James S. Liebman, Shawn Blackburn, David Mattern and Jonathan Waisnor
Abstracted from: James S. Liebman, Shawn Blackburn, David Mattern and Jonathan Waisnor, The Evidence of Things Not Seen : Non-matches as Evidence of Innocence, 98 Iowa Law Review 577 (January, 2013) (534 Footnotes)
ABSTRACT
In 1998 in State v. Hayes, a New Orleans jury convicted juveniles Ryan Matthews and Travis Hayes of the robbery-murder of a New Orleans store owner. There was considerable evidence against the two. Witnesses reported that a black male shot the robbery victim, then leapt into a getaway car through an open passenger-side window and escaped. Police stopped Hayes and Matthews in a vehicle resembling witnesses' description of the getaway car. Hayes confessed to the police that he drove the getaway car, and that Matthews, who was high on marijuana, entered the store and ran out after several gunshots. A witness identified Matthews as the man he saw running from the store.
Although strong, the evidence was deficient in some respects. Hayes was borderline mentally retarded and confessed after six hours of interrogation. His statements were modestly inconsistent with each other and with known details of the crime. The make and model of the car Hayes was driving when the two were arrested were not the same as witnesses reported, and the car had a rolled up passenger-side window that had been inoperable for as long as anyone could remember. The eyewitness who identified Matthews watched the events through a rear-view mirror, and other eyewitnesses could not identify Matthews and described the shooter as 5?6?--six inches shorter than Matthews. DNA on a ski mask left at the crime scene matched neither defendant. None of the deficiencies excluded the possibility of guilt, however, and the jury convicted both men. Hayes was sentenced to life in prison and Matthews to death.
Shortly after the trial, Rondell Love was convicted of an unrelated manslaughter and sentenced to twenty years in the Louisiana State Penitentiary, where Matthews was awaiting execution. Love told other inmates that he had committed the murder for which Matthews was condemned. Matthews got wind of Love's claims and reported them to his lawyers who obtained DNA testing of the ski mask at the scene. The results implicated Love, who at 5?7? and medium build matched the witnesses' descriptions. Matthews and Hayes were eventually exonerated.
In 1983 in State v. DeLuna, a jury convicted Carlos DeLuna of stabbing a young female Hispanic clerk to death at a Diamond-Shamrock gas station in Corpus Christi, Texas. The main evidence at the trial was a night-time show-up identification of DeLuna by the sole eyewitness to the single-perpetrator crime, who had seen the assailant escape on foot. Again, there were discrepancies, including the eyewitness's initial description of the shabbily dressed, mustachioed, and bewhiskered “derelict” he saw struggling with the victim; fingerprints found at the scene; and the blood-soaked scene itself, none of which matched the blood-free white dress shirt, dress pants, and shoes worn by the clean-shaven DeLuna when he was arrested shortly after the killing. The crime was captured on a 911 phone call from the store clerk. Police arrested DeLuna three blocks away, cowering under a pick-up truck. DeLuna testified that he had seen an acquaintance named Carlos Hernandez wrestling with the clerk inside the store and fled when he heard sirens coming because he had been drinking in violation of his parole conditions. A police officer testified that he scoured police records for a “Carlos Hernandez” matching the eyewitness description of the assailant but found none. The prosecutor dubbed Hernandez a “phantom” and DeLuna a “liar” for fabricating him. During DeLuna's appeals, an affirming court expressed “substantial doubt that Carlos Hernandez even DeLuna was quietly executed in 1989.
Seventeen years later, the senior author of this Article, and later the Chicago Tribune, reinvestigated the case and discovered that Carlos Hernandez existed and was well-known to Corpus Christi law enforcement. Three years before the convenience store stabbing, the detective and assistant district attorney handling that case had considered Hernandez a prime suspect in the beating and knifing death of another young Hispanic woman. Although the detective and prosecutor eventually arrested and tried a different man, he was acquitted after the man's attorney marshaled evidence that Hernandez was the culprit. Like DeLuna, Hernandez was 5?8? tall and weighed 160 pounds, as was the man the eyewitness described to police immediately after the crime. In the subsequent reinvestigation, relatives of both Carloses mistook one for the other when shown pictures of the two taken within weeks of the gas station killing.
Shortly after the killing, and for years afterwards, Hernandez told associates that he, not DeLuna, committed the crime. Eight months before DeLuna was executed, Hernandez stabbed another young Hispanic woman nearly to death with a lock-blade buck knife identical to the one found at the Diamond-Shamrock crime scene. Hernandez had previously confessed the Diamond-Shamrock stabbing to this woman, and in the midst of attacking her, told her she was going to suffer the same fate because she insisted on dating another man.
Hernandez's characteristic modes of dress and grooming at the time of the convenience store murder, and many other traits that did not match DeLuna, did match the eyewitness's initial description of the assailant.
Figure 1
TABULAR OR GRAPHIC MATERIAL SET FORTH AT THIS POINT IS NOT DISPLAYABLE
Carlos DeLuna (left) two weeks before the killing for which he was executed. Carlos Hernandez (right) two months after that killing, upon his arrest with a knife behind another convenience store. Hernandez wore a moustache his entire adult life, except in the weeks after the Diamond-Shamrock killing.
When considering what went wrong in cases like Matthews/Hayes and DeLuna, commentators typically focus on the mishandling of what might be called “big” evidence of identity--evidence that by itself is likely to impress a jury, such as Hayes's confession and the show-up identification of DeLuna. These observers point out that identifications, confessions, and “snitch” testimony (another classic example of influential identity evidence) are less reliable than jurors think. To cut such evidence down to size, these critics advocate expert testimony about the foibles of eyewitness identifications and confessions, cautionary instructions about informant testimony, or exclusion of evidence unless it was collected through state-of-the-art techniques such as double-blind lineups and videotaped confessions.
Exoneration cases such as Hayes/Matthews and reinvestigation cases such as DeLuna reveal a second type of indicative evidence, which commentators typically ignore. In almost all of these cases, police arrested, prosecutors tried, and jurors convicted the defendant despite multiple “non-exclusionary non-matches” between the defendant and potentially evidential traces from the crime scene that later were matched to the “real killer.” “Non-exclusionary non-matches” arise when a suspect or defendant demonstrably was not the source of bits of potential evidence associated with a crime that might have been left by the perpetrator but might also have appeared for reasons having nothing to do with the crime. The police, prosecutors, and jurors in these cases no doubt acted against the defendant, despite the non-matches, precisely because the non-matches were “non-exclusionary”: they had explanations consistent with the defendant's guilt, and thus did not exclude him as the guilty party or prove his innocence.
Examples include the non-match in the Hayes/Matthews case between the make and model of the car witnesses linked to the killing and the car Hayes was driving, between the open passenger-side window of the car through which the robber was said to have jumped and the fact that the passenger window on Hayes's car had been stuck in a rolled-up position for years, and between the 5?6? assailant witnesses described and Matthews 6?1? frame. Examples from the DeLuna case are collected in Figure 2. Of twenty-nine traces associated with the crime, three match neither DeLuna (executed for the crime) nor Hernandez (who told associates he committed the crime). Seven more were never tested as to either man due to police inaction--for example, large amounts of blood found at the crime scene and evidence visible in the investigating officers' photographs that they never noticed, including bloody shoe prints and a wad of chewed gum spat onto the floor. Only seven of the twenty-nine traces match DeLuna, all seven of which also match Hernandez: height, weight, ethnicity, sex, hair color, hair style, and cigarette brand. Of the twelve remaining non-matches as to DeLuna, two were never tested as to Hernandez (now deceased), and the remaining ten match Hernandez--including age, clothing, moustache, and weapon of choice. Although the jury convicted DeLuna based in part on the seven matches, it never knew that seventeen or more traits matched Hernandez, including all seven that matched DeLuna.
Figure 2
TABULAR OR GRAPHIC MATERIAL SET FORTH AT THIS POINT IS NOT DISPLAYABLE
It is not surprising that police, prosecutors, and jurors in the Hayes/Matthews and DeLuna cases dismissed these “small” non-matches--“small” because they were easily explained away on grounds unrelated to the defendant's guilt or innocence--and focused instead on confessions, eyewitness identifications, and other “big” evidence of identity. No one begrudges a prospector for gold who ignores tiny flecks of the stuff in a stream bed while seeking the mother lode up the canyon side. Compared to the tedious task of accumulating enough flecks to add up to a real stake, the prospect of striking it rich with the discovery of a large single vein is hard to pass up. This Article argues, however, that systematically aggregating more of the small flecks and putting less faith in the deceptive allure of big lodes of evidence could generate more evidence of identity and a decrease in wrongful conviction.
In Part II, we use both hypothetical and actual examples to illustrate how systematically aggregative analysis of non-matching bits of evidence might improve the accuracy of answers police, prosecutors, and jurors reach on the “Whodunit?” question. The English courts' rejection of aggregative analysis in the actual case discussed in Part II prompts our examination in Part III of the reasons courts give for resisting efforts to quantify the conjoint effect of small bits of non-dispositive evidence in resolving the identity question. The classic case is People v. Collins, which overturned a California robbery conviction premised in part on a prosecutor's effort to quantify the overall effect of several individually non-dispositive matches between the two defendants in the case and what was known about the interracial couple that committed the robbery. Although decided over forty years ago and easily distinguished based on the prosecutor's flawed methodology and barely concealed racial ulterior motive, the case continues today to provide part of the intellectual basis for resisting probabilistic proof generally.
Part IV responds to the Collins critique by showing that the courts themselves have rejected the critique of probabilistic evidence in actual practice through their enthusiastic embrace of inculpatory DNA evidence. With DNA evidence as our prime example, and confession, eyewitness identification, and fingerprint evidence as supporting cases, we elucidate the often acknowledged fact that all evidence is probabilistic with a demonstration of a less obvious fact: the high probabilities associated with these and other supposedly “unique” traits matching suspect to culprit are in fact the aggregate result of multiple matches of non-unique, often very common traits. Other things being equal, the more non-unique matching traits subsumed by any given DNA sample, confession, eyewitness identification, fingerprint, or other piece of “big evidence,” the more probative weight the evidence has. DNA evidence reveals, as well, that there is no reason in principle to shield even lay jurors from the formal quantification of the aggregate effect of non-unique bits of evidence into a stated probability that the defendant was responsible for the existence of evidence associated with a crime. DNA thus gives the lie to Collins's most far-reaching claim--that formal analysis of the aggregate effect of non-unique matching traits is anathema to accepted modes of judicial fact finding. As Part IV also develops, the adversary system's swift and dramatic improvement of forensic DNA analysis, along with the burgeoning capacity of data-mining techniques to reveal the frequency of millions of traits in relevant populations and environments, neutralize Collins's lesser, more technical reasons for rejecting formal aggregative analysis in criminal trials. What is true for trials is even truer for investigations and prosecutorial decision making, where our commitments to adversarial judicial proceedings and jury decision making are not limiting factors.
There are, however, other obstacles to using aggregative analysis of matches and non-matches in criminal investigations, trials, and appeals. Part V addresses three categories of barriers: cognitive, structural, and legal. Focusing mainly on the use of non-exclusionary non-matches to decrease the likelihood of arresting, charging, and convicting the innocent, Part V shows how heuristic economization interacts with the structure of trials to reinforce legal resistance to the aggregative use of “small” evidence and discourage use of the adversarial system to discipline and improve aggregative analysis. In Part VI, we chart two intersecting paths around these obstacles. One is provided by emerging tools designed to make aggregative analysis more intuitively accessible to investigators and jurors. The other is the use of management-based regulation, along with the adversarial system, to discipline systematic steps to pan for small but cumulatively powerful flecks of evidence of identity in criminal investigations and trials.
. Simon H. Rifkind Professor, Columbia Law School.
