Promoters of forensic DNA testing have, from the beginning, claimed that DNA tests are virtually infallible.1,2 In advertising materials, publications and courtroom testimony, the claim has been made that DNA tests produce either the right result or no result.3 This rhetoric of infallibility took hold early in appellate court opinions, which often parroted promotional hyperbole.4 It was supported when the National Research Council, in the second of two reports on forensic DNA testing, declared "the reliability and validity of properly collected and analyzed DNA data should not be in doubt."5 It was further reinforced in the public imagination by news accounts of post-conviction DNA exonerations. Wrongfully convicted people were shown being released from prison, while guilty people were brought to justice, by this marvelous new technology. With both prosecutors and advocates for the wrongfully convicted using it successfully in court, who could doubt that DNA evidence was in fact what its promoters claimed: the gold standard, a truth machine?6
The rhetoric of infallibility proved helpful in establishing the admissibility of forensic DNA tests and persuading judges and jurors of its epistemic authority.7 It has also played an important role in the promotion of government DNA databases. Innocent people have nothing to fear from databases, promoters claim. Because the tests are infallible, the risk of a false incrimination must necessarily be nil. One indication of the success and influence of the rhetoric of infallibility is that, until quite recently, concerns about false incriminations played almost no role in debates about database expansion. The infallibility of DNA tests has, for most purposes, become an accepted fact-one of the shared assumptions underlying the policy debate.
In this article, I will argue that this shared assumption is wrong. Although generally quite reliable (particularly in comparison with other forms of evidence often used in criminal trials), DNA tests are not now and have never been infallible. Errors in DNA testing occur regularly. DNA evidence has caused false incriminations and false convictions, and will continue to do so. Although DNA tests incriminate the correct person in the great majority of cases, the risk of false incrimination is high enough to deserve serious consideration in debates about expansion of DNA databases. The risk of false incrimination is borne primarily by individuals whose profiles are included in government databases (and perhaps by their relatives). Because there are racial, ethnic and class disparities in the composition of databases, the risk of false incrimination will fall disproportionately on members of the included groups.8,9
This article will discuss major ways in which false incriminations can occur in forensic DNA testing, including coincidental DNA profile matches between different people, inadvertent or accidental transfer of cellular material or DNA from one item to another, errors in identification or labeling of samples, misinterpretation of test results, and intentional planting of biological evidence. It will also discuss ways in which the secrecy that currently surrounds the content and operation of government databases makes these issues difficult to study and assess. It will conclude by calling for greater openness and transparency of governmental operations in this domain and a public program of research that will allow the risks discussed here to be better understood.A coincidental match between different people who happen to share the same DNA profile is one way a false incrimination can occur. To understand the likelihood of a coincidental match, it is important to understand what a DNA profile is and how DNA profiles are compared. Forensic laboratories typically "type" samples using commercial test kits that can detect genetic characteristics (called alleles) at various loci (locations) on the human genome. The test kits used in the United States generally examine the 13 STR loci selected by the FBI for CODIS, the national DNA database.10 Some of the newer test kits also examine two additional STR loci.
At each STR locus, there are a number of different alleles (generally between 6 and 18) that a person might have. Each person inherits two of these alleles, one from each parent. Numbers are used to identify the alleles and the pair of alleles at a particular locus constitutes a genotype. Hence, one person can have a genotype (for a locus called D3S1358) of "14, 15;" while another person has the genotype "16, 17." The complete set of alleles detected at all loci for a given sample is called a DNA profile. When describing DNA profiles, people sometimes mention the number of loci they encompass.
In cases I have reviewed over the past few years, evidentiary samples from crime scenes often produce incomplete or partial DNA profiles. Limited quantities of DNA, degradation of the sample, or the presence of inhibitors (contaminants) can make it impossible to determine the genotype at every locus. In some instances the test yields no information about the genotype at a particular locus; in some instances one of the two alleles at a locus will "drop out" (become undetectable). Because partial profiles contain fewer genetic markers (alleles) than complete profiles, they are more likely to match someone by chance (see endnote 1). The probability of a coincidental match is higher for a partial profile than for a full profile.
A further complication is that evidentiary samples are often mixtures. Because it can be difficult to tell which alleles are associated with which contributor in a mixed sample, there often are many different profiles (not just one) that could be consistent with a mixed sample. Because so many different profiles may be consistent with a mixture, the probability that a non-contributor might, by coincidence, be "included" as a possible contributor to the mixture is far higher in a mixture case than a case with a single-source evidentiary sample.
The risk of obtaining a match by coincidence is far higher when authorities search through thousands or millions of profiles looking for a match than when they compare the evidentiary profile to the profile of a single individual who has been identified as a suspect for other reasons. As an illustration, suppose that a partial DNA profile from a crime scene occurs with a frequency of 1 in 10 million in the general population. If this profile is compared to a single innocent suspect, the probability of a coincidental match is only 1 in 10 million. Consequently, if one finds such a match in a single-suspect case it seems safe to assume the match was no coincidence. By contrast, when searching through a database as large as the FBI's National DNA Index System, which reportedly contains nearly 6 million profiles, there are literally millions of opportunities to find a match by coincidence. Even if everyone in the database is innocent, there is a substantial probability that one (or more) will have the 1-in-10 million profile. Hence, a match obtained in a database search might very well be coincidental. Consider that among the 6 billion or so people on planet earth we would expect about 600 to have the one-in-10-million DNA profile; among the 300 million or so in the United States we would expect to find about 30 people with the profile. How certain can we be that the one matching profile identified in a database search is really that of the person who committed the crime?
A number of states have recently begun conducting what is known as familial searches.11 In cases where a database search finds no exact match to an evidentiary profile but finds a near match - that is, a profile that shares a large number of alleles but is not identical - authorities seek DNA samples from relatives of the person who nearly matches in the hope that one of the relatives will be an exact match to the evidentiary sample. In several high-profile cases familial searches have identified suspects who were successfully prosecuted.12 The key questions raised by familial searches, from a civil liberties perspective, are how often they lead to testing of innocent people-i.e., people who do not have the matching profile-and how often they might falsely incriminate innocent people through coincidental matches. Familial searching may increase the number of people falsely incriminated by coincidental matches because it increases the effective size of the population subject to genetic monitoring. The larger the effective size of the database, the greater will be the likelihood that one of those innocent people will be identified.
People have been prosecuted based on cold hits to partial profiles. Defendants in cold-hit cases often face a difficult dilemma. In order to explain to the jury that the incriminating DNA match arose from a database search (in which the government had thousands or millions of opportunities to find a matching profile), the defendant must admit that his profile was in the database, which in many states entails admitting to being a felon, a fact that might otherwise be inadmissible. Courts in some cold-hit cases have, at the urging of defense counsel, opted to leave the jury in the dark about the database search in order to avoid the implication of a criminal record. Jurors are told about the DNA match, but are not told how the match was discovered. The danger of this strategy is that jurors may underestimate the probability of a false incrimination because they assume the authorities must have had good reason to test the defendant's DNA in the first place. In other words, jurors may mistakenly assume the DNA test compared the crime scene sample to the DNA of a single individual who was already the focus of suspicion (a circumstance under which the risk of a coincidental false incrimination is extremely low) and not realize that the defendant was identified through a cold hit (a circumstance under which the risk of a coincidental false incrimination is much higher).
My argument is that jurors'evaluations of the case as a whole may be inaccurate if they are not told the match was found through a database search. I am suggesting that jurors will assume (incorrectly) that the DNA evidence confirms other evidence that made the defendant the subject of police suspicions and hence will underestimate the likelihood that the defendant could have been incriminated by coincidence. This is a process that, in my view, puts innocent people who happen to be included in a database at risk of false conviction.
When DNA evidence was first introduced, a number of experts testified that false positives are impossible in forensic DNA testing. Whether such claims are sinister or not, they are misleading because humans are necessarily involved in conducting DNA tests. Among the first 200 people exonerated by post-conviction DNA testing were two men (Timothy Durham and Josiah Sutton) who were convicted in the first place due partly to DNA testing errors. In both cases a combination of technical problems in the laboratory and careless or mistaken interpretation of the test results produced misleading DNA evidence that helped send innocent men to prison for many years.13 False DNA matches have come to light in a number of other cases as well.14,15
One cause of false DNA matches is cross-contamination of samples. Accidental transfer of cellular material or DNA from one sample to another is a common problem in laboratories and it can lead to false reports of a DNA match between samples that originated from different people. In addition, accidental cross-contamination of DNA samples has caused a number of false "cold hits."
A second potential cause of false DNA matches is mislabeling of samples. The best way to detect labeling errors is to obtain new samples from the original sources and retest them, but this safeguard is not always available. Evidence at crime scenes is typically cleaned up (and thereby destroyed) once samples are taken, and the original samples are sometimes exhausted during the initial round of testing. Retesting is rarely done, even when samples are available. Routine duplicate testing by forensic laboratories is another possible safeguard, but it too is rarely done.
A third potential cause of false DNA matches is misinterpretation of test results. Laboratories sometimes mistype (i.e., assign an incorrect STR profile to) evidentiary samples. If the incorrect evidentiary profile happens to match the profile of an innocent person, then a false incrimination may result. Mistyping is unlikely to produce a false match in cases where the evidentiary profile is compared with a single suspect, but the chance of finding a matching person is magnified (or, more accurately, multiplied) when the evidentiary profile is searched against a database.
The ability of criminals to neutralize or evade crime control technologies has been a persistent theme in the history of crime.16,17 There are anecdotal reports of criminals trying to throw investigators off the track by planting biological evidence. When such planting occurs, will the police be able to figure it out? Will a jury believe the defendant could be innocent once a damning DNA match is found? I have strong doubts on both counts and, consequently, believe that intentional planting of DNA evidence may create a significant risk of false incriminations.
Do innocent people really have nothing to fear from inclusion in government DNA databases? It should now be clear to readers that this claim is overstated. If your profile is in a DNA database you face higher risk than other citizens of being falsely linked to a crime. You are at higher risk of false incriminations by coincidental DNA matches, by laboratory error, and by intentional planting of DNA. There can be no doubt that database inclusion increases these risks, the only real question is how much. In order to assess these risks, and weigh them against the benefits of database expansion, we need more information.
Some of the most important information for risk assessment is hidden from public view under a shroud of governmental secrecy. For example, the government's refusal to allow independent experts to examine the (de-identified) DNA profiles in offender databases is a substantial factor in continuing uncertainty about the accuracy of frequency estimates (and hence the probability of coincidental matches). I believe there is no persuasive justification for the government's insistence on maintaining the secrecy of database profiles, so long as the identity of the contributors is not disclosed. The government's refusal to open those profiles to independent scientific study is a significant civil liberties issue.
William C. Thompson is Professor and Chair of the Department of Criminology, Law & Society at the University of California, Irvine. He co-chairs the Forensic Evidence Committee of the National Association of Criminal Defense Lawyers and is a member of the California Crime Laboratory Task Force, a body created by the state legislature to recommend ways of improving forensic science in California. He studies the way people interpret (and sometimes misinterpret) scientific and statistical data and has also written extensively about the use and misuse of DNA evidence.
1. In general, as the number of alleles in a DNA profile decreases, the probability that a randomly chosen person will, by coincidence, happen to match that profile increases. Because the alleles vary greatly in their rarity, however, it is possible for a profile containing a few rare alleles to be rarer overall that a profile containing a larger number of more common alleles. Consequently, when discussing the likelihood of a coincidental match it is more helpful to focus on the estimated frequency of the profile than the number of loci or alleles encompassed in the profile.
1. J.J. Koehler, "Error and exaggeration in the presentation of DNA evidence," Jurimetrics, 34: 21-39, 1993.
2. W.C. Thompson, "Forensic DNA Evidence," In B. Black & P. Lee (Eds.), ExpertEvidence: A Practitioner's Guide to Law, Science and the FJC Manual. St. Paul, Minn.: West Group, 1997 pp. 195-266.
3. Jay D. Aronson, Genetic Witness: Science, Law and Controversy in the Making of DNA Profiling. New Brunswick, N.J.: Rutgers University Press, 2007.
5. National Research Council, The Evaluation of Forensic DNA Evidence. Washington, D.C.: National Academy Press, 1996, p. 2.
6. Michael Lynch, Simon Cole, Ruth McNally & Kathleen Jordan, Truth Machine: The Contentious History of DNA Fingerprinting. Chicago: University of Chicago Press (2008).
8. Simon A. Cole, "How much justice can technology afford? The impact of DNA technology on equal criminal justice." Science and Public Policy, 34(2) 95-107, March 2007; Simon A. Cole, "Double Helix Jeopardy," IEEE Spectrum, 44-49, August 2007
9. Harry G. Levine, Jon Gettman, craig Reinarman & Deborah P. Small, "Drug arrests and DNA: Building Jim Crow's Database." Paper produced for the Council forResponsible Genetics (CRG) and its national conference, Forensic DNA Databases and Race: Issues, Abuses and Actions held June 19 20,2008, at New York University. Available at www.gene-watch.org.
10. John M. Butler, Forensic DNA Typing: Biology, Technology and Genetics of STR Markers (2nd Ed.). Elsevier/Academic Press, 2005.24
11. Richard Willing, "Suspects get snared by a relative's DNA," USA Today, June 8, 2005, at 1A; David R. Paoletti, Travis E. Doom, Michael L. Raymer & Dan Krane, "Assessing the implications for close relatives in the event of similar but no matching DNA profiles," Jurimetrics Journal 46: 161-175 (2006).
12. Willing 2005.
13. W.C. Thompson, "Beyond bad apples: Analyzing the role of forensic science in wrongful convictions." Southwestern Law Review 37:101-124 (forthcoming).
14. W.C. Thompson, F. Taroni & C.G.G. Atiken, "How the probability of a false positive affects the value of DNA evidence." Journal of Forensic Sciences, 48(1): 47-54 (2003).
15. W.C. Thompson, "Tarnish on the 'gold standard:' Understanding recent problems in forensic DNA testing." The Champion, 30(1): 10-16 (January 2006).
16. Paul Ekblom, "Can we make crime prevention adaptive by learning from other ecological struggles?" Studies on Crime and Crime Prevention. 8: 27-51, 1998.
17. Paul Ekblom, "How to police the future: Scanning for scientific and technological innovations which generate potential threats and opportunities in crime, policing and crime reduction," In. M. Smith and N. Tilley (Eds.) Crime Science: New Approaches to Preventing and Detecting Crime. Cullompton: Willan, 2005.