concurring.

The central question in this case is whether, and to what extent, an appellate court may consider matters not in the record when reviewing a trial court’s decision to admit or exclude scientific evidence. The answer is that the appellate court should consider only material that is in the record and matters that may be judicially noticed. The reliability of a scientific theory or technique should be judicially noticed under the following circumstances: (1) when it is a matter of common knowledge, (2) when widely available court decisions show that reliability has been litigated elsewhere in fact-finding forums to a degree sufficient for the appellate court to conclude that reliability is well-established, and (3) when a prior determination of reliability has been made by an appellate court whose pronouncements are binding in the jurisdiction. As will be discussed below, appellate courts should never conduct their own independent research of the scientific literature.

I. SCIENTIFIC LITERATURE NOT PRESENTED TO THE TRIAL COURT

An appellate court that consults scientific literature on its own initiative thrusts itself into the position of a fact finder — a position appellate courts traditionally do not occupy and for which they are ill-suited. No matter how careful the appellate investigation, there is always the risk that appellate research will fail to uncover scientific sources that are crucial to determining the reliability of a scientific theory or technique.1 Moreover, appellate courts cannot hear live testimony, and such testimony may be important to litigating a particular scientific claim. Experts practicing in the field may have knowledge and experience beyond what is reflected in written treatises.2 As a result, the extent to which a certain scientific treatise or specific assertions within that treatise may be relied upon may be a legitimate subject of litigation that only a fact-finding court can fully address. Consequently, the ability to rebut scientific literature with live expert testimony is an important part of the process of determining the reliability of a scientific theory or technique.3 Cross-*33examination can also be utilized to test the credibility of the sources upon which the parties rely. In line with this reasoning, several appellate courts in other jurisdictions have held that a trial court must conduct a reliability hearing before an appellate court can decide for the first time that a scientific theory’s or technique’s reliability can be judicially noticed.4

The difficulties that result from appellate courts relying upon material outside the record manifest themselves most clearly in the current trend of citing to internet sources in appellate opinions. While some problems arising from this practice, such as evolving content and vanished content, do not carry over completely to the use of outside-the-record texts and treatises, many of the core arguments against the use of internet references apply with full force to the latter situation.5

[W]hen an appellate court goes outside the record to determine case facts ... it ignores its function as a court of review, and it substitutes its own questionable research results for evidence that should have been tested in the trial court for credibility, reliability, accuracy, and trustworthiness.6

Resort could be made to opinion polls, to information on the internet, to articles in psuedo-scientific magazines, or even to authentic treatises that are just wrong. While I trust appellate courts to do their best to sort out the rehable from the unreliable, I trust cross-examination more. Rules of evidence and procedure protect the reliability of the outcome of trial. It is a mistake to hope that appellate diligence can adequately substitute for these tried and true safeguards.

In the past, we have held that judicial notice of scientific literature, as opposed to scientific theory or technique, can be taken even when that literature was not presented by either party at trial or on appeal.7 As I have discussed above, such a holding carries the concept of judicial notice too far. Insofar as they purport to give an appellate court the ability to conduct an independent investigation of outside-the-record scientific literature, those cases are rightfully overruled by the Court’s opinion today.

Judge Keasler contends that scientific reliability should be reviewed de novo and that a de novo standard of review means, or necessarily includes, review of outside-the-record material. Without commenting on whether a de novo review is appropriate, I must express my disagreement. Equating de novo review with an outside-*34the-reeord examination confuses the “how” of appellate review with the “what” of appellate review. The “how” of appellate review involves the amount of deference an appellate court owes a trial court. Whether outside-the-record material may be considered is a separate question, involving the “what” of review. In Jones v. United States, the District of Columbia Court of Appeals recognized that a de novo review, in this context, does not necessarily mean an outside-the-record review:

De novo review in this context, however, is not self-explanatory. Does it simply mean review limited to the trial record, but without required deference to the trial court’s findings and analysis? Or does it also permit appellate court reference to sources outside the record? If so, what kinds of sources: legal and scientific articles, as well as judicial decisions from other jurisdictions? If so, is one type of outside source entitled to greater weight than another? In any event, may the appellate court rely primarily or even exclusively on sources outside the record, or is the probative value of such sources limited to buttressing essential expert testimony of record?8

While holding de novo to be the proper standard of review,9 the D.C. Court declined to rely upon outside-the-record scientific literature; instead the court relied upon other cases in which the validity of the scientific test was litigated at a fact-finding hearing,10 a matter which I address later in this opinion.

II. JUDICIAL NOTICE AS A SUBSTITUTE FOR EVIDENCE

A. Trial versus appeal

We have held that the first two Kelly11 criteria — the reliability of the underlying scientific theory and the reliability of the technique applying the theory — can be determined through judicial notice if the validity of the theory or technique is well-established.12 When a scientific theory or technique is amenable to judicial notice, the proponent is relieved of the burden of producing evidence on that question.13 For this reason, I must disagree with the Court’s contention that “judicial notice on appeal cannot serve as the sole source of support for a bare trial court record concerning scientific reliability.”14 The purpose of judicial notice is to relieve the party of the burden of having to fortify the record. And although the trial court did not proffer an adequate basis for taking judicial notice, if there is such an adequate basis, an appellate court should uphold the trial court’s decision, just as an appellate court would in any case in which a trial court’s decision was correct on an applicable theory of law that was not in fact litigated before the trial court.15 I turn, then, to examining the permissible bases for taking judicial notice.

B. Bases for judicial notice

1. Matters of common knowledge

Some scientific theories and techniques are so well known that they are matters of *35common knowledge.16 Examples of this type of matter are the laws of thermodynamics 17 and the fact that the process of photography is capable of producing a correct likeness of an object.18 Other examples would be the uniqueness of fingerprints and DNA. In such cases, judicial notice may be taken without a fact finding hearing and without consulting any scientific literature.19

2. Consideration in other courts

Appellate courts often look to cases in other courts and other jurisdictions to help determine whether judicial notice should be taken of the validity or reliability of a scientific theory or technique.20 There is good reason for doing so because, as the Supreme Court of Vermont observed, scientific or technical evidence that is novel in one jurisdiction is often not novel in others.21 Even if the trial court in the case under consideration did not conduct an extensive evidentiary hearing, that trial court (or an appellate court reviewing the case) can use the decisions of other appellate courts that had the benefit of an extensive reliability hearing held at trial:

In many cases, like this one ... the issue is whether a certain category of evidence is admissible, often in particular types of cases that are recurring. In some cases both the trial court and this Court can fully evaluate the reliability and relevance of the evidence generally based on the decisions of other appellate courts. In this way, we can avoid conducting our own lengthy and expensive evidentiary hearing aimed at establishing, or attacking, the foundation for the disputed expert testimony. We are not suggesting that the new standard for admissibility has somehow become general acceptance among appellate courts. Irrespective of the decisions of other courts, the responsibility for determining the admissibility of evidence in Vermont courts remains with our trial judges, and on appeal with this Court. However, scientific or technical evidence which is novel to us is frequently not novel to many other state and federal courts. To the extent the evaluation of these courts is complete and persuasive, we can affirmatively rely upon it in reaching our own decision.22

For example, in United States v. Jakobetz, the Second Circuit, based upon an extensive reliability hearing at trial, took judicial notice of the general reliability of *36RFLP-type DNA testing.23 The Eighth Circuit, confronted with the same issue as one of first impression, took judicial notice of the reliability of this type of evidence based upon Jakobetz’s earlier, well-supported holding, saying:

Although Jakobetz was written before Daubert, the court employed a reliability approach to Rule 702 similar to that taken in Daubert. We conclude that the Second Circuit’s conclusions as to the reliability of the general theory and techniques of DNA profiling are valid under the Supreme Court’s holding in Daubert, and hold that future courts can take judicial notice of their reliability.24

In Jones, the District of Columbia Court of Appeals observed that numerous courts have taken this approach:

Perhaps because reliance on articles to establish general acceptance is problematic, more appellate courts, confronted by trial court rulings not based on expert testimony, have been willing instead to sustain the rulings primarily on the basis of judicial notice of other court opinions which themselves were based in substantial part on expert testimony. A number of appellate courts have altogether dispensed with the need for expert testimony at trial if (1) other courts have established, from expert testimony and other sources, the general scientific acceptance of the particular test at issue, and (2) the records and judicial elaborations of these other cases are sufficiently complete and persuasive that the appellate court confidently can incorporate the work of other courts by reference... .Expert testimony in other cases, subject to cross-examination, can be probative of general acceptance of a scientific technique; thus, judicial notice of it is appropriate... .The reviewing court, of course, has to take care that judicial notice of other court opinions is limited to comprehensive expert testimony on the general acceptance issue; otherwise, there is a danger that reliance on other judicial opinions could, in effect amount to delegation of decision-making to another court which itself ruled on the basis of an inadequate record.25

The Jones Court followed this method in the case before it:

We therefore rely primarily on a trial court decision from our own jurisdiction, based upon expert scientific evidence in a record with which we are familiar and in which we have confidence because of the thoroughness with which counsel tried the case and the judge evaluated the evidence. We rely secondarily on, and thus confirm our judgment by reference to, judicial opinions from other jurisdictions which have reached the same result... .(Parenthetically, we emphasize that we do not rely, additionally, on an independent review of scientific or legal literature discussing [the scientific test in question]. We prefer to rely upon judicial decisions which themselves have a trial record — or judicially notice a trial record — that reflects expert testimony, subject to cross-examination about [the scientific test] ).26

*37A less exacting inquiry should be required if a large number of jurisdictions recognize the validity or reliability of a scientific theory or technique. A court could take judicial notice based upon the great weight of authority, which is what occurred when the acceptance of DNA testing among appellate courts had become widespread.27

3. Prior Determination of Reliability

When a scientific theory or technique’s validity or reliability has been shown so convincingly before a fact-finder that an appellate court can conclude the matter should be judicially noticed, the appellate court can decree that the matter will henceforth be the subject of judicial notice. In such instances, judicial notice of the reliability of the scientific theory or technique could then be taken by trial and appellate courts within the decreeing court’s jurisdiction in future cases in which the issue arose.28

III. THE SCIENCE AT ISSUE: THE ADX ANALYZER/FPIA TEST

A. Remand inadvisable

The scientific literature and cases cited by the State in its petition and briefs before us were never presented to the Court of Appeals. For this reason, I must disagree with the dissent’s suggestion that the case be remanded to address these sources — or even with any suggestion that the case be remanded for the more limited judicial notice process I have outlined. No one suggests that the Court of Appeals was required to investigate these sources on its own — only that the lower court could have done so. I see little point in remanding the case to that court to consider, at its discretion, authorities not presented on initial submission. Since the case is currently before us, I will address the question of reliability under the standards I have proposed.

B. The merits

Under those standards, an independent review of the scientific literature is not permitted. So I look to the trial record and to permissible methods of taking judicial notice. It is clear from the trial record that the urinalysis technician did not know how the machine works and was not in a position to give an expert opinion on the reliability of the scientific technique upon which the machine operated. The trial court admitted the evidence because of past litigation in which the same urinalysis operator testified. But we do not have knowledge of these past instances of litigation, and they were not included in this record. And because the State did not present other expert testimony, the present record does not contain sufficient evidence of the reliability of the machine and test.29

The next question is whether reliability can be judicially noticed. The reliability of *38the ADx machine or the FPIA30 test conducted by that machine31 is not a matter of common knowledge. Nor is there a prior case from this Court holding the machine or test to be reliable. The remaining question is whether other courts have addressed this issue in a manner that permits us to take judicial notice of reliability.

In its brief, the State cites three cases: Koenig v. Vannelli,32 Penrod v. State,33 and Carter v. State.34 Koenig involved an ADx test conducted on a prison inmate, who tested positive for marijuana.35 In a footnote, the Ninth Circuit stated “The ADx test is a fluorescein polarization immunoassay test (FPIA) that is used to detect marijuana and other drugs in biological fluids. When properly performed, FPIA tests for cannabinoids are generally accurate.”36 As authority for this proposition, the Ninth Circuit cited a medical article.37 The court did not say how it learned of the medical article and made no reference to a fact-finding hearing at which the reliability of the ADx machine or FPIA testing was litigated. Without any assurance that the Ninth Circuit’s holding is based upon a fact-finding hearing, we should not factor its holding of reliability into a judicial notice analysis. The case does establish that ADx testing occurred as early as 1992.

Penrod is an Indiana intermediate court opinion that addressed whether the trial court had erred in admitting ADx test results.38 In that case, the defendant conceded, apparently at trial,39 that: “there is no difference between” the ADx machine “and the EMIT Cobus or the Seva Corporation ... Those are basically the same technologies. Although the names may be a little different they are basically the same.”40 On appeal, the defendant contended that the State had failed to prove the scientific reliability of the ADx machine.41 Citing the Jones case from the District of Columbia Court of Appeals,42 the Penrod court recognized that the EMIT test had gained general acceptance in the scientific community.43 Referring to the defendant’s concession that the ADx and EMIT tests were basically the same, the Indiana appellate court held that, “[biased upon this statement,” the trial court did not err in admitting the evidence.44 The court articulated no other basis for holding the ADx test to be reliable.

It is unclear whether the Penrod court intended a broad holding that the ADx test was reliable for all future cases or a narrow holding that the trial court did not *39abuse its discretion in admitting the evidence in the case at hand. If the former, the basis for that holding — the defendant’s concession — is tenuous. While the defendant’s concession was adverse to his position, such a concession can hardly be considered an adequate basis for formulating a general rule applicable to all cases. However, the wording of the opinion lends at least some support to the latter interpretation, and in Carter, the same court of appeals held that Penrod’s holding applied only to that case:

In Penrod, we noted that the Record contained a statement that “there is no difference between [the] Abbott Laboratory machine ... or the Seva Corporation.” Penrod, 611 N.E.2d at 654. This reference is insufficient to allow judicial notice concerning the machine at issue in this case. The Penrod decision indicates that this statement was made by the probationer or his counsel; thus, while it was a sufficient basis for our decision in Penrod, it does not form a sufficient basis for acceptance of such evidence in other cases.45

Carter involved, among other things, a question concerning the reliability of a urinalysis test named in the record as “CIVA,” which may have been an inaccurately transcribed spelling of “Seva,” which would denote the test devised by Seva Corporation.46 Aside from holding that Penrod had no application beyond its own facts, the Indiana Court of Appeals also held that Penrod involved only the ADx system and did not apply to the “CIVA” test at issue.47 One judge dissented, citing Penrod’s holding based upon the defendant’s concession, which mentioned ADx, EMIT, and Seva.48

The Indiana Supreme Court reversed.49 Concerning the urinalysis test, the court stated, “we note Judge Friedlander’s dissent where he observed, ‘it is beyond debate that urinalysis has achieved a sufficient level of scientific reliability to be accepted in evidence in our courts’ [citation omitted]. Urinalysis technology is hardly novel and has become a conventional means of drug-testing, the results of which have been deemed reliable in Indiana courts.”50 In support of its statements, the Indiana Supreme Court cited Judge Friedlander’s dissent and Penrod and, relying upon Penrod’s quotation of the defendant’s concession in that case, noted that “Penrod also references ‘Seva’ urinalysis machines at issue in the case before us.”51 The inescapable conclusion is that the Indiana Supreme Court’s holding in Carter flowed solely from the defendant’s concession in Penrod. And in the case of Carter, the concession was used to support the Seva machine’s reliability, a proposition that was not even at issue in Penrod. The State’s cases do not provide sufficient basis to hold that the ADx machine and the FPIA test are reliable.

My research has, however, uncovered a few other cases. In People v. Toran, an Illinois Court of Appeals addressed a claim that a probationer’s confession of cocaine use was involuntary.52 The court held that *40admission of the confession, if error, was harmless because of positive drug test results.53 The defendant’s urine was tested (in 1989) by two different methods: (1) the ADx analyzer and (2) the Toxi-Lab thin layer chromatography test; both tests indicated the presence of cocaine.54 According to the Illinois court, “Testimony was presented that both testing methods were generally accepted as reliable within the chemistry field.”55 So, as early as 1989, the ADx analyzer was in use, and there was testimony at that time that tests conducted by the machine were generally accepted as reliable, at least as to cocaine. It is questionable, however, whether such testimony establishes the test’s reliability for determining the presence of marijuana.

More evidence of reliability surfaced in 1993 in McCoy v. Lockhart, an unpublished opinion by the Eighth Circuit.56 In that case, an inmate filed against the Arkansas Department of Corrections (ADC) a 42 U.S.C. § 1988 action claiming that his due process rights had been violated by a random “unreliable” urinalysis test that returned a positive result for marijuana.57 After a trial, ADC submitted an affidavit from its Administrator of Medical and Dental Services explaining that ADC utilized the “ADx system” and a second type of test to confirm the results.58 The affidavit further attested that “the ADx sys-tern was comparable to the ‘EMIT system’ in accuracy and sensitivity.”59 This post-trial affidavit was never contested.60 In written findings of fact, the magistrate judge held that ADC’s drug testing method was reliable.61

Although not addressing the ADx analyzer, in Love v. State the Georgia Supreme Court discussed what appears to be the underlying test — described by that court as the “Fluorescence Polarization Eminase (FPIA)” test. In that case, expert witnesses discussed the FPIA method as one of two methods used by the Georgia Bureau of Investigation to determine the presence of marijuana in urine.62 In a California civil lawsuit, witnesses testified to three different methods of analyzing urine for the presence of drugs: (1) enzyme multiplied immunoassay technique (EMIT), (2) fluorescence polarization immunoassay (FPIA), and (3) gas chromatography/mass spectrometry (GC/MS).63 The reliability of these various tests was not discussed, except to observe that the practice of the laboratory was to test under one of the first two, and if a positive result occurred, to retest under the third method.64

In 2002, a California court of appeals addressed the admissibility of ADx test results in People v. Nolan.65 The defendant was tested by the ADx machine. He tested positive for marijuana and his pro*41bation was revoked. Id. at 333. The California court began by citing cases holding that urinalysis is a generally accepted method of drug testing.66 The court further stated that “The ADX Abbott test has reached a general ‘level of acceptance’ in the scientific community ... [and] is an accurate test for the presence of marijuana in biological fluids.”67 For support, the court cited Carter, Penrod, and Koenig.68 Citing Taran, the California court also remarked that the ADx test had been used before to test for cocaine,69 and the court observed that the ADx machine had been used to perform thousands of tests in the local drug court program.70 The court compared the ADx test to the EMIT test: “Both the ADX Abbott and the older EMIT (enzyme immunoassay test) use well-accepted immunoassay scientific technique to detect drugs in urine.”71 In support of this reliability comparison, the court cited Koenig, Penrod, Carter, and Spence v. Farrier,72, an Eighth Circuit case addressing the reliability of EMIT testing.73 Finally, the California court indicated that its focus was on the methodology, not the machine carrying out the methodology: “There will always be new devices to implement established scientific methods. But a Kelly74/Frye75 hearing is not required for new devices; it applies to new methodologies.”76

The California court’s observation that urinalysis, in the abstract, is generally accepted is not particularly helpful because there are several different methodologies for conducting urinalysis testing. General acceptance of “urinalysis testing” could be based upon tests not at issue in the case (the gas chromatography/mass spectrometry test, for example). Also unhelpful is the court’s distinction between machine and methodology, and its insistence that the machine itself need not be shown reliable. Not only does this position contradict our Kelly case, which requires that the “technique applying the theory” be shown reliable,77 but analysis of the cases indicates that various urinalysis tests employ different methodologies, even if they may obtain similar results. While the Nolan court points out that FPIA and EMIT are both immunoassay tests, that observation is likewise of limited value because the caselaw suggests that they are different tests, even if they share some common characteristics. And to the extent the California court relied upon the Koenig, Pen-rod, and Carter cases, it relied upon a foundation of sand. On the other hand, the Nolan court did add the fact that the ADx machine had been used in drug court to perform thousands of tests, and it referred to Toran, in which there was live testimony about the reliability of the ADx system.

An issue that has arisen in some of these cases is the comparison between the ADx/ FPIA system and the EMIT system. The significance of that comparison is that the *42EMIT system has been overwhelmingly accepted as reliable.78 The reliability of that test has been litigated extensively before fact-finders, with the parties being able to offer live testimony and to conduct cross-examination.79 A comparison of the ADx test to the EMIT test might allow the ADx test to piggyback onto the well-recognized reliability of EMIT.

Although the references are sparse, there is at least some indication, in litigation at the trial level, that the ADx test is comparable to EMIT in its reliability. There has been at least one hearing in which live witnesses testified that the ADx test was a reliable method of detecting cocaine usage. The ADx system has been in use for fourteen years or more and has been used in thousands of cases and across multiple jurisdictions. And I have not found a single case in which the reliability of the ADx analyzer, or the FPIA method it uses, was questioned by an expert witness or a fact-finder. Although this is some evidence of reliability, and the question is perhaps a close one, the evidence is not sufficiently extensive and convincing for a court to confidently take judicial notice of ADx analyzer’s reliability. Consequently, I agree that the Court of Appeals’s decision should be affirmed.

OPINION

The present case involves the admissibility of scientific evidence under Texas Rule of Evidence 702 and Kelly v. State.1 Specifically, we granted the State’s following three grounds for review:

1) Must a party seeking to introduce evidence of a scientific principle always present evidence sufficient to satisfy the test of Kelly v. State, 824 S.W.2d 568 (Tex.Cr.App. 1992), regardless of the particular scientific principle?

2) Where either the Court of Criminal Appeals or a court of appeals has determined the validity of a particular scientific principle and a technique applying that principle, must a party subsequent^ ly seeking to introduce evidence based upon that scientific principle nevertheless satisfy the first two prongs of the test of Kelly v. State, 824 S.W.2d 568 (Tex.Cr.App. 1992)?

3) Did the Court of Appeals in this case err in holding that the trial court had abused its discretion by admitting evidence of the results of a urinalysis test of the appellant’s urine sample?

The short answers to these three questions are: “No, no, and no.” Therefore, we affirm the court of appeals which held that the trial court abused its discretion in re-*28yoking appellant’s community supervision based upon scientific evidence that was not shown to be reliable. Hernandez v. State, 55 S.W.3d 701 (Tex.App.-Corpus Christi 2001).

I.

Appellant pleaded guilty to possession of marijuana and the trial court placed him on probation for ten years. Almost a full ten years later, the State filed a motion to revoke his probation alleging, inter alia, that appellant failed to avoid the use of controlled substances and had tested positive for marijuana on January 28, 1999.

At the revocation hearing, Alonzo Perez, a laboratory technician, testified that he tested appellant’s urine for the presence of drugs using a machine called an “ADx analyzer.” The test results were positive. Mr. Perez testified that he had worked as a urinalysis lab technician for two and a half years. He explained that he had thirty-two hours of specialized training on the ADx analyzer and about two and a half weeks of extensive on-the-job training. When asked how many urinalysis tests he had performed, he replied, “I couldn’t say. It’s just so many.”

On cross-examination, appellant asked Mr. Perez to explain the scientific theory underlying the test:

The machine uses what you call floures-cence polarization amino acid technology which deals with antigens and antibodies that are in the blood system, and the antigens being the drugs.... Well, what the machine does is, you see, the antigens are in your bloodstream. That’s the drugs, but they are not in our system long enough for your body to produce antibodies to attack them so what the company does is they send you reag-ins [sic] which you use which what they do is they inject lab rats with these drugs so they can produce the antibodies, and what happens is the antibodies attach to the antigens and then the drug machine and these antibodies have a flourescent tag on them and when the light is shown through that is what gives you the reading.

Mr. Perez testified that he had learned that the machine was ninety-five to ninety-six percent accurate, but he conceded that he did not know the technical aspects of the machine’s operation.

At the close of evidence, appellant re-urged his Rule 702 objections to Mr. Perez’s testimony and to the lab report. He argued that Mr. Perez “did not know anything about the scientific theory underlying the test, whether the scientific theory was valid, whether any techniques used in applying the theory [were] valid, whether or not the technique was properly applied in this case.” The trial court, however, ruled that the evidence was reliable and noted that Mr. Perez had testified on this subject in other cases.

The Corpus Christi Court of Appeals held that Mr. Perez’s testimony and the lab report were inadmissible because the State failed to satisfy the scientific reliability requirements set out in Kelly.2, The court of appeals also noted that it had previously held Mr. Perez’s testimony inadmissible for the same reason in an earlier, unpublished opinion involving this same appellant.3

II.

A party seeking to introduce evidence of a scientific principle need not *29always present expert testimony, treatises, or other scientific material to satisfy the Kelly test. It is only at the dawn of judicial consideration of a particular type of forensic scientific evidence that trial courts must conduct full-blown “gatekeep-ing” hearings under Kelly. Once a scientific principle is generally accepted in the pertinent professional community and has been accepted in a sufficient number of trial courts through adversarial Dau-bert/Kelly hearings, subsequent courts may take judicial notice of the scientific validity (or invalidity) of that scientific theory based upon the process, materials, and evidence produced in those prior hearings.4

Similarly, once some courts have, through a Daubert/Kelly “gatekeeping” hearing, determined the scientific reliability and validity of a specific methodology to implement or test the particular scientific theory, other courts may take judicial notice of the reliability (or unreliability) of that particular methodology.5

Trial courts are not required to reinvent the scientific wheel in every trial. However, some trial court must actually examine and assess the reliability of the particular scientific wheel before other courts may ride along behind. Some court, somewhere, has to conduct an adversarial gatekeeping hearing to determine the reliability of the given scientific theory and its methodology.6

In this case, appellant objected under Rules 702-705 to the technician’s testimony concerning the results produced by an ADx analyzer testing machine. Appellant did not argue that the underlying scientific theory of urinalysis, as a mode of determining whether a person has consumed a certain substance, is scientifically invalid. *30His objection went to the machine itself. He stated:

There has been no proper predicate to establish the ... reliability of the testing procedures and this gentleman has not been qualified as an expert. We don’t know anything about the machine, whether it’s reliable. Based on Rule 702 and Rule 705, I would object that this burden by the State has not been met, Your Honor.

The trial court simply stated:

The procedures meet the requirements of law. The Court has upheld the procedure in other cases that the witness Alonzo Perez testified to.

The fact that a trial court has allowed some type of scientific testimony by a particular witness before (perhaps without objection) does not mean that the witness’ testimony is, ipso facto, scientifically reliable in this case. Nor does the fact that the trial court has allowed this witness to testify to these procedures before explain how or why the ADx machine is a scientifically reliable one for determining the presence of a controlled substance. It may well be scientifically reliable, but the trial court’s statement that he has allowed this testimony before does not make it so. Perhaps the trial judge had previously conducted numerous Daubert/Kelly gatek-eeping hearings on precisely this issue and had repeatedly found it scientifically reliable. If he had, then either the State or the trial judge should put that on the record along with materials from those previous hearings.7 There is no other evidence or material in this trial record, however, that would support any finding of the scientific reliability of the ADx analyzer.

Thus, the Corpus Christi Court of Appeals did not err. It found that the trial court abused its discretion in admitting the results of “an ADX analyzer” without any showing of its scientific reliability or any reliance upon other scientific materials or judicial opinions which had found “an ADx analyzer” a reliable methodology for determining whether a person does or does not have marijuana in his body.8

In his brief to this Court, the State Prosecuting Attorney presents a plethora of cites to scientific articles and learned treatises, as well as to some cases from other jurisdictions concerning this general area of scientific endeavor. This is swell stuff. The trial court should have been given this material, and appellant should have been allowed an opportunity to cross-examine any witnesses who sponsored it. The trial court hearing is the main event for Daubert/Kelly gatekeeping hearings; it is not a try-out on the road to an appellate scientific seminar.

The State had the burden of proof at trial (or, as in this case, at the probation revocation hearing) to show, by clear and convincing evidence, that the ADx analyzer is a reliable method of determining the presence of marijuana in a person’s body.9 *31It failed to offer any testimony, any scientific material, or any published judicial opinions from which the trial court might take judicial notice of its scientific reliability.10 It cannot now rely upon the appellate courts to become independent scientific sleuths to ferret out the appropriate scientific materials which could support the trial court’s decision to allow the ADx technician’s testimony.11 Thus, the court of appeals was confronted with a trial ree-ord which did not support the scientific reliability of the ADx machine.12 It cannot be faulted for concluding that, based upon the record before it, the State had failed to show the machine’s reliability.

Although appellate courts may take judicial notice of other appellate opinions concerning a specific scientific theory or methodology in evaluating a trial judge’s Daubert/Kelly “gatekeeping” decision,13 ju-*32dieial notice on appeal cannot serve as the sole source of support for a bare trial court record concerning scientific reliability.

Therefore, we affirm the judgment of the court of appeals.

dissenting in which HERVEY, J., joined.

I agree with the Court’s answers to the State’s first and second grounds for review. But I would answer the third ground “perhaps,” and remand to the Court of Appeals for further review.

In its third ground, the State argues that the Court of Appeals erred because in this case, the scientific theory and technique behind the ADx analyzer are well-established. In support of its argument, the State relies on scientific literature which it did not present to the trial court or to the Court of Appeals.

The Court of Appeals conducted an abuse of discretion review, confining itself to the trial record, and found nothing to establish the reliability of the scientific theory or technique. The dissent, on the other hand, conducted a de novo review, considering scientific literature outside the record, and found the science reliable.

A majority of this Court concludes that the State’s literature is “swell stuff’1 but it can not be considered because the State did not present it to the trial court. So this Court, like the Court of Appeals, is applying an abuse of discretion review.

But determining the proper standard of review for Kelly claims is a complicated endeavor. While we have previously stated that it is an abuse of discretion standard,2 we have not addressed the issue directly. Instead, we have stated the standard of review without discussion. But the appropriate standard of review in this context is a subject of heated debate nationwide.

The confusion began when the Supreme Court decided Daubert but did not set forth the appropriate standard of review on appeal.3 Under Frye, appellate courts had reviewed trial courts’ “general acceptance” conclusions de novo,4 but under *44Daubert ⅛ new test, the standard of review was unclear.5

Courts responded to the confusion in various ways. In Florida, the court retained the Frye standard and reviewed the trial court’s “general acceptance” conclusion de novo.6 The Florida Supreme Court explained that “the general acceptance issue transcends any particular dispute .... Application of less than a de novo standard of review to an issue which transcends individual cases invariably leads to inconsistent treatment of similarly situated claims.” 7

Before Daubert, the Maryland Supreme Court had explained that “[t]he answer to the question about the reliability of a scientific technique or process does not vary according to the circumstances of each case. It is therefore inappropriate to view this threshold question of reliability as a matter within each trial judge’s individual discretion.”8 After Daubert, the court continued to follow the Frye test and review Rule 702 claims de novo.9

Illinois also continued to follow Frye, not because of any reasoned formal decision to do so but simply because litigants there did not argue for adopting the Daubert test.10 But the standard of review applied in that state is unclear. In People v. Miller, the Illinois Supreme Court stated that Frye claims are to be reviewed for an abuse of discretion.11 Despite this statement, the court proceeded to consider scientific journals outside the record.12 Justice McMorrow criticized the majority for setting forth the abuse of discretion standard but then not applying it.13 He argued that “the all-encompassing abuse of discretion standard ... does not permit a reviewing court to adequately address the legal issues” in these cases and advocated a “mixed standard of review” in which trial court decisions concerning the expert’s qualifications and his testimony’s relevance be reviewed for abuse of discretion but decisions regarding whether a technique has achieved general acceptance be reviewed de novo.14 He reasoned that a theory or technique’s general acceptance in the scientific community “transcends any particular dispute” and “[ajpplication of less than a de novo standard of review to an issue which transcends individual cases invariably leads to inconsistent treatment of similarly situated claims.”15

Federal courts also expressed confusion after Daubert. The Sixth Circuit was particularly torn, with one panel setting forth a three-fold standard of review, and a second panel rejecting it. The first panel explained that the abuse-of-discretion standard of review “does not adequately describe the standard of review required by the Federal Rules of Evidence for the *45admissibility of expert opinion evidence.”16 The court set forth its opinion of the proper standard: the trial court’s preliminary fact-finding under Rule 104(a) is reviewed for clear error; the court’s determination of whether the expert’s opinion is properly the subject of scientific, technical, or other specialized knowledge is reviewed de novo; and the court’s determination whether the expert’s opinion will assist the trier of fact is reviewed for an abuse of discretion.17

But the other panel disagreed with this approach, concluding that the three-tiered standard of review lacked any precedential support.18 Nevertheless, even this second panel acknowledged that “the de novo prong has some intuitive appeal to it ... particularly ... with respect to the validity of various types of scientific evidence.”19 The court admitted that “a de novo standard ... ensures that there are not conflicting pronouncements in a circuit regarding the validity of scientific methods,” but cautioned that “the distinction between scientific principles and how those principles are applied in a particular dispute should be recognized.”20

Several states adopted this hybrid standard of review, looking de novo at the trial court’s determination that a theory or technique was reliable, but reviewing the application of the science in the particular case for an abuse of discretion. The West Virginia Supreme Court held that “[t]he trial court’s determination regarding whether the scientific evidence is properly the subject of scientific, technical, or other specialized knowledge is a question of law that we review de novo. On the other hand, ... [ajppellate review of the trial court’s rulings under the relevancy requirement is under an abuse of discretion standard.”21

The Oklahoma Court of Criminal Appeals agreed. It concluded that, “[ajfter ... considering the permanent impact of a trial judge’s decision to admit novel scientific evidence, we find we should subject that decision to an independent, thorough review and not simply ask whether an abuse of discretion was committed.” The court, quoting the District of Columbia court, explained that, “[generally, the decision whether or not to admit expert testimony is addressed to the sound discretion of the trial court. Where the question of the general acceptance of a new scientific technique is raised, however, the proponent will often be asking the court to establish the law of the jurisdiction for future cases_Accordingly, in recognition of the fact that the formulation of the law of this jurisdiction is a quintessentially appellate function, we engage in a broad review of the trial judge’s determination whether the forensic use of DNA technology has gained general acceptance. In doing so, we may consider not only expert evidence of record, but also judicial opinions in other jurisdictions, as well as pertinent legal and scientific commentaries.”22

Massachusetts also set forth a dual standard of review, reviewing the issue of scientific validity de novo but reviewing the application of the technique with deference *46to the trial court’s ruling.23 The Supreme Judicial Court conceded that it was departing from the abuse of discretion standard of review used by federal appellate courts, but concluded that it was appropriate to review scientific validity decisions de novo “because the validity of a scientific methodology does not vary according to the circumstances of a particular case and because a lower court’s conclusion will have applicability beyond the facts in the case before it.”24

New Jersey agreed. In State v. Harvey, that state’s Supreme Court acknowledged that “[generally, appellate courts review a trial court’s determination of the admissibility of evidence for an abuse of discretion.”25 But the court nevertheless concluded that review of the reliability determination should be de novo. The court reasoned that, “[ujnlike many other evi-dentiary issues, whether the scientific community generally accepts a methodology or test can transcend a particular dispute.”26 The court felt that, “[njotwithstanding the trial court’s better position to shape the record and make factual determinations, appellate courts retain an important residual role for questions concerning the admission of scientific evidence. Like trial courts, appellate courts can digest expert testimony as well as review scientific literature, judicial decisions, and other authorities.” 27 The court concluded that, “[wjhen reviewing a decision on the admission of scientific evidence, an appellate court should scrutinize the record and independently review the relevant authorities, including judicial opinions and scientific literature.” 28

During this time period, legal writers rushed to criticize Daubert and urge the Supreme Court to adopt a de novo standard of review. One writer warned that “[ijf Daubert decisions are reviewed [for an abuse of discretion] ..., inconsistent decisions concerning the admissibility of novel scientific testimony may go unchecked from jurisdiction to jurisdiction and from judge to judge. This inconsistent standard of review inevitably may ... confound efforts to provide uniformity under the Rules.”29

Another author urged “courts ... [to] consider developing a hybrid approach in which appellate courts review de novo those decisions involving general scientific propositions, but allow trial courts greater discretion with regard to the particular facts of each case.”30 Another agreed, arguing that “[t]he gatekeeping function assumes that trial judges possess some sophistication and experience in scientific matters ... [but][t]here is little reason to believe that trial judges can readily equip themselves with such expertise.”31 This author proposed “de novo appellate review of district court rulings on the scientific *47knowledge prong of Daubert.” He explained that

[e]xamination of scientific theories or methodologies to determine whether they have evolved sufficiently to amount to scientific knowledge is a task that lends itself to de novo appellate review. Such an exercise would have broad implications on subsequent consideration of the same methodology in future litigation. District court rulings on the second prong of Daubert (the fit requirement), which requires consideration of specific facts in a case, would still be reviewed under an abuse of discretion standard. But de novo appellate review of district court findings on the scientific knowledge prong of Daubert would create a body of appellate opinions that carefully review scientific theories and methodologies. As appellate courts repeatedly face the same sorts of scientific evidence, more uniform adjudication at the trial and appellate levels will result. In addition, careful appellate scrutiny would permit consideration and development of distinct validation criteria for expert testimony relating to different scientific or technical disciplines. Finally, appellate courts are also well situated to consider the broad public policy issues associated with admissibility determinations.32

In 1997, in General Elec. Co. v. Joiner,33 the Supreme Court stated, without elaboration, that federal district judges’ Dau-bert rulings were to be reviewed for an abuse of discretion.34 In that case, the issue before the Court was whether an appellate court could apply a different standard of review when the district court admits scientific evidence versus when it excludes scientific evidence. The Supreme Court firmly rejected this notion, concluding that the same standard of review should apply in either instance.35 The Court then stated that abuse of discretion was the appropriate standard of review, but the Court did not consider the inconsistencies that this standard would cause, nor did it address the possibility of a hybrid standard of review.36

After Joiner, most courts employ an abuse of discretion standard of review. The Sixth Circuit put the dispute in that circuit “to rest” and adopted an abuse of discretion standard.37 And Massachusetts did an about-face on the issue, switching its standard of review based on the Joiner opinion.38

But not every state court danced to the Supreme Court’s tune. An Arizona appellate court continued to apply a de novo standard of review without acknowledging Joiner.39 And Minnesota rejected Dau-bert in its entirety in an effort to retain a de novo standard of review.40 The court stated that “the potential for non-uniformity in the law under Daubert [gave it] ... *48considerable cause for concern,”41 noting that “[c]ases built on similar facts and offering similar scientific techniques could have widely disparate results.”42 The court concluded that its “Frye-Mack”43 standard, with the accompanying de novo appellate review, was “more apt to ensure objective and uniform rulings as to particular scientific methods or techniques.”44

The Alaska Supreme Court followed Joiner in applying an abuse of discretion standard of review, but its decision was met with vigorous dissent. Justice Fabe contended that “[t]he determination of whether a general scientific proposition or process is reliable should not vary from case to case or from judge to judge.”45 He proposed “a hybrid standard under which we would review de novo any threshold rulings on scientific validity but would review for abuse of discretion case-specific rulings on relevance and testing performance.”46 He argued that a straight abuse of discretion standard would “most likely lead to inconsistent treatment of similarly situated claims” which would in turn “compromise the integrity of the judiciary in the eyes of the public.”47

Moreover, the critiques of Joiner were many. One legal writer complained that “[t]he Court gave trial judges too much discretion on a topic they know little about, with no guidance for making the admissibility decisions. The result will undoubtedly be rampant individualized decision-making; judges will apply a number of different criteria or general principles to decide whether to permit a nonscientific expert to testify.”48 Another pointed out that “the broad discretion judges now have in screening all types of experts will likely cause increased uncertainty among lawyers regarding the admissibility of expert testimony.”49

One author explained that “there does not seem to be any sound basis for assuming the appellate court is any less competent in reviewing scientific evidence de novo than the trial judge. The traditional reason for deferring to the trial judge on evidentiary rulings is that the trial judge has the benefit of reviewing the demeanor of the witness to determine whether the witness is lying. Because it is not the witness’s credibility at stake in scientific evidence, but the validity of the underlying methodology and application, it is not necessary to give the trial judge’s observation any particular weight. Moreover, the jurisprudential policy issues transcending the facts of a given case are more appropriately resolved by appellate courts.”50

*49A bifurcated standard of review is not unprecedented in Texas. We have applied such a review to trial court rulings on motions to suppress evidence51 and motions for speedy trial.52 We have recognized that an abuse of discretion review is not necessarily appropriate in the context of application of law to facts when the decision does not turn on the credibility or demeanor of witnesses.53 Instead, an appellate court must conduct a de novo review when “the trial judge is not in an appreciably better position than the reviewing court to make that determination.” 54 With Kelly issues, the trial court is not in an appreciably better position in determining whether a particular scientific theory or technique is valid. Although the credibility of the testifying expert is one factor in Kelly, it is not the only factor, and the other factors do not involve credibility of witnesses. I believe appellate courts should defer to the trial court’s credibility determination but review the validity of the science de novo.

There is good reason to believe that the Supreme Court would agree with a bifurcated standard of review. As explained previously, the Supreme Court was not faced in Joiner with deciding whether to apply a hybrid standard of review. Instead, the Court merely addressed whether the same standard of review should apply to trial court decisions admitting evidence and trial court decisions excluding evidence. That Court would likely adopt a bifurcated standard of review if the issue were directly before it.

But even if the Supreme Court firmly believes that all aspects of a Daubert claim should be reviewed on appeal for an abuse of discretion, its conclusion should not prevent us from applying a different standard of review in Texas. Indeed, our review of Kelly claims already differs substantially from the Supreme Court’s review of Dau-bert claims. First, we hold that the gatek-eeping hearing is mandatory55 while the Supreme Court finds it optional.56 Additionally, we hold that the proponent of the evidence must prove its admissibility by clear and convincing evidence,57 while the Supreme Court requires proof only by a preponderance of the evidence.58 Admission of scientific evidence is already more difficult in Texas than in the federal courts. Consistent with that approach, it is appropriate to apply a heightened standard of review on appeal.

The abuse-of-diseretion standard is wholly inadequate in the context of Kelly claims. We should adopt a bifurcated standard of review, reviewing the reliability of a scientific theory or technique de novo, but reviewing the application of the technique in each particular case for an abuse of discretion.

Under a de novo review of a Kelly claim, the appellate court would be able to consider literature not presented to the trial court. It would be the appellate court’s duty to determine de novo whether the *50particular scientific theory or technique is valid, and it would be able to reach that determination by reviewing any rehable authority it could locate, regardless of whether the trial judge reviewed it.

Courts across the country have considered scientific literature outside the record in conducting a de novo review of Daubert claims.59 Indeed, in all the cases discussed in the previous section, the courts understand that reviewing the validity of a scientific theory or technique de novo means reviewing literature outside the record.

Despite any implications to the contrary, “[t]rial courts are no better situated to assess the validity of scientific methods.”60 This is because in most cases, “[s]uch determinations do not ... depend on assessing the credibility of witnesses or knowledge of local conditions.”61 Indeed, “judicial assessment of validity might be better conducted through written briefs rather than oral testimony.”62

In one of the earlier discussions on the topic, Kenneth Karst explains that “[a] judge without special training finds it hard to criticize an expert’s methods of gathering data or the inferences he draws from them. Yet experts are fallible. Worse, they have axes to grind — at least professional axes.”63 He continues by noting that “ultimately there is no way for the conscientious judge to escape making a determined effort to learn enough about the subject matter of the litigation to formulate his own views with the assistance— not the domination — of the experts.”64

Judge Ferren of the D.C. Appeals Court discussed this issue in depth in Dean v. District of Columbia,65 He noted that “it is questionable whether a hearing with expert testimony about issues of legislative fact Vould reveal more reliable or higher quality information than is available by referring to authorities submitted in briefs by both sides, and, in appropriate cases, by additional research at the appellate level.’ ”66 As he explains, “[t]he advantage of such a costly exercise is likely, on too many occasions, to be marginal at best, and the further away the hearing is from the ideal model I have posited (using the most highly qualified experts), the more the judge will confront distorted, perhaps even biased, presentations and thus have to rely primarily on many non-record sources for critical scientific or social science information.”67

He adds that “[t]he point not to be forgotten ... is that the likelihood of ex*51pert testimony of record being superior to the court’s own resourcefulness in finding legislative facts is often highly questionable. The quality of the paid experts too often will be limited to persons who have not done the most careful studies themselves and, in any event, may have an ax to grind that, absent very skillful cross-examination, can create a record that hides the real truth- Arguably, the more complex the issue ... the greater the risk that a hearing will yield no better, and perhaps less satisfactory, results than non-record sources, including a judge’s own research into primary data, helped along, of course, by counsel’s advocacy.”68

Additionally, consulting literature outside the record as part of a de novo review permits the appellate court to change with the scientific times. Several states recognize that “[i]n the rapidly changing world of modem science, continuing research may affect the scientific community’s acceptance of a novel technology. By reviewing post-trial publications, an appellate court can account for the rapid pace of new technology. The continuing review also recognizes that general acceptance may change between the time of trial and the time of appellate review.”69 Moreover, “by examining such additional information, an appellate court can prevent any injustice rendered by the admission or exclusion of the evidence at the trial level.”70

As the Arizona court recognized, when an appellate court does this, “[i]t is somewhat incongruous to call the trial court’s ruling ‘error.’ ”71 Nevertheless, that court concluded that “neither logic nor authority supports confining ourselves to a snapshot, rather than viewing the motion picture, of technological advancement. If the result obtained is the product of invalid scientific theory, there is no good reason to accept it simply because we were fooled at the inception of the inquiry_Thus, in surveying all information available, we consider scientific literature published, as well as cases decided, after trial.”72

Finally, it is worth noting that we have previously approved reviewing scientific literature outside the record, although we did so under the theory of “judicial notice.” In Emerson v. State,73 we considered scientific articles outside the record in determining whether the HGN test was reliable. We justified this approach under the notion of “judicial notice.”74 And in Mata v. State,75 we cited Emerson as authority for our taking “judicial notice” of scientific literature outside the record. In retrospect, our use of the phrase “judicial notice” was misplaced in both cases. In cases involving scientific evidence, courts often take judicial notice of the validity of a particular scientific theory or technique.76 In these instances, the court is *52taking judicial notice of a fact — the fact that the theory or the technique is valid. It is not taking judicial notice of the literature. Judicial notice applies to facts, not authority. Instead, the courts take judicial notice of the fact that a theory or technique is valid.77

I would adopt a bifurcated standard of review on appeal. I think appellate courts should review the reliability of the scientific theory or technique de novo, and review the application of that science in the particular case for an abuse of discretion. I would remand to the Court of Appeals to give the appellate court an opportunity to apply this standard of review.

concurring.

We live in a time of great expansion in the world of technology. Diagnostic machines can detect and identify more kinds and increasingly smaller amounts of chemicals. In the context of criminal justice, if the chemical in question is a controlled substance, some defendant somewhere will challenge the validity of the theory and of the process used by the machine. With this case, we move toward a process of our own for the evaluation of such challenges.

It is foreseeable, perhaps even inevitable, that one trial court will hear evidence on a new process and find it reliable and another trial court will hear the same or similar evidence and rule that the process is not reliable. When that situation arises, it will fall to the appellate courts to make a ruling that applies to all trial courts. This has long been the case; the law must be uniform for all citizens in all the courts of this state.

When that time comes, the approach described in Judge Keasler’s dissent is the appropriate one. Because the law must be uniformly applied, the decision must lie with the appellate courts. Because they must make a decision, if the record does not offer sufficient data to determine whether the new process is reliable, the courts must be free to consult other sources. What if Judge A hears the testimony of Expert Z and finds the process reliable, and then Judge B hears the testimony of Expert Z and finds the process unreliable? The findings cannot both be true. It is up to the appellate courts to make the choice.

Ruling that our appellate courts are restricted to consulting the opinions of other appellate courts from other states does not really address the issue or solve the problem. If those other courts have not reviewed the literature, either in the record or on their own, de novo, they have ruled in ignorance, and to follow their conclusions would be folly. If those other courts have reviewed the literature in their effects to discern the truth, then we are relying on decisions made after reviewing the literature de novo, while forbidding the courts of this state to do that same review.

It seems far more wise to allow the appellate court, the body charged with *43maintaining uniformity of law, to address the issue of reliability directly. The appellate courts must be free to consult their own sources and test the strength and weakness of the evidence presented at trial. Most judges (indeed, most lawyers) have little scientific training. It is difficult to evaluate data in an area in which one is not knowledgeable; we should not hamper judges’ efforts to educate themselves in their search for truth. Once the decision on reliability has been made, then, as with breath-alcohol testing and many other new processes, the question of whether the theory and process were properly applied will fall to the discretion of the trial court.

Our decision today begins the development of our own process for evaluation of scientific reliability; the first step is to create a record at trial that provides to an appellate court sufficient information on which to base a decision on reliability.

I join the opinion of the Court.