every minute aspect of a new or novel, scientific technology that produces potential legal evidence. However, a cautious judge determining legal reliability of such a complex procedure needs to assess the science in a logical, structured analysis that does not risk overlooking some critical principle or procedure. All of these concerns cause us, for purposes of this order, to explain the supreme court's "first prong" of restriction fragment length polymorphism (RFLP) testing by dividing it into eight steps.²⁶ 1. Collection of the Physical Evidence.

26 . These steps are similar to those described in the Reference Manual on Scientific Evidence. See Judith A. McKenna et al., "Reference Guide on Forensic DNA Evidence," Reference Manual on Scientific Evidence (Fed. Judicial Ctr. 1994). For other judicial attempts to explain this process, see People v. Castro, 144 Misc.2d 956, 545 N.Y.S.2d 985 (N.Y.Sup.Ct. 1959); Yee, 134 F.R.D. 161; State v. Anderson, 118 N.M. 284, 881 P.2d 29 (1994).

First, the physical evidence that may contain DNA is collected at the crime scene. In this step, samples of blood or other DNA source material are also collected from the victim, from the defendant, and from any other human whose DNA is to be compared in the testing. Unlike newer polymerase chain reaction methods, RFLP testing requires a relatively high-quality crime scene sample containing many cells with DNA.²⁷ Although evidence might be excluded in a specific case due to defects in this step, we are not presented with any issue warranting a Frye hearing concerning this initial step. Thus the trial court need not address this-step during the relinquishment.

27 . See Campbell, 910 S.W.2d 475.

2. Laboratory Extraction of Pure DNA for Testing .

The human genetic code is carried in 23 pairs of chromosomes made from DNA. Chromosomes, and thus DNA, are found inside the nucleus of every human cell containing a nucleus.²⁸ DNA is an extremely long and complex molecular structure. It is constructed using only four nucleotides or bases, identified as A, T, C, and G. These nucleotides are connected in pairs to create long strands in the shape of a double helix. It may be helpful to think of this as a twisting ladder, though the structure is actually more complex: There are approximately 3 billion nucleotide pairs in a set of 23 chromosomes.²⁹

28 . See NCR II at 12. Eggs and sperm have half the DNA of normal cells and combine to create a complete DNA structure.

29 . See NRC I at 3.

RFLP testing requires that the DNA chemical substance be extracted from the various sources of physical evidence. For accurate testing, it is important that the sources be relatively pure and undamaged. This step involves a series of sophisticated chemical procedures. As a result of the supreme court's opinion in this case, there is no question that the process the State used in this step is a generally accepted procedure within the relevant scientific community.³⁰ During the relinquishment, the trial court does not need to consider any additional evidence concerning this step.

30 . We would note that the relevant scientific community that must be satisfied for this step may be different than the community involved in the statistical analysis of DNA for forensic purposes in criminal trials. See "Reference Guide on Forensic DNA Evidence" at 286. The processing of DNA is done for a wide variety of medical, research, and forensic purposes. There appears to be no current disagreement that the process used to isolate pure DNA is acceptable to this broad scientific community.

3. Fragmenting the Pure DNA

As its name suggests, the RFLP method requires that the strands of DNA be broken into fragments by a "restriction" process. This process essentially cuts the molecular strands of DNA into many pieces, both horizontally and vertically.³¹ In this step, each strand of DNA is fragmented or cut using an enzyme called HAE-III.

31 . See NRC I at 36-37.

The vertical cut divides the pairs into two separate "half-ladders." With the knowledge that the DNA double helix is constructed using only the four bases, A, T, C, and G, scientists have determined that in the restriction process, the HAE-III enzyme cuts a strand horizontally between C and G bases whenever the strand contains the combination GGCC. This restriction process causes the DNA strand to be broken, into many pieces of varying lengths. Some of these pieces are short and contain only a few bases or nucleotides. Others are longer and contain a large number of nucleotides. Again, after the supreme court's decision in this case, it is clear that this step passes the Frye test The trial court does not need to consider any additional evidence concerning this step.

4. Electrophoresis.

When fragments of DNA are subjected to an electric field, they are drawn by the current somewhat like iron filings are attracted to a magnet Because of this characteristic, the fragmented strands of DNA can be placed on a sheet of electrophoresis gel, subjected to an electric current, and allowed to move or migrate for approximately sixteen hours. Shorter pieces migrate faster and farther than longer pieces. At the end of the sixteen hours, a sample of DNA material is dispersed across the gel in a broad smear in which the longest pieces are at the near end and the shortest pieces are at the far end. Pieces of the same length will tend to congregate together on the gel.

When electrophoresis is performed in forensic cases, several different samples of DNA are placed on one sheet of gel. Each sample is placed in its own lane of travel. One sample is always a known control sample to assure that it distributes in the expected manner. If it distributes as expected, this is an indication that the process has been properly performed. The other samples will include the DNA from the crime scene and from one or more of the suspected perpetrators.

In order to determine the lengths of the fragments, a sizing ladder is placed in the gel. This is essentially a ruler that is marked in nucleotide lengths. It is expected that the electric current will cause virtually all of the fragments of a specific length to move to that portion of the gel that aligns with the marker for that length. In this case, the laboratory used a "Lifecodes 23 KD" ladder.

Again, after the supreme court's decision in this case, it. is clear that this step, in general, passes the Frye test. From our review of the literature, however, we are concerned that the specific ladder used can ultimately affect the statistics presented at trial. Mr. Brim's short motion in limine attacked DNA profiling evidence in very broad terms. At this point, we are uncertain whether he intends to challenge the use of the ladder. We do not suggest that we have located any disagreement within the scientific community concerning this ladder. On the other hand, we have located no express agreement about its use. During the relinquishment, if Mr. Brim maintains that he has a reasonable basis to challenge the general acceptance of this specific ladder or the ladder used in the FBI testing that created the population frequency tables, the State must establish that the ladders were generally accepted by the relevant scientific community at the time of the trial and that, even if more sophisticated ladders are in current use, subsequent research has not discredited the accuracy of the ladders.

5. Southern Blotting.

When the electrophoresis procedure is finished, the dispersed DNA fragments in the electrophoresis gel are then transferred to a nylon membrane by a process known as Southern Blotting.³² The DNA firmly attaches to the membrane so that it can be further tested and processed. After the supreme court's decision in this case, it is clear that this step passes the Frye test. The trial court does not need to consider any additional evidence concerning this step on relinquishment.

32 . Edwin Southern developed the process in 1975. See NRC I at 35.

6. Selection of Multiple Probes.

Given the similarity of all humans, it is not surprising that many locations along the strands of DNA contain identical sequences for all humans. In fact, scientists estimate that approximately 99% of the 3 billion nucleotide pairs fall into this category. These locations are referred to as "monomorphic" and are not useful for discriminating one human from another.³³ Other parts of the strands, however, exhibit considerable variability and are referred to as "polymorphic." Some of these polymorphic locations along the DNA strand are responsible for specific human traits, such as eye color. Others, however, perform no known function and are sometimes referred to as "junk" DNA.

33 . See "Reference Guide on Forensic DNA Evidence" at 282.

Within some of the polymorphic DNA that performs no known function, there are locations that exhibit a tendency to repeat a short pattern of nucleotides or "base pair sequences." The number of repetitions varies considerably among different people. These variable numbers of tandem repeats (VNTR), or paired repeats, allow the scientists to develop chemical "probes" that demonstrate the differences among humans at these locations in the DNA.

Because scientists know, as a matter of chemistry, that the nucleotide A always bonds or pairs with T, and the nucleotide G always pairs with C, they can create a chemical probe, containing a known sequence of nucleotides, that will always bond only with a complementary sequence of nucleotides. The chemical probe thus contains a small strand of about 20 to 30 nucleotides that can be envisioned as an extremely short "half ladder" of DNA. Scientists have developed a number of such probes that are designed to locate DNA fragments containing VNTRs.³⁴

34 . This description applies only to single-locus probes, which are the type used in this case. Multi-locus probes are more complex and may not be generally accepted. See NRC I at 56.

This sixth step of this overall process requires the laboratory to select specific probes, developed by other scientists, to use in its testing of the DNA fragments on the nylon membrane. It is our impression that the laboratory involved in this case had a standard policy of using the same five probes in all of its testing. The probes used were purchased commercially. In the record they are described as (1) YNH (also described as 2-Y and H-24) purchased from Promega Corporation, (2) TBQ purchased from Promega Corporation, (3) D-1 (also described as B-1) purchased from Lifecodes Corporation, (4) MS-1 purchased from Celmar Corporation, and (5) PH-30 purchased from BRL Corporation. The record indicates that other probes exist, such as TBQ-7.³⁵

35 . The record contains little information about these specific probes, and we are not certain that the record reflects accurate names for these probes.

We describe this step in some detail because, as more fully explained later, the selection of probes directly affects the final statistical analysis. We do not regard the supreme court's decision as specifically approving each of these probes. Nothing in the trial record or in the record on remand establishes that each of these probes and the population frequency tables derived from each are generally accepted in the relevant scientific community. The supreme court instructed the trial court to review the population frequency statistics in greater detail, and we conclude that the acceptance of each probe is an essential portion of this process because the frequency statistics are probe specific.

We emphasize that, at this point, our review of legal publications and judicial opinions has not caused this court to question the use of these specific probes. On the other hand, little in the information we have reviewed suggests that judges have ever inquired about the acceptance of each probe. At the hearing on remand, the selection of probes was only briefly discussed. Unless Mr. Brim stipulates otherwise, the trial court must determine, during the relinquishment granted in this order, whether each of the five probes used in his test were generally accepted by the relevant scientific community at the time of the trial and, even if more sophisticated probes are in current use, that subsequent research has not discredited the accuracy of each probe.

7. Probing and Creating the Autorad.

Once the laboratory has selected probes to use in testing samples, the actual probing can begin. The DNA on the membrane is subjected to testing in a bath containing only one probe. The probe will attach to the sample DNA where the known complementary sequence is located. When scientists examine many human samples, they discover that this complementary DNA sequence occurs in fragmented strands of DNA of many different lengths due to the principle of polymorphism. As a result, a segment that attracts the probe in a specific sample can appear almost anywhere along the lane of DNA fragments on the nylon membrane. Because we inherit two copies of DNA, one from our mother and one from our father, each person will usually have two locations on the membrane that attract the probe. If the parents happen to have identical DNA in this respect, only one location attracts the probe.³⁶

36 . See NRC I at 55.

In addition to the short sequence of DNA, each molecule of the probe contains radioactive atoms. Because the probe is chemically attracted to only those fragments of DNA that contain the compatible sequence, the radioactive atoms tend to accumulate in two bands on the membrane in the two locations where the specific lengths of fragmented DNA migrated in the electrophoresis gel. After some additional processing, the membrane is x-rayed and the radioactive atoms produce visible bands on the x-ray film. This film is normally referred to as an autorad. The bands on the autorad are reliable, in the scientific sense that the experiment could be repeated and the bands would appear at almost exactly the same locations.

After the supreme court's decision in this case, it is clear that probing, as a general process, passes the Frye test. The trial court does not need to consider any additional evidence concerning this step on relinquishment.

8. Testing for Exclusion.

When each probe is completed, trained technicians examine the results on the autorad to determine whether the probe excludes the suspect as a possible perpetrator. As explained in greater detail in the next step, the bands created by a specific probe on an autorad can be examined to determine whether the bands in the crime scene sample occur at the same locations on the ladder, or at different locations on the ladder, than those in the suspect's sample. If the bands in the two samples occur at different locations on the autorad, then the probe selected a different fragment of DNA of a differing length in each sample. This difference effectively excludes the suspect as a possible perpetrator. If this occurs, no further probing or statistical analysis is necessary because the test has excluded the possibility that the suspect is the perpetrator.³⁷ The laboratory obviously did not find an exclusion in this case. There is no reason for the trial court to further examine this subject.

37 . To coin a phrase, if the bands do not fit, you must acquit.

VIII. Prong Two: DNA Statistical Analysis.

In this case, the bands in Mr. Brim's blood sample all appeared to align with the bands in the crime scene sample. As the supreme court explained, at the end of "this first step in the DNA testing process[,] [the results] simply indicate that two DNA samples look the same." Brim, 695 So.2d at 269. It is useful to consider that the five samples in a case such as this have compared only a combined total of approximately 150 to 200 nucleotides out of the more than 3 billion nucleotide pairs within the human DNA. Thus, it is not intuitive that the five tests establish any significant probability that the two samples of DNA come from the same person.³⁸

38 . By rough analogy, this is comparable to identifying a 4000-mile highway by examining only 5 pieces of road totaling less than 1½ feet in length.

Before any meaning could be attributed to points of similarity in two such samples, scientists had to accomplish two general steps. First, the scientists needed to create a valid and reliable "population frequency table" for each probe by testing a sufficiently large random population of humans to determine the range of human responses to the probe. Then they needed to decide the best statistical method or methods to explain what the population

1. Creating a Population Frequency Table.

The creation of a valid and reliable population frequency table is a critical step that must occur before a probe is of any use in any forensic case for identification. This general step contains at least two parts. First, the scientists must perform accurate RFLP testing with the specific probe on a random samples of humans to obtain sufficient data about the probe. Then the data must be properly summarized on a statistical table.

In the first part of the process, the scientists test hundreds of humans. In general, the larger the number of humans tested, the more accurate the population frequency tables. These tests create one autorad for each tested human with one or two bars along the ladder in the lane. The position of the bars on the ladder indicates roughly the length of the DNA segment for that tested person containing the complementary sequence of DNA that attracted the probe.

As explained earlier, short segments of DNA will have migrated a greater distance down the lane than longer segments. The scientists attempt to group similar segment lengths by dividing the lanes on these autorads into "bins." The creation of bins is rather like drawing lines across the autorad about every quarter inch. Each space between the lines is a bin. This process divides the length of the lane into as many as 25 bins.³⁹

39 . An appropriate number of bins is determined for each probe by examining its results. This process is sometimes referred to as sizing.

The scientists then examine all of the samples. They create a list to show how many samples have a band in each bin. Hypothetically, for example, after testing 100 humans, 5 tests of a given probe might produce a band in bin 15. The scientists could then predict to some level of certainty that a 5% chance or probability exists that any person randomly selected has DNA that would create a band in bin 15 when tested with the specific probe. The table that translates the data into such probabilities is the population frequency table.⁴⁰

40 . See, e.g., NRC II at 20 (example of population frequency table).

It is helpful to remember that the fact that two bands appear in the same bin does not mean that the segments of DNA contain identical sequences of nucleotides, except for the short sequence of the probe. It does not mean that the two segments are even identical in length. It merely establishes that the two samples each have the same short sequence of nucleotides in segments of DNA that are similar in length.

Obviously, there are many issues that a good scientist would consider and attempt to control in creating such a population table. The full range of those issues is well beyond the scope of this order and the competence of this court. It is clear that the sample size must be adequate, and the people tested must be selected randomly. In addition, the RFLP test must be accurately performed on each sample.

Two issues in the creation of the table are especially important. They revolve around the identity of the population tested and the binning process. Initially, scientists were concerned that the sequence of DNA used in a specific probe could appear in a segment length of DNA from a subgroup of humans more or less often than in the human population as a whole. By way of an extreme example, people from an isolated island might have DNA that is more similar in regard to a probe than humans as a whole. This would cause the bands produced by the subpopulation to fall into different bins more often than those reflected by the population frequency table for all humans. Scientists have, in fact, created different population frequency charts for Caucasians and African-Americans to adjust for this effect.⁴¹

41 . Determining race for these purposes is not challenged in this case.

Initially, some scientists were concerned that this risk of a subpopulation effect might cause national population frequency tables to conceal the importance of genetic tendencies in a local population. In other words, there was a concern that the frequency of a probe appearing in a particular segment length of DNA in Hillsborough County, Florida, might be significantly different from the results achieved in the United States population as a whole, or in a sample derived from another locality. If this were true, one could not use a general population frequency table to obtain accurate results in some localities.⁴²

42 . This debate was occurring at the time of Mr. Brim's trial. See R. Stephen Kramer, Comment, Admissibility of DNA Statistical Data: A Proliferation of Misconception, 30 Cal. W.L.Rev. 145 (Fall 1993).

From our review of legal publications and judicial opinions, it appears that the relevant scientific community now generally agrees that the concern over subpopulations is unfounded. The supreme court in its opinion accepted the use of the product formula and rejected the need to limit statistical testimony by any ceiling principle. The ceiling principles are designed, in part, to protect against the possible influence of subpopulation distortion. Accordingly, we conclude that the supreme court did not mandate the trial court to further examine the impact of subpopulations upon the population frequency tables.

Concerning the second matter, the binning process, the probabilities stated in a population frequency table are strongly influenced by the size and number of bins. If the lane in the autorad were divided into two bins, for example, the table would contain only two probabilities and they might each approach 50%. By dividing the lanes into many bins, the percentages become lower and the probabilities expressed by the forensic experts become more favorable to the State. An increase in the number of bins also results in more autorads that contain a band that straddles two bins. Thus, it is important that scientists who fully understand this process agree upon the number and placement of the bins used to quantify the results for any given probe. Our review of legal publications and judicial opinions leads us to believe that the relevant scientific community has generally accepted the binning process. We conclude, however, that because the bins are apparently set for each type of probe, it is necessary to have testimony that the binning processes used for the specific probes relied upon in this case, and likewise used to generate the population frequency tables relied upon in this case, were generally accepted at the time of trial and are still accepted at this time.

We would emphasize that it is unclear to this court whether population frequency tables have a scientific use outside the field of forensics. We are unaware of any current medical or academic use of such tables. We are concerned about the sufficiency of peer review if these tables are designed for use only in legal testimony. If only forensic experts have an interest in these tables, then the tables may not have been carefully examined and generally accepted or rejected by a neutral, objective group of scientists.⁴³ Our confidence in these tables would be enhanced if the evidence on relinquishment established that these tables were used for serious decision-making by people other than jurors.

43 . It is arguable that a court can use the Frye test to evaluate the admissibility of novel scientific evidence only when a sizable, objective scientific community exists and that community has already peer-reviewed the new principle or procedure. If a scientific procedure has limited applicability outside the context of litigation, it is possible that no objective peer group will ever adequately evaluate the novel scientific procedure. In such a case, a court may be compelled to evaluate the science, as science, to decide whether it is sufficiently reliable to be admitted into evidence.

At the hearing on remand in this case, the experts testified at some length about the procedures performed in Orlando to create a data base and population frequency table for Caucasian-Americans. In this case, however, that population frequency table is largely irrelevant because Mr. Brim is an African-American. The data base and population frequency table for African-Americans, which apparently was used in this case to perform the statistical calculations, was created by the FBI. Our record contains very little information about that table.

During the remand, the trial court did not receive any evidence about the general acceptance of the procedures used to create the relevant FBI tables for any of the probes involved in this case. Nothing in our record establishes general acceptance by the relevant scientific community of the FBI table as a scientifically valid and reliable table.⁴⁴ We conclude that when the supreme court required the trial court on remand to "clarify the exact methods used by the State in calculating its population frequency statistics at the time of the plea and trial,"⁴⁵ it intended the trial court to examine more than the final formula the State used in calculating statistics relating to Mr. Brim. The general acceptance of the FBI population frequency table, which was critical to the forensic testimony in this case, needed to be established.

44 . See Murray v. State, 692 So.2d 157, 162-63 (Fla. 1997).

45 . See Brim, 695 So.2d at 275.

During the relinquishment ordered by this order, the trial court shall confirm that the population table used in calculating the statistics used in Mr. Brim's test for each specific probe was generally accepted by the relevant scientific community at the time of the trial and, even if more sophisticated tables are in current use, that subsequent research has not discredited the reliability of the calculations at trial.

2. Calculating a Probability for Use in a Particular Case.

Once scientists have generated population tables for all the relevant probes, the next question is how to use these probability estimates to explain the test results in any given case. The supreme court's decision in this case established that forensic experts may rely upon the product or multiplication formula to explain the results. See Brim, 695 So.2d at 273.

The generic product formula is well-accepted within the field of statistics. It derives from Bayes' Theorem,⁴⁶ which is taught to every student in an introductory level course in statistics. Even without a course in statistics, anyone who has ever played Monopoly has sensed that the odds of one die landing on "1" are 1 in 6, and the odds of two dice landing in that manner are 1/6 x 1/6 or 1:36. When random events are independent, the odds that both will occur is the product of the odds that each will occur.

46 . See NRC II at 132; Ian W. Evett Bruce S. Weir, Interpreting DNA Evidence: Statistical Genetics for Forensic Scientists 17 (Sinauer Assocs., Inc. 1998).

The specific product formula used in interpreting DNA evidence leads to a mathematical statement that is difficult for the layperson to comprehend or objectively assess.⁴⁷ Whether the odds under the product formula are 1 in 3.5 billion or 1 in 2.7 billion in Mr. Brim's case, most jurors know that there are fewer than 15 million people in Florida. Such a high statistic suggests a positive identification to a layperson, but not necessarily to a scientist.⁴⁸

47 . It should not be overlooked that courts have traditionally prohibited the use of statistical evidence to prove guilt in criminal trials. See People v. Collins, 68 Cal.2d 319, 66 Cal.Rptr. 497, 438 P.2d 33 (1968) (noting, "[m]athematics, a veritable sorcerer in our computerized society, while assisting the trier of fact in the search for truth, must not cast a spell over him"); see also Laurence H. Tribe, Trial by Mathematics: Precision and Ritual in the Legal Process, 84 Harv. L.Rev. 1329, 1377 (Apr. 1971) (concluding that utility of mathematical methods is greatly exaggerated, that the methods inherently conflict with other important values, and thus "the costs of attempting to integrate mathematics into the fact-finding process of a legal trial outweigh the benefits"). Thus, the admissibility of DNA statistical evidence can be viewed as a departure from the general rule.

48 . Technically, the 1 in 3.5 billion statistic is the probability of a false positive. It is the probability of randomly selecting a sample of DNA that just happens to produce probe results that are the same as the results from the crime scene sample, See Interpreting DNA Evidence: Statistical Genetics for Forensic Scientists at 227-29. We do not know whether Mr. Brim happens to have an identical twin brother. If so, the forensic DNA Statistical against him would be the same as the evidence against Mr. Brim. It is noteworthy that the FBI's internal guidelines apparently do not permit a forensic expert to claim that the DNA test has established a positive identification unless the statistical probability reaches 1 in 260 billion-a number that is many times the total population of Earth. See Random Samples Edited by Constance Holden, "DNA Fingerprinting Comes of Age," Science, Nov. 12, 1997, vol. 278, issue 5342 at 1407.

Because the product formula reaches its answer by multiplying the probabilities derived from five different probes, it multiplies the possible error within the statistics for each probe.⁴⁹ It also results in unreliable outcomes if the sample used to establish the probabilities in the table was not randomly selected or if the probes are not truly independent of one another. By analogy, the odds on the Monopoly dice change if they are shaved or weighted. The odds of a little girl having black hair and brown eyes cannot be calculated by the product of these two characteristics because they are not independently inherited.

49 . See Ryan McDonald, Note and Comment, Juries and Crime Labs: Correcting the Weak Links in the DNA Chain, 24 Am. J.L. Med. 345, 355 (1998); Jonathan J. Koehler, On Conveying the Probative Value of DNA Evidence: Frequencies, Likelihood Ratios, and Error Rates, 67 U. Colo. L.Rev. 859 (Fall 1996). See also Schwartz, 447 N.W.2d 422 (highlighting errors in laboratory testing and limiting use of statistical evidence of DNA).

The counting method, the ceiling principles, and the modified ceiling principles described in the supreme court's decision are methods that adjust the statistical analysis in DNA cases to reflect scientific concern that the pure product formula overstates the true level of scientific certainty in this new and fast-developing field. Although concern about the precision of the product formula may still, be a proper ground for cross-examination, the supreme court's ruling in this case put aside any doubt that the product rule was generally accepted under the Frye standard. Thus, on relinquishment, the trial court does not need to further address the propriety of the use of the product formula to explain DNA evidence to a jury.

However, the product formula necessitates that the forensic expert calculate the statistical evidence by inserting specific probabilities from the FBI's population frequency tables into the formula. As we understand the formula used for DNA calculations from examining it in textbooks outside the trial court's record, it will vary depending upon whether each probe creates one band or two on the autorad.⁵⁰

50 . See NRC II at 35.

We conclude that the supreme court intended the trial court to determine exactly what formula was used in this case, and what probabilities were inserted into the formula to reach the result of 1 in 3.5 billion. The trial court needed to determine that the generally accepted product formula was utilized and not some other formula.⁵¹ It should have confirmed that the correct probabilities from the tables were placed into that formula when calculating the statistic presented to the jury.

51 . See generally Interpreting DNA Evidence: Statistical Genetics for Forensic Scientists ch. 5 at 206-16.

There are two additional concerns with the statistical calculations and testimony in Mr. Brim's case. First, if we understand the testimony correctly, the bands on Mr. Brim's autorad did not fall squarely within single bins in all instances. Because the bands overlapped bins, the formula was adjusted to account for the overlap. The method used in 1993 was different from the method NRC II recommended in 1996. This is the change that caused the statistical probability to shift from 1 in 3.5 billion to 1 in 2.7 billion. We are uncertain whether any refinements in binning have altered these calculations during the last three years.

Moreover, the forensic expert for the State was not required to testify that the estimate was subject to any range of error. Because probabilities are essentially predictions, and are based on limited sampling, they always have a degree of uncertainty or a range of error. The level of uncertainty itself can often be statistically estimated. NRC II suggested that a factor of 10 be used to describe a scientist's level of confidence in the statistical estimate.⁵² In this case, that would require a forensic expert to testify that the probability was as low as 1 in 270 million or as high as 1 in 27 billion. We do not know whether forensic testimony that omits a disclosure on the level of uncertainty is generally accepted within the relevant scientific community, and we have not decided whether such a disclosure should be required by the court, as a matter of law, even if experts are willing to forgo the disclosure.⁵³

52 . See NRC II at 34.

53 . See Jonathan J. Koehler, Why DNA Likelihood Ratios Should Account for Error (Even When a National Research Council Report Says They Should Not), 37 Jurimetrics J. 425 (Summer 1997).

IX. Instructions to Trial Court on Relinquishment

As explained earlier, we are relinquishing jurisdiction due in large part to the extraordinary burden that the supreme court has placed upon the district court when reviewing a Frye determination. We have an obligation to review de novo the trial court's determination that both the underlying scientific principles and all of the testing procedures used in this case to generate DNA statistical evidence were and are generally accepted by the relevant scientific community. In addition to reviewing the trial court determination — which was based upon a preponderance of the evidence — we must make our own appellate "determination," evaluating current science and law. See Brim, 695 So.2d at 274. The supreme court has instructed us to consider, in our de novo review, scientific literature that is outside the record created in the trial court. Id. To accomplish this task expeditiously and efficiently, while providing notice to the parties and a record subject to review in the supreme court, both the trial court and this court must take special precautions.

First, at the beginning of the period of relinquishment, the trial court shall conduct a case management conference to establish the precise issues to be addressed at the evidentiary hearing. Mr. Brim's December 1992 motion to exclude DNA evidence was a brief two-page motion attacking "DNA profiling evidence." He filed a lengthy memorandum in support of the motion that questioned virtually all aspects of DNA testing and testimony. We do. not question the sufficiency of the 1992 motion, but at this point, the issues have been considerably narrowed by this order and the opinions that preceded it. Even as to the issues left open by this order, it may be that Mr. Brim can no longer allege in good faith that a particular principle or procedure is not generally accepted. Although it is the State's burden to prove general acceptance within the relevant scientific community, we believe the defendant has an obligation to plead reasonably narrow reasons why he maintains there is a lack of consensus once he is informed about the testing methods used. A defendant has an obligation to preserve an error for review. The defendant's argument or objection needs to be "sufficiently precise" that the trial court knows the grounds for the objection. See § 924.051(1)(b), Fla. Stat. (1999). Frye hearings involve expensive expert time and testimony. There is no reason to require the State to prove the acceptance of a long list of scientific principles and procedures unless the defendant has raised and preserved an issue concerning those matters.⁵⁴

54 . In Castro, 144 Misc.2d 956, 545 N.Y.S.2d 985, the New York Superior Court sought to establish a pretrial procedure for the presentation of DNA evidence, complete with discovery requirements and burdens of proof, which would allow the trial court to narrow the issues presented. See also Polk v. State, 612 So.2d 381 (Miss. 1992).

Second, the trial court shall determine, based either upon stipulation of the parties or upon evidence, the relevant scientific community in which these principles and procedures must reach general acceptance. From our review, we are inclined to believe that the community may not be so expansive as to include all doctors and scientists who study DNA nor so narrow as to include only forensic genetic experts. It appears most likely from this court's extra-record readings that the relevant scientific community, at least for the statistical issues, is composed of population geneticists,⁵⁵ but we lack the scientific expertise to make this determination as a matter of law on appeal. The trial court may need to make this determination prior to the final evidentiary hearing to allow the parties time to retain the necessary experts.

55 . See "Reference Guide on Forensic DNA Evidence" at 286.

Third, and in summary of the instructions contained in this order, the trial court should consider the following issues at the evidentiary hearing unless Mr. Brim no longer contests the specific principles or procedures:

1. Determine whether the specific ladder used in the DNA test performed in Mr. Brim's case was generally accepted by the relevant scientific community at the time of the trial. If more sophisticated ladders are currently in use, determine that subsequent research has not discredited the accuracy of this type of ladder

2. Determine whether each probe used in Mr. Brim's test, and in the creation of the FBI's population frequency tables applied to Mr. Brim, were generally accepted by the relevant scientific community at the time of the trial and, even if more sophisticated probes are in current use, that subsequent research has not discredited the accuracy of each probe.

3. Determine whether the binning methods used in Mr. Brim's test, and in the creation of the FBI's population frequency tables applied to Mr. Brim, were generally accepted in the relevant scientific community at the time of trial and, even if more sophisticated methods are in current use, that subsequent research has not discredited the reliability of these methods. This inquiry should include questioning as to whether the method NRC II recommended in 1996, after Mr. Brim's trial, discredited the method used in Mr. Brim's case.

4. Confirm that the FBI population frequency table used in calculating the statistics used in Mr. Brim's case for each specific probe was generally accepted by the relevant scientific community at the time of the trial and, even if more sophisticated tables are in current use, that subsequent research has not discredited the reliability of these tables. Consider whether these tables have a scientific use outside the field of forensics and, if not, whether that should influence their legal reliability.

5. Determine exactly what formula was used in this case to calculate the statistical probability presented at trial and whether it was the generally accepted product formula. Then, confirm that the correct probabilities from the tables were placed into that formula when calculating the statistics presented to the jury. Determine whether that formula is still generally accepted, and even if an altered formula is now in current use, that subsequent research has not discredited the formula used in this trial.

6. Determine whether the failure to disclose the level of uncertainty in these calculations is something that is generally accepted within the relevant scientific community. Examine whether a court should require such a disclosure in the presentation of the evidence, as a matter of law, even if experts are willing to forego the disclosure, or whether this can be adequately addressed during cross-examination.

Finally, we add that the trial court, independent of the above, should examine the sentences imposed on Mr. Brim in these two cases to ensure that the corrections required in Brim v. State, 654 So.2d 184 (Fla. 2d DCA 1995), and Brim v. State, 754 So.2d 823 (Fla. 2d DCA 2000), have been made.

We conclude that the trial court is authorized to accept testimony from qualified expert witnesses who express an opinion that a specific scientific principle or testing procedure is or is not generally accepted within the identified relevant scientific community. The expert may base his or her opinion on personal knowledge and on a review of the literature upon which such experts regularly rely. See § 90.703-704, Fla. Stat. (1999). We do not believe it is incumbent upon the trial judge to read and fully comprehend all of the technical literature upon which the expert relies. As we have explained earlier, a court reviews this literature in a Frye hearing to assess the level of consensus or disagreement within the scientific community, not to evaluate the scientific validity or scientific reliability of the principles and procedures.

On the other hand, we caution that the trial court cannot base its decision purely on the credibility of an expert witness who does not explain his or her reasoning in considerable detail. Due to the unusual de novo standard of review, this court is not authorized to defer to the trial court on these issues.

If the parties expect the trial court to read and review scientific literature, the trial court may require that literature be introduced into evidence through an expert so that the trial court can understand the significance of the literature and the reason that the trial court must specially review it. The trial court has discretion to decline to receive literature if it is written at a technical level that is beyond the trial court's competence to evaluate and instead require that the experts provide understandable explanations of the literature.⁵⁶

56 . For example, this court has independently reviewed Interpreting DNA Evidence: Statistical Genetics for Forensic Scientists. Without expert clarification, portions of this text are well beyond our competence.

During the evidentiary hearing, the trial court should be especially cognizant of this court's need to review the determination. The formulas, population frequency tables, information about the probes, and the actual calculations involved in this case should be reduced to exhibits that this court can read and understand. Special care should be taken so that the transcription of testimony accurately reflects the testimony of witnesses who use scientific jargon. All literature introduced in evidence before the trial court must be included in the record on appeal when this case returns to this court.

If a party wishes this court to review any additional scientific literature during the pendency of the appeal after the relinquishment, it must file a copy of the literature with this court as a supplemental scientific authority. These filings may briefly explain how the new literature demonstrates either general acceptance or disagreement within the relevant scientific community. This court has created a formal record of the major materials reviewed to date upon which it relies to issue this order. This material will be available to the trial court and the parties during the relinquishment and will become part of the record in Mr. Brim's case.

In this order, we have repeatedly instructed the trial court to decide whether the calculations used at trial are still legally reliable even if subsequent research has developed more sophisticated methods. We confess considerable uncertainty about the degree of change in the statistical estimates that would warrant the trial judge determining that the old statistics were improperly admitted and that a new trial should be required. It is our impression that this is ultimately a legal question and not a question for the experts from the relevant scientific community.

This is not a situation in which an established trial error is evaluated under the harmless-beyond-a-reasonable-doubt standard. See State v. DiGuilio, 491 So.2d 1129 (Fla. 1986). Instead, the trial judge is deciding whether a trial error has even occurred. While declining to announce today what level of change would compel us to reverse the verdict at the conclusion of this appeal, we agree with the trial court that, if the changes in technology favor Mr. Brim with only a shift from 1 in 3.5 billion to 1 in 2.7 billion, then the new developments do not appear to require a new trial.⁵⁷

57 . It should not be overlooked that if a new trial could be lawfully ordered in this case, newer DNA testing with additional probes might either exclude Mr. Brim as the perpetrator or increase the statistics against him. During the pendency of this appeal, this court attempted to determine whether updated testing could be performed on the existing samples, but Mr. Brim convinced this court that it should not order additional tests. It is his position that if the testing had properly been excluded in the first trial, he would have been acquitted and, thus, protected from a new trial by double jeopardy.

This order is not the final opinion of this court in this case, and no mandate will issue from this order. This court is simply relinquishing jurisdiction to the trial court through December 29, 2000, to complete the hearing on remand that the supreme court ordered in 1997⁵⁸ Once the trial court has completed its work on relinquishment, this court will review the new order in an expedited manner. This court would note that the appellate lawyers for both Mr. Brim and the State have done an exceptional job presenting this difficult case. It would facilitate our final review of this matter if they were allowed to assist trial counsel during the relinquishment.

58 . The trial court should make every reasonable effort to complete this assignment during the allotted time. If the trial court cannot complete this task during the allotted time, the State may seek an extension of the relinquishment for the minimum amount of time needed to complete the hearing and revised order.

Jurisdiction relinquished through December 29, 2000, for an additional evidentiary hearing.

FULMER and SALCINES, JJ., Concur.