The Case of the Bloody Brick In March 2003, a drunk in southern England threw a brick off a bridge late at night, striking and killing a truck driver traveling along the freeway below. Armed with DNA from the blood on the brick, the British police searched the United Kingdom’s national DNA database, which includes convicted felons and people who have been arrested, but failed to get a direct match. They then conducted a DNA dragnet, asking hundreds of young men in the area to donate a sample voluntarily, but still came up short. Without any other leads, the police decided to conduct what’s called a “familial search” of the national DNA database. They were looking not for perfect matches to convicted offenders but for near matches, in the hope of using them to identify a relative who might have committed the crime.
Ordinarily, when searching for actual offenders, the British police look for a perfect match to a DNA profile that contains 10 pairs of peaks, or “alleles,” with one number in each pair provided by the father and the other by the mother. Only identical twins share genetic profiles on all 20 alleles, so if you get a perfect match between the DNA you find at the scene and the DNA database profile, you have very strong evidence that the person in the database committed the crime. But the police can also program the search to look for partial matches, identifying profiles that are similar but not identical to those in the database. A partial match can suggest that the person in the database didn’t commit the crime, but a close relative whose DNA pattern varies slightly on some of the 20 alleles may have done so. Accordingly, the British authorities programmed the search to pull up any offender in the database who matched at least 11 alleles out of 20 from the blood on the brick.
Initially, this familial search produced more matches than the police could follow up. But then the authorities limited the search to young men from two counties near the crime scene. This narrowed the number of partial matches to around 25. After interviewing the person whose profile represented the closest match—16 out of 20 alleles—the police found he had a brother who lived in one of the nearby counties. They went to the brother, Craig Harman, who agreed to give a DNA sample. It turned out to match the DNA on the brick. Harmon confessed after being confronted with the match, and in 2004, he was convicted of manslaughter.
Two years later, the district attorney of Denver, Mitch Morrissey, learned about the trail from the blood on the brick to Craig Harman on a trip to the United Kingdom, where he met with officials from the home office, police, and Forensic Science Service. They praised the success of familial searches. According to the country’s Forensic Science Service, the Brits have done 70 such searches since 2004, leading to 18 matches and 13 convictions. The success rate is estimated at around 10 percent, which seems low but which the police see as worthwhile.
Morrissey returned to the United Sates and pitched the FBI, urging the agency to expand the search capacity of its Combined DNA Index System for “familial searches.” He met resistance from Thomas F. Callaghan, who was then head of the FBI laboratory’s CODIS unit, which encompasses the national DNA database. Callaghan is a former forensic scientist with the Pennsylvania State Police who has a doctorate in molecular biology; he balked at Morrissey’s suggestion. His concern was that if the FBI began doing familial searches without congressional or judicial authorization, there could be a political and legal backlash over privacy and civil liberties that would imperil the federal government’s recent decision to begin storing DNA samples not merely from convicted felons but from anyone who is arrested.
Last March, at an FBI conference about genetic privacy, critics of familial searches made their case to the administrators of the national, state, and local DNA databases. Partly as a result, the FBI decided not to reconfigure the national CODIS software to allow familial searches. Nevertheless, current FBI policy allows individual states to decide whether to proceed with familial searches on their own. As a result, last April, California Attorney General Jerry Brown (who is considering a run for governor) began to allow familial DNA searching in the largest state database in the country. The decision may provoke precisely the legal and political backlash that Callaghan predicted.
The California DNA database now contains approximately 1.2 million convicted people. And it’s about to grow dramatically. In January, Brown announced a “major expansion” of the California database, as the state began to add the DNA of arrestees. California expects this to increase the pool of new DNA profiles from 200,000 to nearly 390,000 a year.
All of this could mean a slew of legal challenges on the horizon—not only over familial searching but also over the decision to include people who have been arrested in DNA databases at all, as 14 states and the federal government have done. In December, the European Court of Human Rights held that Britain’s decision to store the DNA of unconvicted people violates European privacy guarantees—throwing the future of the British database into question.
The legal limits on family searches and DNA databases are murky, but the political implications are explosive for one big reason in particular: race. African-Americans, by several estimates, represent about 13 percent of the U.S. population but 40 percent of the people convicted of felonies every year. The CODIS database of 6.6 million now includes samples from convicted offenders. As arrestees are added to this mix, CODIS may soon grow to 50 million samples, which might be even more disproportionately African-American. Hank Greely of Stanford Law School has estimated that 17 percent of African-American citizens could be identified through familial searches, as opposed to only 4 percent of the Caucasian population. Once the implications of the racial disparity become clear, there may be a reaction against ever-more-expansive forms of DNA collection that makes the debate about racial profiling look tame.
The Dilemma of “CODIS Creep”
Not long ago, I drove to Quantico, Va., to meet with Tom Callaghan, who recently left CODIS, and to tour the CODIS unit, located on a bucolic Marine base in the gleaming new FBI Laboratory on Investigation Parkway. As we sat around a conference table, Callaghan gave me a history of CODIS, which contains national, state, and local DNA databases of convicted offenders, arrestees, crime scenes, and missing persons. The story he told was one of steady expansion. States began to set up their own databases in the early 1990s, and in 1994 Congress authorized the creation of the National DNS Index System. It launched in 1998 and originally included only nine states. Federal offenders were added two years later. In the beginning, only violent felons went into the federal database; that later expanded to all felons and then to all felony arrestees. Today, the national database includes 178 labs, and the CODIS software is distributed to all 50 states and 44 labs in 30 foreign countries.
Flipping through his PowerPoint slides, Callaghan explained to me how CODIS works. Each CODIS profile contains 52 characters representing 13 genetic locations, with two results per location and two digits for each result. Every Monday at 9 a.m., the national database automatically conducts two searches, looking for matches between the DNA of convicted offenders and the DNA at crime scenes. The automatic search also compares all crime-scene DNA samples with one another in search of serial criminals. A request for a specific search—say, from Florida to search the national database for a serial killer at large—goes to the custodian of the national database. As of January, the FBI claims that CODIS searches have “aided” more than 83,000 investigations. In addition, state investigators have provided names for state investigations more than 60,000 times, and states have traded the names of offenders 8,818 times. Still, the feds don’t keep statistics on how many of these “cold hits” have actually led to convictions.
As I watched a mock report of a cold hit flash across the screen, a jarring statistic appeared: The FBI ranks the probability of its matches by using racial categories. In other words, if there is a perfect match of 23 alleles at all 13 genetic locations (the super-sized American version of the British 10 peaks), and an African-American with the targeted profile goes on trial, the jury might be told that those alleles are found in, say, one out of 14 quadrillion African-Americans while being only slightly more (or less) common in Caucasians or southeastern Hispanics. The point of the statistic, according to the FBI, is to reassure jurors who think the police may not have correctly identified a suspect whose race is hard to determine visually: Rather than taking the government’s word for it, the jurors can see for themselves that the perpetrator must be the suspect whom the government has charged. But the decision to record the probabilities of each match in racial terms gives a creepy whiff of eugenics to the CODIS database. And this might become all the more unsettling as the racial disparities in the database increase.
The first political controversy about the national database focused not on race but on partial matches. These occur when, in an initial search of the database, investigators don’t look for perfect matches at each of the 23 genetic locations, but a routine search allows for a little imprecision at each location because of so many different laboratories and agents. In 10 years of operation, the FBI is aware of only seven partial matches using CODIS. But they led to a dilemma. With a partial match in hand, the FBI is confident that the offender in the database isn’t the source of the DNA at the crime scene. But investigators might be inclined to release the innocent person’s name to law enforcement so that his family members can be investigated as possible exact matches.
Sometimes this works in the best way possible: to exonerate the innocent and convict the guilty. The most famous example of this led to the release of Daryl Hunt, a North Carolina man who, as 60 Minutes has documented, spent 19 years behind bars for a brutal rape and murder of a newspaper editor. Nine years after DNA testing first cleared him of rape, which happened 10 years after his initial conviction, the state ran DNA from the crime scene through its state database. The result was a near match to a convicted felon named Anthony Brown, indicating that Hunt could not have committed the murder, but a relative of Brown’s might have. FBI rules at the time allowed states to share the results of partial matches within their own borders, and further investigation revealed that Brown had a brother named Willard in a nearby county. Investigators tracked down Willard Brown, offered him a cigarette, and, as soon as the interview ended, tested the DNA on it. It matched the DNA at the crime scene perfectly. Based on the partial match, Willard Brown confessed and Daryl Hunt was eventually freed.
In 2005, Mitch Morrissey asked the FBI to authorize states to share the results of partial matches across state lines. He had discovered partial matches between the genetic evidence left by three rapists in Colorado and the profiles of convicted offenders in Oregon, Arizona, and California. Morrissey wanted Oregon to test its sample to determine whether its convicted offender shared a Y chromosome with the material found at Morrissey’s crime scene. This technique is called YSTR analysis, and it’s a way of narrowing down a long list of suspects with similar DNA to determine whether they are, in fact, related. Close male relatives share a Y chromosome: My two sons and I, for example, have the same Y chromosome as my father and my father’s brother. In a YSTR test, the police analyze the Y chromosome of the convicted offender in the database who didn’t commit the crime and then compare it with the Y chromosome on the genetic evidence from the crime scene. If the DNA is different, it means the brother, or son, of the offender in the database didn’t commit the crime. If the Y chromosome is the same, he might well have.
Faced with Morrissey’s request, Tom Callaghan told Morrissey that the national database procedures prohibited states from sharing information about people who weren’t suspected of committing crimes. Releasing the names of offenders except in cases of a confirmed match, he told Morrissey, might be viewed by courts as an expansion of the database beyond its original purpose: to solve crimes by surveilling convicted criminals. Morrissey, who had used Callaghan as an expert witness, was surprised by Callaghan’s caution. He decided to go over Callaghan’s head and eventually spoke to Robert Mueller, the director of the FBI. “He said, ‘We have a problem?’ I said, ‘Yup.’ He said, ‘Give me 10 days and we’ll get it fixed.’ ” In 2006, on Mueller’s orders, the FBI established an interim policy allowing states to establish their own policies and procedures to follow up on partial matches that emerged from standard database searches. Armed with the new policy, Oregon and Arizona agreed to cooperate with Morrissey, but YSTR testing, which compared the Y chromosomes, revealed that neither of the profiles in the state databases was related to the Colorado rapists.
Morrissey wasn’t satisfied with his victory on partial matches. Because they occur unexpectedly and infrequently, they’re unlikely to produce lots of investigative leads. Accordingly, Morrissey began to pressure Callaghan and the FBI to change federal policy to allow not only partial matches but also familial searches. The main difference between the two techniques is that partial matches emerge inadvertently from a routine search of the database while family searches represent a second, deliberate trolling of the database for close biological relatives after the first search has failed to produce a perfect match.
The prospect of familial searches alarmed Callaghan. He feared that they might imperil the entire CODIS system since courts might view the searches as an even more troubling example of “CODIS creep”—an attempt by the government to use samples collected for one purpose for a very different purpose. Moreover, even though federal law requires the FBI to inform Congress every time it intends to change the genetic loci in the national database, proponents of familial searches wanted the FBI to act on its own. In Callaghan’s view, the database was an invaluable resource that generates more than thousands of leads a month. Given the fact that familial searching has a success rate of only about 10 percent in the United Kingdom, he reasoned, why jeopardize solving hundreds if not thousands of cases in the future to adopt a controversial and fringe technique that might solve just a handful of cases at most? So when he was asked to organize an FBI symposium on genetic privacy and familial searching, he jumped at the chance to invite critics to challenge a proposal he viewed as ill-advised.
What’s the Law on Expansive DNA Searches?
Last February, I got an e-mail from Callaghan, whom I’d never met, asking me to speak about the constitutionality of genetic privacy and familial searching at an FBI symposium on the topic. And so, in March, I showed up at the Sheraton Crystal City in Arlington, Va., where the FBI had assembled the legal advisers and administrators of all 50 state DNA databases.
The most enthusiastic boosters of familial searches spoke first. Mitch Morrissey explained that a pilot familial searching program in his city, using specially designed software, had identified three cases in which there was a 90 percent chance of a brother or a father-son link to the sample left at the crime scene and someone in the local database. Follow-up YSTR testing suggested a match in chromosome types, although, for different reasons, none of the leads actually led to convictions. Morrissey said he was trying to convince Colorado authorities to begin familial searching and criticized the FBI for resisting. “I liken it to having a Porsche and driving it like a Pinto,” he said.
Another enthusiastic proponent, Dr. Frederick Bieber of Harvard’s Brigham and Women’s Hospital, spoke in language that eerily echoed the eugenic family studies of the early 20th century. “Does crime cluster in families?” he asked. “We know that it does.” Nearly half of prison inmates in federal and state institutions had a family member who had been incarcerated, he announced. “If crime didn’t occur in clusters of families, all this would be an academic conversation.”
Then it was the critics’ turn. Barry Scheck of the Innocence Project said that he could imagine supporting familial searches if Congress authorized and carefully regulated them with safeguards such as the requirement of a judicial warrant and corroborating evidence. But, he argued, “I don’t think there can be any doubt that when the U.S. Congress passed the DNA identification Act of 1994, it did not think for a nanosecond that it was authorizing a database that was going to be used for purposes of familial searches.” Everyone who testified before Congress expected the database to be used for “law enforcement purposes”—by which they meant finding past offenders who could be linked to crime scenes. Several witnesses noted that the National Research Council of the National Academies, which provide scientific advice to the federal government, had warned in 1992 about the dangers of familial searches, citing concerns about “privacy and fairness” for “relatives who have committed no crime.” The council concluded, “Such uses should be prevented both by limitations on the software for search and by statutory guarantees of privacy.”
When my turn came, I said that the constitutional and legal arguments against familial searches weren’t clear and that courts might come down on both sides of the question. The main constitutional objections are that these searches violate the long-established principle that the Fourth Amendment prohibits searches—of a house or a database—for general law enforcement purposes without individualized suspicion of wrongdoing. And in the case of a familial search, the police already know that no one in the database committed the crime. On the other hand, if a familial search is backed up by a YSTR test, it seems less troubling from a privacy point of view because this means the name of a family member will be released for investigation only when there’s a high probability that the person is connected to the crime scene. Courts have often said that searches are reasonable when they’re highly effective at identifying the guilty and don’t invade the privacy of the innocent.
In addition, the framers of the Constitution were concerned about “corruption of blood.” They believed that you should be punished for what you do, rather than for the sins of your fathers. But the analogy isn’t perfect: In the case of familial searches, relatives are being investigated, not punished, and the ones identified through YSTR testing are likely to be guilty, not innocent.
The strongest legal argument against familial searches is that they’re not what Congress intended when it set up the database. In the leading case upholding the collection of DNA samples, U.S. v. Kincade, the U.S. Court of Appeals for the 9th Circuit stressed in 2004 that the government had two good reasons for requiring people on probation to provide a DNA sample: the diminished expectation of privacy that people have once they’re on probation, and the state’s strong interest in ensuring that they reform rather than becoming recidivists and commit new crimes in the future. Familial searches can’t be justified by either rationale. The family members of offenders have done nothing to reduce their expectation of privacy, and the state is investigating new crimes, not stopping repeat offenders.
If the legal implications were murky, the political implications were clear. Given the dramatic racial disparities of family searches, African-American families might be four times as likely to be put under genetic surveillance as white families. For this reason, I predicted that a national decision to begin familial searches without explicit congressional approval might cause a political firestorm that would imperil political support for the entire CODIS system. As custodians of the national and state databases, I concluded with a melodramatic flourish, the officials at the symposium might want to proceed cautiously rather than risk being accused of violating a public trust.
Several months later, I called Tom Callaghan for an update. He sounded relieved. The criticisms of familial searches at the symposium, especially the predictions of a political firestorm, had dampened enthusiasm for implementing them at a national level. The scientific working group advising the FBI in July, he said, provided scientific recommendations about partial matches but was silent about whether CODIS should adopt familial searches. Individual states, however, remain free to proceed with familial searches on their own.
The Grim Sleeper Slips the DNA Dragnet. But the Rest of Us Might Not.
Between 1985 and 2007, a California serial killer shot and strangled at least 11 victims, most of them African-American women. Newspapers called him the “Grim Sleeper” because 13 years elapsed between two of the murders. Starting in 2004, DNA analysis linked several of the crime scenes to one another, but a search of the CODIS database failed to identify the killer.
Soon after the FBI symposium I attended, California decided to forge ahead with familial searching on its own, partly in the hope of identifying the Grim Sleeper. According to the Los Angeles Times, Jerry Brown, the state’s attorney general, overruled his legal adviser’s concerns that judges might strike down familial searches on constitutional grounds and decided to authorize them in April. The policy that California adopted limits the use of the searches to cases with “critical public safety implications,” in which no search of the offender’s crime scene DNA has produced a direct hit or partial match. There are a few safeguards for privacy, most notably the requirement that any DNA evidence be subjected to YSTR testing before a name is released to law enforcement to confirm a probable link between the sample at the crime scene and the target of the familial search.
Since California adopted its new familial searching policy, the attorney general’s office has authorized two searches that failed to produce a YSTR match between the potential-offender sample and the crime scene sample. Most prominently, a familial search did not identify the “Grim Sleeper.” State officials downplayed the failure, emphasizing that they always estimated the chance of finding the serial killer this way as something like 1 in 10.
At the moment, California is limiting its familial searches to convicted offenders. Still, the searches may provoke lawsuits challenging the expansion of the state database to include arrestees—just as Tom Callaghan feared. And courts could well be troubled by the open-ended idea that once you’re arrested and cleared, the state can subject you and future generations of your family members to permanent genetic surveillance.
Indeed, that’s precisely why the European Court of Human Rights last December ruled that the United Kingdom’s decision to store the DNA of arrestees violated European privacy guarantees. The court was especially concerned about the possibilities of familial searches of the DNA of arrestees. The court was also troubled that “the processing of DNA profiles allows the authorities to assess the likely ethnic origin of the donor.”
U.S. courts may be similarly skeptical of the decision to include arrestees in state and federal databases. When Congress, in 2006, authorized the FBI to place the DNA profiles of federal arrestees in the CODIS database, the FBI further required that state databases allow arrestees to expunge their DNA from state databases if they are subsequently cleared or not charged. Thirty-eight states have laws with explicit expungement procedures (although they are not always easy to use). Nevertheless, the Minnesota Court of Appeals in 2006 struck down a state law authorizing the collection of DNA from arrestees for violating the constitutional requirement that searches may not be conducted without a warrant. The court emphasized that under the state law it was striking down, no one had to consider whether the DNA was in any way related to the charged crime or any other criminal activity. On the other hand, the Virginia Court of Appeals reached the opposite conclusion, in 2007, holding that taking a DNA sample on arrest is no different than taking a fingerprint.
How the Supreme Court would rule on familial searches of arrestees is an open question. Two years ago, in the foreword to a book about the technology of justice, Justice Stephen Breyer wrote, “DNA identification may raise privacy concerns. Suppose a check of a convict DNA database reveals a near miss, thereby implicating a relative who has no record of conviction and was consequently not included in the bank. What kind of legal rules should apply?”
Stephen Mercer, a defense attorney who convinced the Maryland legislature to ban familial searches, predicts that courts will be skeptical of the expansion of DNA databases to include arrestees. As genetic research—led by private companies such as 23andMe—reveals increasing ties between genes and predisposition to violence and other antisocial behavior, there may be growing discomfort with the idea of giving the government access to DNA, which could lead to people being surveilled, detained, or suspected for their behavioral tendencies rather than their actions. Especially given the risk of racial bias.
The standard answer to the racial bias charge is this: Expand the database to include everyone. Some progressive scholars, such as Akhil Amar of Yale Law School, have argued that a universal database, such as the one created in Iceland, “would be a godsend to innocent convicts.” For this reason, Amar has argued, U.S. citizens should be compelled to donate their DNA to a universal database, as long as there are strict privacy controls. He would limit “testing to so-called junk DNA—parts of the DNA code that identify individuals without revealing other medicals facts” and “allowing the government to search the database only for important needs, as certified by a special DNA court.”
This seems utopian. As Mercer notes, “the problem is that the FBI keeps the original sample, and that firewall that the FBI said exists between the genetic sample and the edited profile is being breached through familial searching. Now that they’re willing to go back to the original samples for YSTR testing, to determine familial relationships, how long will it be before they say: let’s test those snips to see if someone is a sociopath?”
Moreover, it’s hard to imagine a scenario in which Congress would limit the searching of the DNA database only to serious crimes. Lawmakers refused to impose a similar limitation on foreign surveillance searches during debates over the Patriot Act, succumbing to the bipartisan arguments that those who have nothing to hide should have nothing to fear.
Nor does the Obama administration seem likely to encourage courts to impose the kinds of complicated and nuanced controls on information sharing that a universal database would require. Last month, the administration disappointed privacy advocates by arguing before the Supreme Court that there is no constitutional right for convicts to obtain DNA evidence that might exonerate them.
Neither the United States nor the United Kingdom have any models for the kind of comprehensive privacy regulations that would prevent the government from sharing DNA profiles in law enforcement databases with insurance companies, employers, schools, and the private sector. For this reason, while a perfectly regulated universal database may be conceivable in theory, it’s nearly impossible to imagine in practice. And a universal database that can be consulted for any crime, serious or trivial, is one that many citizens would resist. It opens us to a world in which, based on the seemingly infallible evidence of DNA, people can be framed or tracked, by their enemies or by the government, in ways that liberal societies have traditionally found unacceptable.
What stands between us and this unsettling future are decisions by conscientious public officials, such as Tom Callaghan’s efforts to prevent the FBI from surreptitiously expanding its database to include familial searches without congressional authorization. Thanks to Callaghan’s determination to abide by the law rather than push the envelope, national familial searches are on hold for now. But they may soon become reality as states implement them on their own. There are relentless pressures—well-intentioned but shortsighted—to expand DNA databases without meaningful regulations or controls. And as California’s decision to adopt familial searches shows, all the political incentives are on the side of expansion rather than regulation. It’s unfortunate, in any event, that scenarios previously limited to movies like Minority Report are unfolding quietly, before most of us have thought about the consequences.