Revisiting Dreyfus: A More Complete Account of a Trial by Mathematics

Legal literature and case law depicts the infamous conviction of Alfred Dreyfus for treason and espionage in 1894 as a prime example of the irresistible power of even grossly fallacious mathematical demonstrations to overwhelm a legal tribunal. This essay shows that Dreyfus is not a case of mathematics run amok, unchecked and uncomprehended. To the contrary, the defects in the mathematical proof were dramatically exposed, and this evidence did not lead Dreyfus's judges to condemn him. This more accurate understanding of the case undercuts the reliance of modern courts and commentators on Dreyfus as an indication or illustration of the alleged dangers of probability evidence in criminal cases. A Trial by Mathematics Page 2 1. By "probability evidence," I mean quantitative expressions, derived with the aid of the mathematical theory of probability, of the chances that certain events would occur. Such evidence or argument is not relevant in itself, but is supposed to assist the judge or jury in evaluating the probative force of facts established by other testimony, such as an apparent match in handwriting characteristics, in blood types, in hair fibers, and so on. For surveys of leading cases of "probability evidence" and efforts to extract useful principles from them, see 1 MCCORMICK ON EVIDENCE § 209 (John Strong ed., 5th ed. 1999); D.H. Kaye, The Admissibility of "Probability Evidence" in Criminal Trials)Part I, 26 JURIMETRICS J. 343 (1986); D.H. Kaye, The Admissibility of "Probability Evidence" in Criminal Trials)Part II, 27 JURIMETRICS J. 160 (1987). A deeper analysis can be found in D.H. KAYE ET AL., THE NEW WIGMORE, A TREATISE ON EVIDENCE: EXPERT EVIDENCE § 12 (2004). The most recent, notorious abuse of criminal probability evidence occurred in the prosecution of Sally Clark for allegedly murdering her two children. See, e.g., Kevin Barraclough, Book Review, 329 BRIT. MED. J. 177 (2004) (reviewing JOHN BATT, STOLEN INNOCENCE: A MOTHER'S FIGHT FOR JUSTICE—THE AUTHORISED STORY OF SALLY CLARK (2004)); Stephen J. Watkins, Editorial, Conviction by Mathematical Error? Doctors and Lawyers Should Get Probability Theory Right, 320 BRIT. MED. J. 2 (2000). 2. State v. Schwartz, 447 N.W. 2d 422 (Minn. 1989); State v. Kim, 398 N.W. 2d 544 (Minn. 1987); State v. Boyd, 331 N.W. 2d 480 (Minn. 1983); State v. Carlson, 267 N.W. 2d 170 (Minn. 1978). But see State v. Bloom, 516 N.W.2d 159 (1994) (creating an exception to the Minnesota rule prohibiting testimony as to some probabilities for DNA evidence). 3. See D.H. Kaye, The Probability of an Ultimate Issue: The Strange Cases of Paternity Testing, 75 IOWA L. REV. 75 (1989). 4. See Kaye, Probability Evidence)Part II, supra note 1. The courts have had a hard time with "probability evidence." A few have 1 tried to expel nearly all quantitative assessments of evidence. Others have 2 propounded complex and arbitrary rules of admissibility. Still others have 3 uncritically, and perhaps uncomprehendingly, accepted such assessments. This essay 4 examines one notorious case that has been said to fall in the last category. In a brilliant and influential article published over thirty years ago, Professor Laurence Tribe presented the case as a prime example of the irresistible power of even grossly A Trial by Mathematics Page 3 5. Laurence Tribe, Trial by Mathematics: Precision and Ritual in the Legal Process, 84 HARV. L. REV. 1329 (1971). 6. Hull v. United States, 404 U.S. 893, 895–96 n.3 (1971) (Douglas, J., dissenting from denial of certiorari). Oddly, Hull involved little or no quantified evidence. Justice Douglas would have reversed a smuggling conviction because the government lacked direct evidence that defendants, who were walking on a highway, close to two knapsacks of marijuana, and three-quarters of a mile from "possibly the hottest spot on the Mexican border for smuggling," had crossed the border with the marijuana. Id. at 894. He reasoned that because the agent's testimony about the "hot spot" was based on "anecdotal experiences in four prior investigations," it amounted to "statistical evidence" that "[c]ourts have been hesitant to admit . . . because of the ease with which it can be abused." Id. at 895–96. 7. THE EVOLVING ROLE OF STATISTICAL ASSESSMENTS AS EVIDENCE IN THE COURTS 215 (S. Fienberg ed. 1989). For a more complete description of the panel's work, see D.H. Kaye, Improving Legal Statistics, 24 LAW & SOC'Y REV. 1255 (1990). For reviews of psychological research questioning the alleged tendency of jurors to overvalue probability evidence, see D.H. Kaye & J.J. Koehler, Can Jurors Understand Probabilistic Evidence? 154(A) J. ROYAL STAT. SOC'Y 75 (1991); William Thompson, Are Juries Competent to Evaluate Statistical Evidence? 52 LAW & CONTEMP. PROB. 9 (1989). 8. See State v. Alt, 505 N.W.2d 72, 72 (Minn. 1993); State v. Johnson, 498 N.W.2d 10 (Minn. 1993); State v. Jobe, 486 N.W.2d 407 (Minn. 1992); State v. Nielsen, 467 N.W.2d 615 (Minn. 1991); State v. Schwartz, 447 N.W.2d 422 (Minn. 1989). fallacious mathematical demonstrations to overwhelm a legal tribunal. Not long 5 afterward, Mr. Justice Douglas reiterated this view of the extraordinary power of mathematical evidence to confound court and counsel. A distinguished national 6 panel also described the case to show "the ability of mathematical evidence to paralyze critical examination." And, as the Minnesota Supreme Court attempted for 7 the sixth time to fashion reasonable rules for the admission of expert testimony about DNA evidence, one Justice insisted that "the infamous Dreyfus case" proved that if 8 "an erroneous statistical probability plays any significant role in the conviction of an A Trial by Mathematics Page 4 9. State v. Bloom, 516 N.W.2d 159 (1994) (Coyne, J. dissenting). The majority of the court held that an expert may testify to the frequency of occurrence of a DNA profile when this frequency is computed according to a "conservative" procedure devised by a committee of the National Academy of Science. Not long after that, a second Academy committee wrote that this “ceiling” procedure was not necessary. In its view, more extreme probabilities could be justified scientifically. COMMITTEE ON DNA TECHNOLOGY IN FORENSIC SCIENCE: AN UPDATE, NATIONAL RESEARCH COUNCIL, THE EVALUATION OF FORENSIC DNA EVIDENCE (1996). 10. Histories of the Dreyfus affair are plentiful and opinionated. A detailed account by a Parisian lawyer and professor is JEAN-DENIS BREDIN, THE AFFAIR: THE CASE OF ALFRED DREYFUS (J. Mehlman transl. 1986), originally published as L'AFFAIRE (1983). A shorter essay, which focuses on the legal proceedings, is Martin, The Dreyfus Affair and the Corruption of the French Legal System, in THE DREYFUS AFFAIR: ART, TRUTH AND JUSTICE 37 (Norman L. Kleeblatt ed., 1987). innocent person, the error has not only destroyed the life of the innocent person but has in some sense dehumanized the community." The case in question is the court-martial of Captain Alfred Dreyfus of the French Army at the turn of the Nineteenth Century. It is a case of such injustice that 10 it toppled a government. But despite the legend in the literature on legal statistics, Dreyfus is not a case of mathematics run amok, unchecked and uncomprehended. To the contrary, the defects in the mathematical proof were dramatically exposed, and this evidence did not lead Dreyfus's judges to condemn him. Accordingly, Dreyfus’s contrived conviction, as intolerable as it was, does not militate against the admission of "probability evidence." The Dreyfus Cases: An Overview Dreyfus is not a single case, but rather a series of connected military, civil and criminal proceedings. They began in 1894 with a court-martial that convicted Dreyfus of transmitting military secrets to Germany and sentenced him to life imprisonment on Devil's Island. The verdict, which was tainted by a secret dossier, fabricated evidence and widespread anti-Semitism in the French army and populace, became an international cause célèbre. As evidence of Dreyfus's innocence began to mount and as it became known that French military authorities had manufactured A Trial by Mathematics Page 5 11. Martin, supra note 10, at 47, argues that this judgment of acquittal (as opposed to remanding the case to the army) was procedurally improper but political expedient. 12. The actual author was a French officer named Esterhazy who was providing information to the German military attaché. additional evidence to keep the case from being reopened, the army and the government were shaken. The discovery of one forgery, purporting to be a letter from an Italian military attaché, prompted the suicide of the colonel working in military intelligence who had prepared it and produced the resignations of the chief of the Army's General Staff and the Minister of War. France's highest court, the Cour de Cassation, sitting en banc as a result of special legislation, vacated the judgment of the military court. After five years of brutal conditions on Devil's Island, Dreyfus returned to a second court-martial. At Rennes in 1899, this court again found Dreyfus guilty of treason. Issuing a compromise verdict referring to extenuating circumstances (and prompting Dreyfus to ask "Since when are there extenuating circumstances for treason?"), it sentenced Dreyfus, to another five years confinement. The verdict was so poorly received that within two weeks, the president of the republic pardoned Dreyfus. After further political upheavals and a War Office report finding that most of the evidence at Rennes either did not relate to Dreyfus or had been altered to make it appear that it did, the Cour de Cassation granted a petition for review. In 1906, declaring that no credible evidence of treason ever existed, the court annulled the verdict of the Rennes court-martial. Dreyfus, the man twice convicted of treason, 11 returned to the army and was awarded the cross of the Legion of Honor. Bertillon's Analysis of the Bordereau The document that initiated the 1894 prosecution was a letter, known as the bordereau. Retrieved from a wastepaper basket in the German embassy, the bordereau listed several relatively unimportant accompanying documents about French artillery and troops. French intelligence officers decided that Dreyfus was the culprit. (Why not? He was a Jew and an Alsatian.) They collected and contrived evidence to support this thesis and ignored or suppressed all contrary evidence. The 12 A Trial by Mathematics Page 6 13. Bertillon, who had invented a system of identifying individuals from various body measurements, headed an office of criminal identification. 14. Tribe, supra note 5, at 1332. 15. Mary Gray, Statistics and the Law, 56 MATHEMATICS MAG. 67, 68 (1983), and Mode, Probability and Criminalistics, 58 J. AM . STAT. ASS'N 628 (1963), rely on a 1928 account, The Reality of Mathematical Processes, by Hedrick, in THE NATIONAL COUNCIL OF TEACHERS OF MATHEMATICS THIRD YEARBOOK 40-41, that refers to the frequency of the letters of the French alphabet within the bordereau as contrasted to the proportions found in typical French prose as constituting the statistical evidence against Dreyfus. I found no references to such testimony in the documents collected in LA REVISION DU PROCÈS DE RENNES (1909). 16. Tribe, supra note 5, at 1333 n.8. resulting dossier was enough to convince the French government to convene the 1894 court-martial. Handwriting experts contacted by the army and the Ministry of Justice who studied the bordereau came to conflicting conclusions. The most notorious analysis came from the anthropologist Alphonse Bertillon, who claimed that Dreyfus wrote 13 it in a way that would make it look like a forgery of his own handwriting. Bertillon advanced this "self-forgery" theory at both the military trials and at the criminal libel trial of the novelist, Emile Zola, for his vitriolic public letter, J'accuse, which denounced the army for "one of the greatest iniquities of the century" in its handling of the case. To Bertillon, the proof of "self-forgery" was scientific, incontestable and infallible)in a word, geométrique. This proof included computations of the probabilities of selected coincidences between the lengths of certain words and letters in this document and the lengths of certain words and letters in correspondence taken from Dreyfus's home. Furthermore, from obscure lexicographical and graphological 14 coincidences within the document itself, Bertillon divined that the letter contained coded information. For example, he stressed the presence of four coincidences out 15 of the 26 initial and final letters of the 13 repeated polysyllabic words in the bordereau. He evaluated at .2 the probability of an isolated coincidence and calculated a probability of .2 = .0016 that four such coincidences would occur in 4 normal writing. Likewise, to establish that the letters had been traced over the word 16 A Trial by Mathematics Page 7