United States v. Patrick Leroy Crisp

MICHAEL, Circuit Judge,

dissenting:

The majority believes that expert testimony about fingerprint and handwriting identification is reliable because the techniques in these fields have been accepted and tested in our adversarial system over time. This belief leads the majority to excuse fingerprint and handwriting analysis from the more careful scrutiny that scientific expert testimony must now withstand under Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), before it can be admitted. In Patrick Leroy Crisp’s case the government did not prove that its expert identification evidence satisfied the Daubert factors or that it was otherwise reliable. I respectfully dissent for that reason. In dissenting, I am not suggesting that fingerprint and handwriting evidence cannot be shown to satisfy Daubert. I am only making the point that the government did not establish in Crisp’s case that this evidence is reliable. The government has had ten years to comply with Daubert. It should not be given a pass in this case.

I.

The Daubert case lists five factors for assessing the reliability of expert scientific testimony: (1) whether the expert’s theory can be or has been tested; (2) whether the theory has withstood peer review and publication; (3) whether there is a known or potential rate of error; (4) whether standards exist for the application of the theory; and (5) whether the theory has been generally accepted by the relevant scientific community. Daubert, 509 U.S. at 593-94, 113 S.Ct. 2786. These factors are not meant to be exclusive or necessarily dis-positive. Id. However, when “the Daubert factors are reasonable measures of the [expert] testimony’s reliability, the Supreme Court has instructed that the trial judge should consider them.” United States v. Lewis, 220 F.Supp.2d 548, 551 (S.D.W.Va.2002) (citing Kumho Tire Co. v. Carmichael, 526 U.S. 137, 152, 119 S.Ct. 1167, 143 L.Ed.2d 238 (1999) (emphasis added)).

The majority excuses fingerprint and handwriting analysis from any rigorous Daubert scrutiny because these techniques are generally accepted and have been examined for nearly one hundred years in our adversarial system of litigation. These circumstances are not sufficient to demonstrate reliability in the aftermath of Dau-bert. To say that expert evidence is reliable because it is generally accepted is to say that it is admissible under Daubert because it was admissible under the old rule articulated in Frye v. United States, 293 F. 1013, 1014 (D.C.Cir.1923) (allowing expert evidence that had “gained general acceptance in the particular field in which it belongs”). Frye’s “general acceptance” rule was replaced by Fed.R.Evid. 702, which now requires expert testimony to be “the product of reliable principles and methods.” Daubert, of course, outlines the factors that are relevant to the determination of reliability. Nothing in the Supreme Court’s opinion in Daubert suggests that evidence that was admitted under Frye is grandfathered in or is free of the more exacting analysis now required. See United States v. Saelee, 162 F.Supp.2d 1097, 1105 (D.Alaska 2001) (“[T]he fact that [expert] evidence has been generally accepted in the past by courts does not mean that it should be generally accepted now, after Daubert and Kumho.”).

Nor is fingerprint and handwriting analysis necessarily reliable because it has been subjected to the adversarial process *273of litigation. In a criminal case like this one, adversarial testing simply means that the defense lawyer cross-examines the government’s expert. That, I concede, is important, but it only goes part way. In most criminal cases, particularly those in which the defendant is indigent, the defendant does not have access to an independent expert who could review the analyses and conclusions of the prosecution’s expert. Simon Cole, Suspect Identities: A History of Fingerprinting and Criminal Identification 280 (2001) [hereinafter Cole, Suspect Identities ] (noting that defense lawyers rarely challenge fingerprint evidence, in part because they often do not have the funds to hire experts). Lack of money is only one problem. Lack of independent crime laboratories is another. The great majority of crime laboratories are operated by law enforcement agencies. Paul C. Giannelli, The Abuse of Scientific Evidence in Criminal Cases: The Need for Independent Crime Laboratories, 4 Va. J. Soc. Pol’y & L. 439, 470 (1997); Paul C. Giannelli, “Junk Science”: The Criminal Cases, 84 J.Crim. L. & Criminology 105, 118 (1993). More important, criminal defendants do not appear to have access to experts who could challenge the basic principles and methodology of fingerprint and handwriting analysis. Jennifer L. Mnookin, Fingerprint Evidence in an Age of DNA Profiling, 67 Brooklyn L.Rev. 13, 38-39 (2001) [hereinafter Mnookin, Fingerprint Evidence ] (explaining that fingerprint evidence came to be seen as particularly powerful in part because it was so rarely challenged by the defense); Cole, Suspect Identities, supra at 280 (reporting that New York City police officers caught fabricating evidence chose to create fingerprint evidence because it was so unlikely to be challenged). Our adversarial system has much to commend it, but it is not a general substitute for the specific Daubert inquiry. The system without Daubert did not work to ensure the reliability of fingerprint and handwriting analysis. As I point out in parts II.B. and III infra, fingerprint and handwriting analysis was admitted with little judicial scrutiny for decades pri- or to Daubert.

Nothing in the history of the use of fingerprint and handwriting evidence leads me to conclude that it should be admitted without the scrutiny now required by Dau-bert. The government, of course, has the burden to put forward evidence “from which the court can determine that the proffered testimony is properly admissible” under Daubert. Md. Cas. Co. v. Therm-O-Disc, Inc., 137 F.3d 780, 783 (4th Cir.1998). The government utterly failed to meet its burden here.

II.

A.

At Crisp’s trial the government’s fingerprint identification evidence failed to satisfy any of the Daubert requirements for establishing scientific reliability. The first Daubert factor is whether the technique has been tested. The government did not offer any record of testing on the reliability of fingerprint identification. See J.A. 361 (testimony of the government’s fingerprint expert, an employee of the North Carolina Bureau of Investigation, stating that she was not aware of any testing on the validity of the science). Indeed, it appears that there has not been sufficient critical testing to determine the scientific validity of the technique. See United States v. Llera Plaza, 188 F.Supp.2d 549, 564 (E.D.Pa.2002); Robert Epstein, Fingerprints Meet Daubert: The Myth of Fingerprint “Science” Is Revealed, 75 S. Cal. L.Rev. 605, 624-26 (2002); David A. Stoney, Fingerprint Identification: The Scientific Basis of Expert Testimony on Fingerprint Identification, in 3 Modem *274Scientific Evidence: The Law and Science of Expert Testimony § 27-2.0, § 27-2.1.2[6] (David L. Faigman et al. eds., 2002). Specifically, with respect to forensic fingerprint examination, there have not been any studies to establish how likely it is that partial prints taken from a crime scene will be a match for only one set of fingerprints in the world. Stoney, supra at § 27-2.3.2 (“The issue is not the finding of two fingerprints that are alike, but rather the finding of prints from two different fingers that can be mistakenly judged to be alike by a fingerprint examination.”). Although the government introduced evidence that its fingerprint expert in this case had taken and passed proficiency tests, see J.A. 362-63, this evidence gave no basis for a conclusion that these proficiency tests reflect real world conditions. Proficiency testing is typically based on a study of prints that are far superior to those usually retrieved from a crime scene. Llera Plaza, 188 F.Supp.2d at 565 (acknowledging that proficiency tests may not reflect real world conditions); compare also Lewis, 220 F.Supp.2d at 554 (noting that proficiency tests are inadequate when everyone passes), with J.A. 362 (testimony of the government’s fingerprint expert in this case, saying that she always achieved a perfect score on proficiency tests). The government did not introduce evidence of studies or testing that would show that fingerprint identification is based on reliable principles and methods.

The second Daubert factor is whether the science or technique has been subjected to peer review and publication. Again, the government offered no evidence on this factor at trial. Fingerprint examiners, like other forensic scientists, have their own professional publications. Epstein, supra at 644. But unlike typical scientific journals, the fingerprint publications do not run articles that include or prompt critique or reanalysis by other scientists. Indeed, few of the articles address the principles of fingerprint analysis and identification at all; rather, most focus on the process of lifting fingerprints from crime scenes. Epstein, supra at 644. This lack of critical analysis in the fingerprint identification field has had a predictable effect. Unlike traditional scientific fields where criticism and vibrant exchange of ideas have led to dramatic advances, the techniques used by fingerprint analysts have changed little over the years. Simon Cole, What Counts for Identity? The Historical Origins of the Methodology of Latent Fingerprint Identification, Sci. in Context, Spring 1999, at 139, 165 (noting that little change has taken place in the methodology of analyzing latent prints).

The third Daubert factor calls for consideration of the known or potential rate of error. The government has not tested the reliability of fingerprint identification, so it ignored the error rate factor in this case. J.A. 360 (testimony of government’s expert that “[a]s far as statistics, off the top of my head at this point, I cannot give you any. I do know that ... errors have been made in the field of fingerprints.”); see also Epstein, supra at 633. Some courts have merely assumed that the rate of error in fingerprint identification is low. See Llera Plaza, 188 F.Supp.2d at 566 (concluding that the absence of evidence of high error rates means that the error rate is not unacceptably high). And that may be. But an error rate must be demonstrated by reliable scientific studies, not by assumption. Nor is it sufficient after Dau-bert for a proponent simply to show that a particular fingerprint examiner scores well on proficiency tests. First, it is unclear whether the proficiency tests taken by the examiner in this case were representative of real life conditions. Cf. Llera Plaza, 188 F.Supp.2d at 565 (acknowledging that proficiency tests may not reflect real world *275conditions). Second, where tests have attempted to imitate actual conditions, the error rates have been alarmingly high. Epstein, supra at 634. In a 1995 test conducted by a commercial testing service, less than half of the fingerprint examiners were able to identify correctly all of the matches and eliminate the non-matches. On a similar test in 1998, less than sixty percent of the examiners were able to make all identifications and eliminations. Id. at 634-35. An error rate that runs remarkably close to chance can hardly be viewed as acceptable under Daubert.

The fourth Daubert factor asks whether there are universal standards that govern the application of the technique. The government did not establish that there are such standards. Its expert asserted that her department had controlling standards, yet when pressed on the point, she admitted that the degree of similarity required to find that prints are matching “is left up to each individual examiner.” J.A. 363. As one forensic expert contends, “[a]ny unbiased, intelligent assessment of fingerprint identification practices today reveals that there are, in reality, no standards.” Stoney, supra § 27-2.3.1 [2], Many fingerprint examiners testify in terms of matching points, that is, the number of similarities between the ridges in the print taken from the crime scene and the ridges in the defendant’s known print. But the trend has been toward eliminating any requirement for a minimum number of matching points before an opinion can be given that a latent print and a known exemplar are attributable to the same person. See J.A. 363 (testimony of the government’s fingerprint expert that no minimum number of points is required); Llera Plaza, 188 F.Supp.2d at 570 (somehow concluding that the fingerprint examination field has uniform standards because most examiners agree that no minimum number of points is required to confirm a match). The trend away from a minimum-point requirement may not be unreasonable because the requirement, although adopted by some agencies (and countries), is not based on scientific study. Epstein, supra at 637 (quoting a fingerprint expert as saying that the point system is based on “educated conjecture”); Cole, Suspect Identities, supra at 270 (discussing Britain’s eventual rejection of the sixteen-point minimum). Examiners have not, however, been able to replace the point system with anything more concrete. Epstein, supra at 638-39; Cole, Suspect Identities, supra at 268-69. There is even disagreement as to what aspects of the fingerprint the examiner should rely on. One prominent expert rejects traditional reliance on ridge characteristics and calls on examiners to look at other details such as sweat pores and ridge edges. Epstein, supra at 639; Cole, Suspect Identities, supra at 267. Others, however, vehemently reject this approach, explaining thatvariations in these particular details are especially common because of differences in pressure, residue on the fingers, the condition of the surface on which the print is left, and processing techniques. Epstein, supra at 639-40. All of this leads one expert to conclude that “[t]he criteria for absolute identification in fingerprint work are subjective and ill-defined. They are the product of probabilistic intuitions widely shared among fingerprint examiners, not of scientific research.” Stoney, supra § 27-2.3.1[1]. See also Cole, Suspect Identities, supra at 268-69.

Further, even the safety checks that are thought to be universally accepted are not consistently followed. For example, fingerprint experts are supposed to reject as matching a pair of prints that contain even one dissimilarity. Epstein, supra at 640. At least one expert, however, has said that when fingerprint examiners believe the *276prints are a match, they explain away the differences rather than discounting the match. Epstein, supra at 640-41. Moreover, independent verification of a match by a second examiner is considered to be essential. See Cole, Suspect Identities, supra at 269; Epstein, supra at 641. Yet in many cases, including this one, no verification takes place. See ante at 268 n. 4 (noting that no independent review took place in this case); Epstein, supra at 641; Cole, Suspect Identities, supra at 282 (explaining that an error made by Scotland Yard was attributed to the fact that independent verification did not take place); see also Cole, Suspect Identities, supra at 280-81 (detailing extensive fabrication of fingerprint evidence in the New York City Police Department that was not uncovered sooner in part because no independent verification took place). Moreover, any verification that does take place is not independent in the truest sense. The reviewer is usually a supervisor or colleague in a forensic lab associated with law enforcement, so the reviewer may share the same inclinations as the original examiner. See Cole, Suspect Identities, supra at 269.

In short, the government did not establish that there are objective standards in the fingerprint examination field to guide examiners in making their comparisons.

The fifth (and final) Daubert factor is whether the technique has been generally accepted in the relevant scientific community. I acknowledge, of course, that the general public, which sees movies and television programs that regularly portray fingerprinting and other forensic techniques as key to crime solving, regards fingerprint identification as perfectly reliable. Moreover, several circuit courts since Daubert have held — without going deeply into the question — that fingerprint evidence is admissible. See United States v. Hernandez, 299 F.3d 984, 991 (8th Cir. 2002); United States v. Havvard, 260 F.3d 597, 601 (7th Cir.2001); United States v. Sherwood, 98 F.3d 402, 408 (9th Cir.1996). But “[t]he Daubert court did not suggest that acceptance by a legal, rather than a scientific community, would suffice.” United States v. Starzecpyzel, 880 F.Supp. 1027, 1038 (S.D.N.Y.1995). The fingerprint examination community is certainly a proponent of the technique. That community’s enthusiasm, however, must be subjected to objective scrutiny if Daubert is to have any meaning. One author asserts that “mainstream scientists, by and large, have ignored the question of whether individuals can be reliably identified through small, distorted latent fingerprint impressions.” Epstein, supra at 646. At least two forensic commentators have expressed concern about the lack of objective scientific research into the reliability of the technique. Id. Nothing in the record in this case shows that the fingerprint examination community has challenged itself sufficiently or has been challenged in any real sense by outside scientists. Accordingly, the government did not establish that the technique has valid, general acceptance in the scientific community. The fifth factor is not satisfied. The government thus failed to demonstrate in this case that fingerprint identification is reliable under the specific Daubert criteria.

B.

Even if the proponent of scientific expert evidence does not satisfy the Daubert factors, the evidence may be admissible if it is otherwise shown to be reliable. Cf. Daubert, 509 U.S. at 593-94, 113 S.Ct. 2786; ante at 268. The government also failed to provide other reasons to establish that its fingerprint evidence in this case is rehable.

Fingerprint identification’s long history of use does not by itself support the deci*277sion to admit it. Courts began admitting fingerprint evidence early last century with relatively little scrutiny, and later courts, relying on precedent, simply followed along. To put it bluntly, the precedent of prior admission, rather than exacting scientific scrutiny, led to its universal acceptance. Cole, Suspect Identities, supra at 186 (“Fingerprint evidence won acceptance without being subjected to the kind of organized skepticism and careful scrutiny that is supposed to be inflicted upon scientific and legal facts.”); id. at 259 (noting that fingerprint evidence had became widely accepted although “latent fingerprint identification was ... not based on scientific research at all[ ][but][i]nstead ... was based on anecdote, experience, and nineteenth century statistics”); Michael J. Saks, Merlin and Solomon: Lessons from the Law’s Formative Encounters with Forensic Identification Science, 49 Hastings L.J. 1069, 1104 (1998) (noting the lack of serious inquiry into the admissibility of fingerprint evidence in the early years). As a matter of fact, other forms of evidence in vogue at the time fingerprint ing began to be commonly used were generally believed to be more credible. Cole, Suspect Identities, supra at 93, 146, 159. For example, experts in the Bertillon technique took minute measurements of the human body — including the bones in the face, arms, and feet, and the shape and size of the ears — to identify criminals. Id. at 34-44. The Bertillon system and its offshoots were widely used in France and were recognized by many states in the United States. Cole, Suspect Identities, supra at 146-49. It, like fingerprinting, was admitted as evidence in criminal cases. See, e.g., State v. Hill, 145 Kan. 19, 64 P.2d 71, 75 (1937); see also Downs v. Swann, 111 Md. 53, 73 A. 653, 654-55 (1909) (upholding as constitutional the use of Bertillon measurements for identification purposes in a criminal case); Cole, Suspect Identities, supra at 146-47 (noting use of Bertillonage and similar, systems in the United States). Today, we consider the Bertillon system to be absurd. See People v. King, 266 Cal.App.2d 437, 72 Cal.Rptr. 478, 483-84 (1968) (noting that we should heed the “tragic lessons of the Bertillon system”). Fingerprinting replaced the Bertillon system. But Bertillonage did not fall out of favor because anyone demonstrated its unreliability or fingerprinting’s superiority. Rather, law enforcement officials found the Bertillon system too cumbersome to use and too complicated to entrust to untrained technicians. See Cole, Suspect Identities, supra at 91, 93, 159. Fingerprinting, on the other hand, rose in popularity because the prints could be taken and analyzed quickly by those with little training or experience. Id. at 159 (“Fingerprinting, then, emerged not as a method of criminal identification superior to anthropometry [Bertillonage] but rather as a quick and cheap, supposedly less scientific way of identifying those whose crimes did not justify the expense of an-thropometry.”). These advantages were seen to outweigh fingerprinting’s primary drawback' — that it was believed to be considerably less reliable than the Bertillon system. Cole, Suspect Identities, supra at 87-88, 93-94. Fingerprint identification’s long history of use, therefore, does not itself establish its reliability.

Fingerprint identification may also be seen as reliable because the examination community prevents its experts from testifying to a match unless they are certain of the match. Fingerprint experts, in other words, refuse to hedge their testimony in terms of probability. 3 David L. Faigman et al., Modem Scientific Evidence: The Law and Science of Expert Testimony § 27-1.0, § 27-1.0 (2002 & Supp.2003). This practice seems to have hastened the technique’s acceptance by courts, who have *278been attracted to its seeming infallibility. Mnookin, Fingerprint Evidence, supra at 36. Professions of absolute certainty by an expert witness, however, seem out of place in today’s courtroom. Even a DNA match has a small chance of being in error. Indeed, there is some suggestion that the certainty requirement for fingerprint identification is a false comfort. In one case, two prints found at a crime scene — identified with certainty by FBI experts as matching the defendant’s exemplars — were sent along with the defendant’s exemplars by the FBI to all fifty state crime laboratories. See D. Michael Risinger et al., The Daubert/Kumho Implications of Observer Effects in Forensic Science: Hidden Problems of Expectation and Suggestion, 90 Cal. L.Rev. 1, 41 (2002). Seven state labs found that one print could not be conclusively matched; five labs said the same about the second print. Id. Faced with this result that violated the tenet that no identification should be made if there was room for disagreement, the FBI sent annotated versions of the fingerprints, indicating important points of similarity, back to the twelve labs that did not find complete matches. Id. With time to reconsider, the initially dissenting labs changed their conclusions to support the FBI’s original identification. Id. The amount of maneuvering it took to reach the certain match requirement in this one case raises doubts about whether this requirement can be relied upon to ensure reliability.

The history of fingerprint identification and the dogged certainty of its examiners are insufficient to show that the technique is reliable. Because of that and the government’s failure to show that its fingerprinting evidence is reliable under the Daubert standards, I conclude that the district court’s decision to admit the fingerprint evidence was an abuse of discretion. Cf. 3 Faigman et al., supra § 27-1.0 (“A judge who takes Daubert’s commands seriously would be hard pressed to write a coherent opinion justifying a decision to admit the expert [fingerprinting] opinion.”)

III.

Handwriting identification evidence has been greeted with more skepticism by courts in the wake of Daubert. Some courts have refused to admit it. See Lewis, 220 F.Supp.2d at 554; Saelee, 162 F.Supp.2d at 1106; see also Starzecpyzel, 880 F.Supp. at 1028 (admitting handwriting evidence only after concluding, prior to Kumho, that Daubert did not apply, and explaining that if Daubert did apply, it “might well have concluded that forensic document examination constitutes precisely the sort of junk science that Dau-bert addressed”); United States v. Hines, 55 F.Supp.2d 62, 68 (D.Mass.1999) (noting that a rigorous application of the Daubert standards to handwriting evidence would reveal “serious problems”). Other courts have allowed testimony about the similarities between handwriting samples without permitting the expert to testify to conclusions about the authorship. See United States v. Rutherford, 104 F.Supp.2d 1190, 1193 (D.Neb.2000) (excluding testimony on the authorship of a document, although allowing uncontested evidence of the similarities and differences between two samples); Hines, 55 F.Supp.2d at 63-64 (allowing admission of testimony about similarities and differences but denying admission of testimony drawing conclusions about authorship). I believe that the government’s evidence on handwriting, like its evidence on fingerprinting, does not demonstrate its reliability, and the evidence should therefore have been excluded. Cf. Andre A. Moenssens, Handwriting Identification Evidence in the Post-Daubert World, 66 UMKC L.Rev. 251, 276-77 (1997) (noting that if Daubert factors were applied to forensic *279sciences, many expert opinions would no longer be admissible).

I will again run through the Daubert factors, considering first whether the technique of handwriting analysis has been tested. The proposition that forensic document examiners can reliably identify handwriting was not established in this case. See Saelee, 162 F.Supp.2d at 1102 (noting the lack of testing); Hines, 55 F.Supp.2d at 68 (concluding that handwriting has never been “subject to meaningful reliability or validity testing”). Starzecpyzel, 880 F.Supp. at 1086 (noting the lack of evidence to support the principle that no two people write identically). This case aside, it appears that no one has ever assessed the validity of the basic tenets of handwriting comparison, namely, that no two individuals write in precisely the same fashion and that certain characteristics of an individual’s writing remain constant even when the writer attempts to disguise them. The government asserted in this case that because these premises had not been disproven, they must be true. See J.A. 884-85; Moenssens, supra at 319-20 (asserting these premises to be true but providing no evidence to support them); cf. Jennifer L. Mnookin, Scripting Expertise: The History of Handwriting Identification Evidence and the Judicial Construction of Reliability, 87 Va. L.Rev. 1723, 1806 (2001) [hereinafter Mnookin, Scripting Expertise ] (discussing basic tenets of handwriting analysis). One researcher has attempted to compare the ability of professional examiners to identify handwriting with the ability of lay persons. See J.A. 332-34. Even with this study, which is discussed below, the data on handwriting analysis is “sparse, inconclusive and highly disputed.” Starzecpy-zel, 880 F.Supp. at 1037; D. Michael Risinger with Michael J. Saks, Science and Nonscience in the Courts: Daubert Meets Handwriting Identification Expertise, 82 Iowa L.Rev. 21, 65 (1996) [hereinafter Ri-singer & Saks, Science & Nonscience ] (“Put simply, if courts trust handwriting experts to be experts, little incentive exists to advance the field’s knowledge or to test its claims. And so, in the past century virtually no research of that kind has been done.”). Moreover, although the government’s expert here testified to his success on proficiency tests, the government provides no reason for us to believe that these tests are realistic assessments of an examiner’s ability to perform the tasks required in his field. See J.A. 342 (testimony of the government’s handwriting expert that he has always achieved a perfect score on proficiency tests); Lewis, 220 F.Supp.2d at 554 (noting that proficiency tests are inadequate when everyone passes); Saelee, 162 F.Supp.2d at 1102 (noting the problems with studies on the error rate of individual examiners as well as the lack of data supporting the underlying premises of the field); see also D. Michael Risinger, Handwriting Identification: The Scientific Status of Handwriting Identification Expertise, in Modem Scientific Evidence, supra § 28-2.0, § 28 — 2.3.8[3] (2002 & Supp.2003) [hereinafter Risinger, Handwriting Identification]. If what little the government said in this case is any indication, the premises upon which handwriting analysis is based have not been exposed to a sufficient amount of objective testing.

The next Daubert question is whether handwriting examination has been subjected to peer review and publication. The government did not present any evidence about peer review or critical scholarship in the field. See, e.g., Hines, 55 F.Supp.2d at 68 (concluding that handwriting analysis has not been subjected to meaningful peer review); Starzecpyzel, 880 F.Supp. at 1037 (explaining that articles on handwriting analysis are “significantly different from scholarly articles in such fields as medicine *280or physics, in their lack of critical scholarship”). Those within the field have failed to engage in any critical study of the basic principles and methods of handwriting analysis, and few objective outsiders have taken on this challenge. Starzecpyzel, 880 F.Supp. at 1038 (concluding that the literature on handwriting analysis “fails to meet the expectations of the Daubert court— that a competitive, unbiased community of practitioners and academics would generate increasingly valid science”); D. Michael Risinger et al., Brave New “Post-Daubert World,” — A Reply to Professor Moenssens, 29 Seton Hall L.Rev. 405, 441 (1998) [hereinafter Risinger et al., Reply ] (“No members of the handwriting identification community are rewarded for doing empirical testing and for examining the claims of the enterprise skeptically.”); Starzecpyzel, 880 F.Supp. at 1038 (identifying relevant fields of science that could be expected to have an interest in document examination but concluding that experts in these fields “are either unfamiliar with forensic document examination, or are critical of the field”). This lack of critical review has hampered the advancement of methodology in the field. Indeed, the field of handwriting analysis, unlike most other technical fields, relies primarily on texts that were written fifty to one hundred years ago. J.A. 335 (“The methodology of handwriting examination, handwriting identification, is consistent with those proposed back as early as the 1900s.”); Star-zecpyzel, 880 F.Supp. at 1038 (describing “the apparent stagnation of research within the [forensic document examiner] community”). The second Daubert factor, peer review and publication, is not satisfied.

The next Daubert factor requires a look at the technique’s known or potential rate of error. Under pressure from courts, handwriting analysis appears to have been subjected to more testing than fingerprint analysis. See Risinger, Handwriting Identification, supra § 28-2.3. In this case, however, the government failed to introduce any evidence about what the error rate might in fact be. See J.A. 332-33 (testimony of the government’s handwriting expert discussing studies in general terms); J.A. 336 (“I would hesitate to say that it has a known rate of error....”); J.A. 338 (“So I, again, would have to say that I’m not aware of any set error rate.... ”). The testing that has been done suggests that experts, on average, do better than non-experts at avoiding false positives, that is, in identifying someone as an author who in fact is not. See Risinger et al., Reply, supra at 421; Risinger, Handwriting Identification, supra § 28-2.3.6[4]. On some tests, however, the best of the non-experts did as well as some of the experts. See Risinger et al., Reply, supra at 421. Even these modest results have been challenged. Id. at 423-29 (noting problems in the methodology of the testing, including motivational differences between experts and non-experts, the lack of controls to prevent sharing of answers among experts, and the lack of similarity between the test and the day-to-day work of document examiners); Risinger, Handwriting Identification, supra § 28-2.3 (discussing the tests). Moreover, other more challenging studies that more accurately reflect real world conditions show higher rates of error. One study found that as many as nine percent of document examiners misidentified a forgery as being written by the named author, and almost one-quarter of the examiners incorrectly concluded that a disguised writing was written by someone other than the true author. Risinger, Handwriting Identification, supra § 28-2.3.8[l]. The error rates in the testing that has been reported are disquieting to say the least. In any *281event, the government did not satisfy the third Daubert factor in this case.

The next Daubert factor focuses on whether there are standards or controls that govern the expert’s analysis. In this ease the government’s expert asserted that handwriting examiners follow the same methodology, J.A. 335-36, but he provided no listing of objective criteria that are used to form an opinion. There does not seem to be any list of universal, objective requirements for identifying an author. J.A. 342-43; Lewis, 220 F.Supp.2d at 554 (explaining that handwriting experts had no set number of similarities required to proclaim the handwriting a match); Saelee, 162 F.Supp.2d at 1104 (“The technique of comparing known writings with questioned documents appears to be entirely subjective and entirely lacking in controlling standards.”); Risinger & Saks, Science and Nonscience, supra at 39 (explaining that because document examiners base their conclusions on their own empirical observations rather than publicly available data, the results are “only as good as the unexaminable personal database of the practitioner[ ] and the practitioner’s not-fully-explainable method of deriving answers”).

The last factor is whether the technique is generally accepted in the scientific community. The general acceptance of handwriting analysis appears to come only from those within the field. Saelee, 162 F.Supp.2d at 1104 (explaining that handwriting analysis has been generally accepted by those in the field); Hines, 55 F.Supp.2d at 68 (concluding that handwriting evidence has only been generally accepted by those in the field, not by disinterested experts in other fields). And those within the field have not challenged or questioned its basic premises. More is required to meet the “general acceptance” factor. Lewis, 220 F.Supp.2d at 554 (noting that general acceptance in the forensic community was insufficient to satisfy the fifth Daubert factor); Starzecpyzel, 880 F.Supp. at 1038 (discounting general acceptance among the community of forensic document examiners because it is “devoid of financially disinterested parties”).

The government did not show that there are factors beyond the Daubert list that credibly demonstrate the reliability of handwriting evidence. Like fingerprint experts, document examiners' have long been allowed to testify in judicial proceedings. Saelee, 162 F.Supp.2d at 1104-05 (“Testimony from these experts has, until recently, been uncritically accepted as reliable in the courts.”). But, like the case of fingerprint evidence, there is no reason to believe that longstanding use of handwriting evidence demonstrates its reliability. The testimony of handwriting experts was initially admitted into evidence because courts saw it as no less reliable than that of lay witnesses who claimed to be able to identify the writers of documents. Mnoo-kin, Scripting Expertise, supra at 1763-64, 1784; D. Michael Risinger et al., Exorcism of Ignorance as a Proxy for Rational Knowledge: The Lessons of Handwriting Identification “Expertise”, 137 U. Pa. L.Rev. 731, 762 (1989). But that does not make handwriting analysis a reliable science.

Because the government has failed to demonstrate either that its handwriting evidence satisfies the Daubert factors or that it is other-wise reliable, I would reverse the district court’s decision to admit it as an abuse of discretion. See Starzecpyzel, 880 F.Supp. at 1028 (“The Daubert hearing established that forensic document examination, which clothes itself with the trappings of science, does not rest on carefully articulated postulates, does not employ rigorous methodology, and has not *282convincingly documented the accuracy of its determinations.”).

IV.

Because the government failed to show that its fingerprint and handwriting evidence meets Daubert’s requirements or is otherwise reliable, the evidence should have been excluded. The government conceded at oral argument that this evidence was necessary to prove Crisp’s guilt beyond a reasonable doubt. Because the evidence was inadmissible, I would reverse Crisp’s conviction. I must therefore respectfully dissent.