As of today, the Ninth Circuit Court of Appeals has twice upheld the constitutionality of DNA collection before conviction (DNA-BC). The first opinion came in United States v. Pool, 621 F.3d 1213 (9th Cir. 2010). There, the panel applied a totality-of-the-circumstances balancing test to uphold the federal arrestee law. However, the full court promptly agreed to rehear the case en banc (thus vacating the opinion in Pool).
At that point, another challenge to DNA-BC was pending. Haskell v. Brown, 677 F.Supp.2d 1187 (N.D. Cal. 2009), was a class action brought to enjoin enforcement of the California Proposition that initiated DNA arrestee sampling in that state. Determining that this law was, on balance, constitutionally reasonable, District Judge Charles Breyer, denied the request for a preliminary injunction. Plaintiffs appealed this ruling.
With eleven Ninth Circuit judges stirring the pot in Pool, Haskell moved to the back burner. But then Pool pled guilty, and the en banc court dropped his case as moot. Pool, 659 F.3d 761 (9th Cir. 2011) (en banc). Hence, the Haskell panel returned to work.
Today the case boiled over with the release of the panel’s opinion. As in Pool, the court split 2-1 in favor of DNA-BC. The two judges in Haskell to side with the state were Milan D. Smith, Jr., and James Dale Todd, a senior district judge for the Western District of Kentucky. Judge William A. Fletcher wrote a sharp dissent. Judge Smith devoted much of his majority opinion to refuting Judge Fletcher’s dissent. As a preliminary matter, they disagreed over whether Friedman v. Boucher, 580 F.3d 847 (9th Cir. 2009), a previous Ninth Circuit case not even involving a DNA database forced the court’s hand here. Judge Smith correctly wrote that “very broad dicta” in the case did not compel any particular result in Haskell.
With the path cleared, Judge Smith asked whether the “totality of the circumstances” were such that DNA-DB under California law was “reasonable” within the meaning of the Fourth Amendment. Judge Fletcher disputed the use of this amorphous standard. On this issue, neither opinion is convincing. Time after time, in case after case, the Supreme Court has repeated an important mantra—full fledged searches and seizures intended to produce evidence of criminal conduct are per se unreasonable, regardless of the other circumstances, when they are undertaken without a warrant and when they do not fall within a categorical exception to this warrant requirement. E.g., Kentucky v. King, 131 S.Ct. 1849 (2011); City of Ontario v. Quon, 130 S.Ct. 2619 (2010); Katz v. United States, 389 U.S. 347, 357 (1967). Therefore, it will not do to state, as Judge Smith blandly did, that totality-of-the-circumstances balancing is the norm in Fourth Amendment cases. To the contrary, it is an anomaly that has been used only in two Supreme Court cases (for searches involving probationers or parolees). Upholding DNA-BC therefore requires showing that the program falls within an established exception—or recognizing a new exception to accommodate the acquisition of fingerprints, photographs, and now, DNA profiles for use in databases. The former move is possible but takes considerable effort. The latter would be the more candid and convincing mode of analysis.
But just as the Haskell majority’s leap to totality balancing is largely unjustified [update of June 16, 2013: The Supreme Court in Maryland v. King made an attempt to justify such balancing and used it to uphold a system of routine DNA sampling before conviction], the dissent’s theory of why DNA-BC is unconstitutional is groundless. According to dissenting Judge Fletcher, the Supreme Court held in Hayes v. Florida, 470 U.S. 811 (1985), and Davis v. Mississippi, 394 U.S. 721 (1969), that identifying information (fingerprints, in those cases) “may not be taken from an arrestee solely for an investigative purpose, absent a warrant or reasonable suspicion that the [information] would help solve the crime for which he was taken into custody.” This characterization overlooks the most critical facts in the cases and the explicit reasoning of the opinions. Justice White’s majority opinion in Hayes could not have been clearer in explaining that the cases simply hold “that transportation to and investigative detention at the station house without probable cause or judicial authorization together violate the Fourth Amendment.” 470 U.S. at 815. That the fingerprints were used for investigative purposes and not just to establish the true identities of the arrested suspects had nothing to do with the constitutional infirmity. “[T]he fingerprints ... were the inadmissible fruits of an illegal detention” (id. at 813) because “the police at that time were without probable cause for an arrest, there was no warrant, and [the suspects] had not consented to being taken to the station house.” Id. at 313-14. Not a single Justice dissented from this understanding of the issue. Yet, Judge Fletcher transformed the simple and unremarkable result in Hayes and Davis into a condemnation of all forms of DNA-BC. Judge Smith was kind to call this aggressive reading of the cases a “novel interpretation.”
The Haskell court also split on the outcome of the balancing test. The majority was quick to accept the state’s litany of interests in DNA-BC as weighty, whereas the dissent demanded more rigorous proof. Conversely, the dissent was quick to accept nonscientific speculations about the dangers of DNA profiles and the risks of misuse of retained DNA samples, whereas the majority was loath to invalidate today’s laws on the grounds that they might be misused or changed.
Further developments are likely in the Ninth Circuit. Of the six judges to vote in Haskell and Pool (three per case), two of them came from outside the circuit and sat by designation. Counting only the Ninth Circuit noses, the vote to uphold DNA-DB has been 2-2 — not exactly a stable alignment. Obviously, there is considerable turmoil among the Ninth Circuit judges, and an en banc order in Haskell seems probable.
Acknowledgments: Thanks to Jen Wagner for sending me a copy of today’s slip opinion in Haskell.
Cross-posted from The Double Helix Law Blog.
Friday, 24 February 2012
Saturday, 18 February 2012
Government-sponsored Report on Latent Fingerprint Work in Criminal Investigation and Prosecution
From NIST's Office of Law Enforcement Standards (OLES), as posted on February 17, 2012: “Expert Working Group Reduces Human Error Analyzing Fingerprints.” Sounds great. So how has the group succeeded in reducing these errors? Alas, clicking on the link reveals that, as yet, the group has not prevented a single error, human or inhuman. But NIST has released a long overdue report prepared by an expert working group that it assembled and funded for an extended period.
The final 234-page report from the committee of 34 experts contains 34 recommendations to improve latent print examination in criminal investigations and in presentations of the findings to police, lawyers, judges, and juries. The committee believes that implementation of these recommendations eventually could reduce the incidence of errors in latent print work and in the understanding of the results. The recommendations are a kind of wish list with entries that range from studying how to test the vision of fingerprint analysts; to educating examiners “in the scientific method” and in probabilistic and statistical thinking; to creating “a culture” in which examiners can confess honest error without fear of punishment; to monitoring the day-to-day work and courtroom testimony; and to certification for all analysts and accreditation for all laboratories.
Lawyers will be interested in the admonishments in the report about insulating examiners from unnecessary extraneous information; contemporaneously documenting the thought processes of examiners; making complete reports available; and curbing exaggerated claims of conclusive identifications. Concluding that existing data does not permit reasonably accurate estimates of error rates in real-life fingerprint identifications, for example, the report insists that an “expert should not state that errors are inherently impossible or that a method inherently has a zero error rate.” Likewise, the report states that “latent print examiners should not report or testify, directly or by implication, to a source attribution to the exclusion of all others in the world.” However, the group of latent print examiners, forensic laboratory scientists and managers, psychologists, engineers, statisticians, and lawyers, was unable to agree on what statements should replace such well entrenched testimony.
Given these tensions and ambitions, will NIST's premature assertion that “Expert Working Group Reduces Human Error Analyzing Fingerprints” come true? One can only hope for the best.
--DH Kaye, Feb. 18, 2012
Irony: The NIST statement accompanying this report on error and human factors has errors of its own. The actual publication date was February 17, not February 9. Authors’ names are not always spelled correctly, and support staff are listed as authors. No doubt, this says something about human fallibility if not human factors.
Disclosure: The author of this note was a working group member and the editor of the report described here. See Expert Working Group on Human Factors in Latent Print Analysis, Latent Print Examination and Human Factors: Improving the Practice through a Systems Approach, Feb. 2012.
The final 234-page report from the committee of 34 experts contains 34 recommendations to improve latent print examination in criminal investigations and in presentations of the findings to police, lawyers, judges, and juries. The committee believes that implementation of these recommendations eventually could reduce the incidence of errors in latent print work and in the understanding of the results. The recommendations are a kind of wish list with entries that range from studying how to test the vision of fingerprint analysts; to educating examiners “in the scientific method” and in probabilistic and statistical thinking; to creating “a culture” in which examiners can confess honest error without fear of punishment; to monitoring the day-to-day work and courtroom testimony; and to certification for all analysts and accreditation for all laboratories.
Lawyers will be interested in the admonishments in the report about insulating examiners from unnecessary extraneous information; contemporaneously documenting the thought processes of examiners; making complete reports available; and curbing exaggerated claims of conclusive identifications. Concluding that existing data does not permit reasonably accurate estimates of error rates in real-life fingerprint identifications, for example, the report insists that an “expert should not state that errors are inherently impossible or that a method inherently has a zero error rate.” Likewise, the report states that “latent print examiners should not report or testify, directly or by implication, to a source attribution to the exclusion of all others in the world.” However, the group of latent print examiners, forensic laboratory scientists and managers, psychologists, engineers, statisticians, and lawyers, was unable to agree on what statements should replace such well entrenched testimony.
Given these tensions and ambitions, will NIST's premature assertion that “Expert Working Group Reduces Human Error Analyzing Fingerprints” come true? One can only hope for the best.
--DH Kaye, Feb. 18, 2012
Irony: The NIST statement accompanying this report on error and human factors has errors of its own. The actual publication date was February 17, not February 9. Authors’ names are not always spelled correctly, and support staff are listed as authors. No doubt, this says something about human fallibility if not human factors.
Disclosure: The author of this note was a working group member and the editor of the report described here. See Expert Working Group on Human Factors in Latent Print Analysis, Latent Print Examination and Human Factors: Improving the Practice through a Systems Approach, Feb. 2012.
Sunday, 5 February 2012
Genetic Geneology: Is Sarah Yarborough’s Killer a Mayflower Descendant?
In The Double Helix and the Law of Evidence, I remarked that “[s]urname prediction is a concept that has yet to be implemented in forensic science . . . ” (1, p. 211). Now, cold-case detectives investigating the 1991 rape and strangulation of a 16-year old girl in Seattle are looking for “a distant relative of Edward and Samuel Fuller, who crossed the Atlantic on the Mayflower in 1620 [and who] might even share their last name” (2).
Sarah Yarborough left home to take part in a dance-team competition one morning in December. She never returned. Six people had seen a man in his 20s, with shoulder-length blond hair, a trench coat and dark pants speaking with her or leaving a brushy area where her body was found. In the succeeding months, more than 3,000 tips poured in. DNA analysis excluded some suspects, but trawling offender DNA databases for an exact match produced no leads, and the killer remains unidentified.
Yet, other databases can be searched. “Genetic geneology databases” (3, p. 384) that use the paternally inherited Y-STR haplotypes that are of interest in anthropology are growing in popularity. Indeed, some are available on the web. At the suggestion of a scientist at the state crime laboratory, police turned to Identifinders of Huntington Beach, California. Presumably, the firm combed through the online databases to arrive at its conclusion that “there's a good chance the killer's last name is or was Fuller” (4) because “the killer is a descendant of Robert Fuller, who arrived in Salem, Mass., in 1630. Fuller was not himself on the Mayflower, but he was related to three passengers: Edward Fuller, as well as Edward Fuller's brother, Samuel, and 12-year-old son” (2).
The value (and accuracy) of this surmise remains to be seen. None of the tips the police received involve a Fuller (2), and many Fullers had nothing to do with that historic voyage. Moreover, the Mayflower Fullers may be multitudinous. After all, “there are tens of millions of people descended from the 102 passengers and about 25 crew members who arrived on the Mayflower . . .” (2).
Acknowledgment: Thanks to Gene Johnson for telling me about the case.
References
Sarah Yarborough left home to take part in a dance-team competition one morning in December. She never returned. Six people had seen a man in his 20s, with shoulder-length blond hair, a trench coat and dark pants speaking with her or leaving a brushy area where her body was found. In the succeeding months, more than 3,000 tips poured in. DNA analysis excluded some suspects, but trawling offender DNA databases for an exact match produced no leads, and the killer remains unidentified.
Yet, other databases can be searched. “Genetic geneology databases” (3, p. 384) that use the paternally inherited Y-STR haplotypes that are of interest in anthropology are growing in popularity. Indeed, some are available on the web. At the suggestion of a scientist at the state crime laboratory, police turned to Identifinders of Huntington Beach, California. Presumably, the firm combed through the online databases to arrive at its conclusion that “there's a good chance the killer's last name is or was Fuller” (4) because “the killer is a descendant of Robert Fuller, who arrived in Salem, Mass., in 1630. Fuller was not himself on the Mayflower, but he was related to three passengers: Edward Fuller, as well as Edward Fuller's brother, Samuel, and 12-year-old son” (2).
The value (and accuracy) of this surmise remains to be seen. None of the tips the police received involve a Fuller (2), and many Fullers had nothing to do with that historic voyage. Moreover, the Mayflower Fullers may be multitudinous. After all, “there are tens of millions of people descended from the 102 passengers and about 25 crew members who arrived on the Mayflower . . .” (2).
Acknowledgment: Thanks to Gene Johnson for telling me about the case.
References
- David H. Kaye, The Double Helix and the Law of Evidence (2010).
- Gene Johnson, Awkward Twist over DNA Link in Teen's Slaying, Seattle Times, Jan. 12, 2012
- John M. Butler, Advanced Topics in Forensic DNA Typing: Methodology (2011).
- Gene Johnson, Mayflower Relative Sought in '91 Wash. Killing, Seattle Post-Intelligencer, Jan. 11, 2012
Cross-posted from The Double Law Blog
Thursday, 5 January 2012
When Is DNA Collection a "Search or Seizure"?
In a brief opinion, the Supreme Court of Louisiana invoked the constitutionality of routine DNA sampling of custodial arrestees to uphold an order for a cheek swab to an indicted defendant. The case presents an unusual set of facts in that investigators were seeking, for no apparent reason, what the court assumed to be a second sample from the defendant.
After "a shooting in which three individuals were killed" and another wounded, "a witness informed [police] that he observed [Charles] Franklin and Dwayne Johnson changing out of their clothing and into clothing that was black." Franklin and Johnson then were arrested. At their preliminary examination, a detective "testified that, pursuant to a search warrant for a nearby residence, he recovered clothing similar to what Mr. Franklin was observed wearing and subsequently removing on the night of the offense." State v. Franklin, No. 2010-K-0564 (La. Ct. App. Aug. 11, 2010), rev'd, No. 2011-1909 (La. Dec. 16, 2011),
Rather than apply for a search warrant for a DNA sample, the state obtained a court order for Franklin to give a buccal swab. Id. Before the order could be enforced, however, the Court of Appeals held that this procedure constituted an unreasonable search or seizure because
This observation is puzzling. If the state already had a sample for the state databank, why demand that defendant provide another before it had even recovered and analyzed any DNA from the clothing? More fundamentally, the court failed to articulate clearly why DNA sampling "as a routine incident of booking" supported later warrantless DNA sampling. Merely comparing a recorded profile to one developed from the clothes would not rise to the level of a "search" requiring constitutional justification (see DNA Database Trawls and the Definition of a Search in Boroian v. Mueller), but this fact does not explain why the state can compel a suspect to open his mouth to a scraping of cells from the inside of his cheek. Does the intrusion into the mouth fall short of a search in Louisiana?
The Louisiana Supreme Court came close to taking this position. It expressed some skepticism of the notion that
In any event, both probable cause and reasonable suspicion involve probabilities, not certainties. The Franklin court is correct in that regard. But this tells us next to nothing about the level of probability that is required to justify a search under either standard. There is no obvious reason that investigators cannot analyze clothing from the vicinity of a crime before demanding a DNA sample from a suspect.
In this case, however, "the state refined its motion by presenting ... the trial judge with a crime lab report ... indicating that the evidence, clothing taken from defendant's residence in a prior search, contains blood and a single hair strand with follicle, which may be tested and which may yield a DNA profile ... ." Still, "may yield a DNA profile" is a rather weak promise from the laboratory. Any item of clothing "may yield" a profile.
References
Acknowledgments
After "a shooting in which three individuals were killed" and another wounded, "a witness informed [police] that he observed [Charles] Franklin and Dwayne Johnson changing out of their clothing and into clothing that was black." Franklin and Johnson then were arrested. At their preliminary examination, a detective "testified that, pursuant to a search warrant for a nearby residence, he recovered clothing similar to what Mr. Franklin was observed wearing and subsequently removing on the night of the offense." State v. Franklin, No. 2010-K-0564 (La. Ct. App. Aug. 11, 2010), rev'd, No. 2011-1909 (La. Dec. 16, 2011),
Rather than apply for a search warrant for a DNA sample, the state obtained a court order for Franklin to give a buccal swab. Id. Before the order could be enforced, however, the Court of Appeals held that this procedure constituted an unreasonable search or seizure because
The State has not developed a DNA profile from the seized clothing; therefore, the State has not established that the DNA to be seized from Mr. Franklin, standing alone, could 'constitute evidence tending to prove the commission of an offense.The Louisiana Supreme Court reversed. At first, it seemed to maintain that judicial review and the equivalent of a warrant were superfluous. Louisiana was the first state to authorize DNA sampling as part of a custodial arrest, and the court was willing to "presume that the Louisiana State Police Crime Laboratory already possesses a sample of defendant's DNA to which its CODIS DNA section has access for testing as part of a cooperative effort between state and federal authorities to develop a comprehensive identification data base."
This observation is puzzling. If the state already had a sample for the state databank, why demand that defendant provide another before it had even recovered and analyzed any DNA from the clothing? More fundamentally, the court failed to articulate clearly why DNA sampling "as a routine incident of booking" supported later warrantless DNA sampling. Merely comparing a recorded profile to one developed from the clothes would not rise to the level of a "search" requiring constitutional justification (see DNA Database Trawls and the Definition of a Search in Boroian v. Mueller), but this fact does not explain why the state can compel a suspect to open his mouth to a scraping of cells from the inside of his cheek. Does the intrusion into the mouth fall short of a search in Louisiana?
The Louisiana Supreme Court came close to taking this position. It expressed some skepticism of the notion that
a warrant is required to take another buccal sample from defendant for purposes of testing by the Crime Lab's Forensic DNA Section as part of the effort to solve a particular crime . . . because collection of the sample, although minimally intrusive, nevertheless constitutes a search."This skepticism seems difficult to square with Cupp v. Murphy, 412 U.S. 291 (1973). There, the U.S. Supreme Court described the extraction of material beneath a suspect's thumbnail to be a "'severe, though brief, intrusion upon cherished personal security' that is subject to constitutional scrutiny." The Louisiana court seemed to recognize the problem (although it did not mention Cupp), for it assumed arguendo that the cheek swabbing "constitutes a search." On that assumption, the court reasoned that
the state ... need only show a "fair probability" the sample may yield evidence to establish the requisite probable cause, not that the search will produce evidence probative one way or another on the question of guilt. Illinois v. Gates, 462 U.S. 213, 231 (1983) ("'The process does not deal with hard certainties, but with probabilities.'") ... . Given probable cause that defendant participated in the charged crimes, it suffices that the state has established that its evidence, removed from defendant's residence, contains recoverable biological samples that may, but not necessarily will, provide a DNA profile useful for forensic purposes.This part of the opinion poses another puzzle or two. Why demand probable cause? In Cupp, Justice Stewart was quoting from Terry v. Ohio, 392 U.S. 1 (1968), to show that Fourth Amendment protections came into play. But the invasion in Terry was a stop-and-frisk rather than a prolonged arrest or a thorough search. The Court held the fourth amendment applied to the lesser invasion, but it was unwilling to demand probable cause. Instead, it countenanced pat downs on the basis of reasonable suspicion. Likewise, "nontestimonial orders" for blood, hair, fingerprints, voice exemplars, and the like, do not require probable cause.
In any event, both probable cause and reasonable suspicion involve probabilities, not certainties. The Franklin court is correct in that regard. But this tells us next to nothing about the level of probability that is required to justify a search under either standard. There is no obvious reason that investigators cannot analyze clothing from the vicinity of a crime before demanding a DNA sample from a suspect.
In this case, however, "the state refined its motion by presenting ... the trial judge with a crime lab report ... indicating that the evidence, clothing taken from defendant's residence in a prior search, contains blood and a single hair strand with follicle, which may be tested and which may yield a DNA profile ... ." Still, "may yield a DNA profile" is a rather weak promise from the laboratory. Any item of clothing "may yield" a profile.
References
- State v. Franklin, No. 2011-1909, 2011 WL 6312400 (La. Dec. 16, 2011)
- Cupp v. Murphy, 412 U.S. 291 (1973)
- Terry v. Ohio, 392 U.S. 1 (1968)
- David H. Kaye, DNA Database Trawls and the Definition of a Search in Boroian v. Mueller, 97 Va. L. Rev. in Brief 41 (2011)
Acknowledgments
- Thanks to Lisa Hurst for calling the opinion to my attention.
- Cross-posted from The Double Helix Law Blog.
Monday, 19 December 2011
Finegan's Wake: A Partial DNA Match in Rhode Island
A Rhode Island newspaper reports that "[t]he Cranston Police Department has arrested 49-year-old David Finegan, of no permanent address, for the burglary and sexual assault of an 81-year-old woman." [1] Police "collected [DNA] at the crime scene on May 2." On June 10, the Rhode Island Department of Health reported that it had a DNA profile from the crime with which to query an offender database. The Rhode Island laboratory did not use software designed for kinship searching, but on July 8, the department advised detectives that it had found a partial DNA match to a female inmate.
The article does not speculate on why it took a month to complete a routine computer search and report the results.Was it because of legal concerns? Was the partial match trawl intentional, or was the discovery inadvertent? Whichever it was, the detectives turned their attention to five male siblings. They discovered that one of them, David Finegan, was "in close proximity . . . on the night of the incident."
Why this roundabout method of identifying Finegan? He was on parole in July. Was the underlying offense not one that triggered entry into the DNA database? Was there a backlog in entering offender profiles into the database? Whatever the explanation, Finegan had the misfortune of being picked up on July 14 on a parole violation and held for the weekend. Detectives quickly obtained a search warrant and took a sample of DNA from him before he made bail and dropped out of sight. A week later, they learned that it matched the crime-scene DNA.
Pursuing an anonymous tip, police found and arrested Finegan in Providence. He "is being charged with burglary and first-degree sexual assault." Interestingly, he has an arrest record (including domestic assault, felony DWI, resisting arrest and other assaults) dating back to 1991. A bill that would expand the state database to include arrestees is before House and Senate committees in Rhode Island.
Reference
1. Joe Kernan, Arrest Made in Rape of Elderly Woman, Cranston Herald, Dec. 19, 2011
Acknowledgment
Thanks to Frederick Bieber for informing me of the Cranston Herald article.
The article does not speculate on why it took a month to complete a routine computer search and report the results.Was it because of legal concerns? Was the partial match trawl intentional, or was the discovery inadvertent? Whichever it was, the detectives turned their attention to five male siblings. They discovered that one of them, David Finegan, was "in close proximity . . . on the night of the incident."
Why this roundabout method of identifying Finegan? He was on parole in July. Was the underlying offense not one that triggered entry into the DNA database? Was there a backlog in entering offender profiles into the database? Whatever the explanation, Finegan had the misfortune of being picked up on July 14 on a parole violation and held for the weekend. Detectives quickly obtained a search warrant and took a sample of DNA from him before he made bail and dropped out of sight. A week later, they learned that it matched the crime-scene DNA.
Pursuing an anonymous tip, police found and arrested Finegan in Providence. He "is being charged with burglary and first-degree sexual assault." Interestingly, he has an arrest record (including domestic assault, felony DWI, resisting arrest and other assaults) dating back to 1991. A bill that would expand the state database to include arrestees is before House and Senate committees in Rhode Island.
Reference
1. Joe Kernan, Arrest Made in Rape of Elderly Woman, Cranston Herald, Dec. 19, 2011
Acknowledgment
Thanks to Frederick Bieber for informing me of the Cranston Herald article.
Cross-posted from The Double Helix Law Blog
Friday, 16 December 2011
Abusing AFIS -- Conviction by Computer?
On television shows such as NCIS, seconds after a forensic scientist inserts an image of a latent fingerprint into a machine connected to an automated fingerprint identification system (an AFIS), a computer screen blinks MATCH FOUND ... MATCH FOUND ... MATCH FOUND ... , and the chase is on. Apparently, some police and prosecutors think this is real.
Robert Garrett, a past president of the International Association for Identification, writing in the Evidence Technology Magazine, describes three cases in which governments have taken very serious actions against individuals based solely on the output of an AFIS search — with no review by any latent print examiner [1]. AFIS searches generate a list of potential matches — often the 20 closest matches as determined by an algorithm that looks at prints very differently from the way humans do. There is no proof that the AFIS ranks the candidates in the way that a skilled human examiner, relying on more information in the images, would. A number 1 candidate can be an obvious mismatch (to the human eye and brain). The Scientific Working Group on Friction Ridge Analysis, Study, and Technology (SWGFAST) insists that “AFIS ranks and scores have no role in formulating and stating conclusions based on ACE-V,” the steps that latent print examiners follow [2] and that “[t]he practice of relying on current AFIS technology to individualize latent prints correctly is not sound” [3].
According to Mr. Garrett, U.S. Customs and Immigration Enforcement relies on raw AFIS results to initiate deportation proceedings, and grand juries issue indictments relying on this information without testimony from a qualified examiner that the AFIS match means anything [1, p.10]. He concludes that “AFIS hits must be examined by a qualified fingerprint examiner and the results of that examination verified before any proceedings are commenced against a potential suspect. It is unethical, unprofessional, and—most likely—unconstitutional to do otherwise” [1, p. 11].
I would not presume to question Mr. Garrett’s professional judgment of what is unprofessional conduct in the fingerprint expert community, but it seems fair to ask on what basis he concludes that the practice of using unreviewed AFIS output is “most likely ... unconstitutional.” The article offers two possible bases for this judgment. First, Mr. Gerrett writes that
The second suggestion regarding the constitutionality of AFIS evidence is that
In any event, the article is an eye-opener. Police and prosecutors who rely on unexamined AFIS matches are not acting professionally or responsibly. Mr. Garrett deserves thanks for shedding some light on this remarkable practice.
References
1. Robert J. Garrett, Automated Fingerprint Identification Systems (AFIS) and the Identification Process, Forensic Sci. Mag., July-Aug. 2011, at 10–11.
2. SWGFAST, Position Statement on the Role of AFIS Ranks and Scores and the ACE-V Process, Oct. 15, 2011.
3. SWGFAST, Press Kit, May 19, 2004, at 14.1.3.
Robert Garrett, a past president of the International Association for Identification, writing in the Evidence Technology Magazine, describes three cases in which governments have taken very serious actions against individuals based solely on the output of an AFIS search — with no review by any latent print examiner [1]. AFIS searches generate a list of potential matches — often the 20 closest matches as determined by an algorithm that looks at prints very differently from the way humans do. There is no proof that the AFIS ranks the candidates in the way that a skilled human examiner, relying on more information in the images, would. A number 1 candidate can be an obvious mismatch (to the human eye and brain). The Scientific Working Group on Friction Ridge Analysis, Study, and Technology (SWGFAST) insists that “AFIS ranks and scores have no role in formulating and stating conclusions based on ACE-V,” the steps that latent print examiners follow [2] and that “[t]he practice of relying on current AFIS technology to individualize latent prints correctly is not sound” [3].
According to Mr. Garrett, U.S. Customs and Immigration Enforcement relies on raw AFIS results to initiate deportation proceedings, and grand juries issue indictments relying on this information without testimony from a qualified examiner that the AFIS match means anything [1, p.10]. He concludes that “AFIS hits must be examined by a qualified fingerprint examiner and the results of that examination verified before any proceedings are commenced against a potential suspect. It is unethical, unprofessional, and—most likely—unconstitutional to do otherwise” [1, p. 11].
I would not presume to question Mr. Garrett’s professional judgment of what is unprofessional conduct in the fingerprint expert community, but it seems fair to ask on what basis he concludes that the practice of using unreviewed AFIS output is “most likely ... unconstitutional.” The article offers two possible bases for this judgment. First, Mr. Gerrett writes that
U.S. Supreme Court decisions in Melendez-Diaz v. Massachusetts and, more recently, Bullcoming v. New Mexico, reiterated a defendant’s Sixth Amendment right “to be confronted with the witnesses against him.” Reports of a laboratory or investigative finding do not satisfy the requirement.This won’t wash. The Sixth Amendment right to confrontation is a trial right. It does not apply to administrative and grand jury proceedings. Even the usual rules for expert evidence are not binding in these proceedings. At trial, an AFIS hit, not used as part of the basis for an examiner’s opinion, would be impermissible under the rules of evidence, since this kind of scientific evidence is neither scientifically valid (under Daubert) nor generally accepted in the scientific community (per Frye). But this does not make it unconstitutional. An argument could be made that it deprives the defendant of due process to be convicted on the strength of such evidence. Cf. McDaniel v. Brown, 130 S.Ct. 665 (2010). However, the Confrontation Clause line of cases on which Mr. Garrett relies suggests that “machine-generated” test results can be introduced without the separate judgment of a human examiner. The state of Illinois currently is seeking to exploit this idea — unpersuasively in the judgment of many commentators — in the pending Supreme Court case of Williams v. Illinois (involving DNA tests and discussed on other postings on this blog).
The second suggestion regarding the constitutionality of AFIS evidence is that
Our society and its government have embraced technology in various forms for its efficiency and economy. In the areas of law enforcement and public safety, these technological advances have included AFIS, the Combined DNA Index System (CODIS), airport security screening devices, and red light/traffic cameras. But these advances bring with them compromises of privacy and our right “…to be secure in their persons, houses, papers, and effects…”It would be extravagant to assert that AFIS, CODIS, airport magnetometers and ordinary scanners, and red light cameras are “most likely unconstitutional,” and this may not be what Mr. Garrett intended to state or imply.
In any event, the article is an eye-opener. Police and prosecutors who rely on unexamined AFIS matches are not acting professionally or responsibly. Mr. Garrett deserves thanks for shedding some light on this remarkable practice.
References
1. Robert J. Garrett, Automated Fingerprint Identification Systems (AFIS) and the Identification Process, Forensic Sci. Mag., July-Aug. 2011, at 10–11.
2. SWGFAST, Position Statement on the Role of AFIS Ranks and Scores and the ACE-V Process, Oct. 15, 2011.
3. SWGFAST, Press Kit, May 19, 2004, at 14.1.3.
Thursday, 15 December 2011
Williams v. Illinois (Part II: More Facts, from Outside the Record, and a Question of Ethics)
This morning, Professor Richard Friedman posted a revealing report that Cellmark sent to the Illinois State Police (ISP). As he explains, and as my previous posting on the facts of Williams v. Illinois indicated, the report consists of much more than "machine-generated" statements. But the "lodging" (not part of the record in the case) and Professor Friedman’s remarks warrant a few additions or revisions to my presentation of the facts of the case.
On cross-examination, ISP analyst Karen Kooi Abbinanti, who examined the blood sample that Williams gave under court order in another case, testified to William’s STR profile. Because ISP analyst Sandra Lambatos, who provided the state’s only evidence of a DNA match, testified that “there [was] a computer match generated of the male DNA profile found in semen from the vaginal swabs of [LJ] to a male DNA profile that had been identified as having originated from Sandy Williams,” I presumed that the Cellmark report listed this profile as coming from the male fraction of DNA in the vaginal swab. Indeed, Lambatos testified that the “allele chart” in the Cellmark report “included data that [she] used to run [her] data bank search.” Joint Appendix at 61. Thus, I wrote that
Ms. Lambatos, however, testified on cross-examination that the Cellmark report’s “deduced male donor profile” (to quote the report itself) was not actually a deduced profile, but only a list of deduced alleles. Joint Appendix at 71. Interpreting it in this fashion (which may well be the correct understanding what the unknown analyst meant to write), she searched the unspecified database for certain two-allele subsets of the three alleles— namely, (13, 13), (10, 13), and (12, 13). Id. This made sense because, if Cellmark had correctly identified the victim’s profile — something that Lambatos did not check — then the rapist rather than the victim had to be the source of the 13-repeat allele.
The circumscribed nature of Ms. Lambatos’s testimony on direct examination about the “DNA match” is worthy of comment. Full disclosure would have required a scientist to reveal that other male profiles than just Williams’ profile were “consistent with” the vaginal-swab mixture and could have been picked out of a database in her trawl. Instead, Ms. Lambatos acquiesced in or suggested confining her testimony to Williams’s matching profile and the random-match probability associated with that one profile. In other words, she chose not to acknowledge possibilities that were inconsistent with the state’s theory. Does such selectivity contravene the professional responsibility of forensic scientists to “[a]ttempt to qualify their responses while testifying when asked a question with the requirement that a simple ‘yes’ or ‘no’ answer be given, if answering ‘yes’ or ‘no’ would be misleading to the judge or the jury”? [1]
The answer, I think, depends on how misleading Ms. Labatos’s answers on direct examination were. This was not a case of a single profile that probably could exclude everybody except for a twin brother. The analysts were unable to distinguish between Sandy Williams and other males with similar, but not identical profiles, as possible sources of the male DNA. By not disclosing this fact, Ms. Lambatos and the prosecutor made the DNA “match” sound especially compelling. The prosecutor asked about “the male DNA profile found in the semen.” Ms. Labatos made no effort to correct or clarify even though she firmly believed that Cellmark was reporting at least three different male profiles for the semen (and that Williams was, of course, a match to only one of them). Hammered with Ms. Lambatos’s figures for the Williams’ profile frequency, a judge surely would think that only Williams or a mythical twin could have been the rapist. In contrast, a judge who understood that Cellmark's tests also pointed to men with other DNA profiles might have been more willing to entertain some doubt.
The counterargument is that the probative value of the evidence for the ambiguous profile is essentially the same as the probative value of the evidence for the unambiguous profile that Ms. Lambatos was asked about. Assuming that the vaginal swab DNA is a mixture of the victim’s DNA and one man’s DNA, and assuming that the laboratory called all the alleles correctly, the likelihood ratio for the hypotheses of Williams versus that of a random, unrelated man is 1/[p(10,12,13) + p(13,13) + p(10,13) + p(12,13)], where p is the random-match probability for the full genotype, including the alleles shown in parentheses. Ms. Lambatos computed the probability p(12,13) as falling in the quadrillionths. Although I have not consulted allele frequency tables, it is a safe bet that similarly small probabilities would pertain to the profiles with the (13,13) and (10,13) genotypes. The random-match probability for the profile with the tri-allelic pattern would be even smaller. (When asked by the defense, Ms. Lambatos testified that a tri-allelic male was not a real possibility.) Therefore, I would predict that the correct computation would not differ from the number given to the judge by more than an order of magnitude. Hence, the witness’ failure to clarify or correct the prosecutor in her questioning affected the probative value of the evidence minimally.
Nevertheless, for the expert to present such oversimplified testimony without any qualification seems problematic to me. When confronted with the omissions on cross-examination, the expert owned up to them, but did she not ask the prosecutor to present the expert's reasoning accurately in the first place? And if she did, why did the prosecutor not do it?
References
1. G. Mertensemail, S. Rand, E. Jehaes et al., Observation of Tri-allelic Patterns in Autosomal STRs During Routine Casework, 2 Forensic Sci. Int’l: Genetics Supplement Series 38-40 (2009).
2. NIST STR-base, Tri-Allelic Patterns, June 2, 2011, http://www.cstl.nist.gov/strbase/var_D5S818.htm#Tri
3. American Society of Crime Laboratory Directors Laboratory Accreditation Board, ASCLD/LAB Guiding Principles of Professional Responsibility for Crime Laboratories and Forensic Scientists, Principle 19, Version 1.1, 2009.
On cross-examination, ISP analyst Karen Kooi Abbinanti, who examined the blood sample that Williams gave under court order in another case, testified to William’s STR profile. Because ISP analyst Sandra Lambatos, who provided the state’s only evidence of a DNA match, testified that “there [was] a computer match generated of the male DNA profile found in semen from the vaginal swabs of [LJ] to a male DNA profile that had been identified as having originated from Sandy Williams,” I presumed that the Cellmark report listed this profile as coming from the male fraction of DNA in the vaginal swab. Indeed, Lambatos testified that the “allele chart” in the Cellmark report “included data that [she] used to run [her] data bank search.” Joint Appendix at 61. Thus, I wrote that
The unnamed analyst believed that the semen had the following profile: D3 (16, 19), DWA (17, 17), FGA (18.2, 22), D8 (14, 14), D21 (29, 30), D18 (13, 17), D5 (12, 13), D13 (11, 11), D7 (10, 12), D16 (9, 11), TH01 (7, 7), TPOX (11, 11), and CSF (8, 10). The analyst’s report included this profile . . . .Now that the report is lodged, it is clear that this singular profile is not what the anonymous Cellmark analyst and Cellmark’s two laboratory directors, Robin Cotton and Jennifer Reynolds, signed off on. Their table, which was Lambatos’s “data,” has the entry of (10, 12, 13) instead of (12, 13) for the D5S818 locus. Had Ms. Lambatos used this tri-allelic genotype, Williams would have been excluded! (Tri-allelic, single locus profiles are rare, but they are not unheard of. For example, one paper reports three cases of tri-allelic patterns observed during routine forensic casework on 5964 Belgian residents [1], and the D5S818 (10, 12, 13) profile has been observed [2].)
Ms. Lambatos, however, testified on cross-examination that the Cellmark report’s “deduced male donor profile” (to quote the report itself) was not actually a deduced profile, but only a list of deduced alleles. Joint Appendix at 71. Interpreting it in this fashion (which may well be the correct understanding what the unknown analyst meant to write), she searched the unspecified database for certain two-allele subsets of the three alleles— namely, (13, 13), (10, 13), and (12, 13). Id. This made sense because, if Cellmark had correctly identified the victim’s profile — something that Lambatos did not check — then the rapist rather than the victim had to be the source of the 13-repeat allele.
The circumscribed nature of Ms. Lambatos’s testimony on direct examination about the “DNA match” is worthy of comment. Full disclosure would have required a scientist to reveal that other male profiles than just Williams’ profile were “consistent with” the vaginal-swab mixture and could have been picked out of a database in her trawl. Instead, Ms. Lambatos acquiesced in or suggested confining her testimony to Williams’s matching profile and the random-match probability associated with that one profile. In other words, she chose not to acknowledge possibilities that were inconsistent with the state’s theory. Does such selectivity contravene the professional responsibility of forensic scientists to “[a]ttempt to qualify their responses while testifying when asked a question with the requirement that a simple ‘yes’ or ‘no’ answer be given, if answering ‘yes’ or ‘no’ would be misleading to the judge or the jury”? [1]
The answer, I think, depends on how misleading Ms. Labatos’s answers on direct examination were. This was not a case of a single profile that probably could exclude everybody except for a twin brother. The analysts were unable to distinguish between Sandy Williams and other males with similar, but not identical profiles, as possible sources of the male DNA. By not disclosing this fact, Ms. Lambatos and the prosecutor made the DNA “match” sound especially compelling. The prosecutor asked about “the male DNA profile found in the semen.” Ms. Labatos made no effort to correct or clarify even though she firmly believed that Cellmark was reporting at least three different male profiles for the semen (and that Williams was, of course, a match to only one of them). Hammered with Ms. Lambatos’s figures for the Williams’ profile frequency, a judge surely would think that only Williams or a mythical twin could have been the rapist. In contrast, a judge who understood that Cellmark's tests also pointed to men with other DNA profiles might have been more willing to entertain some doubt.
The counterargument is that the probative value of the evidence for the ambiguous profile is essentially the same as the probative value of the evidence for the unambiguous profile that Ms. Lambatos was asked about. Assuming that the vaginal swab DNA is a mixture of the victim’s DNA and one man’s DNA, and assuming that the laboratory called all the alleles correctly, the likelihood ratio for the hypotheses of Williams versus that of a random, unrelated man is 1/[p(10,12,13) + p(13,13) + p(10,13) + p(12,13)], where p is the random-match probability for the full genotype, including the alleles shown in parentheses. Ms. Lambatos computed the probability p(12,13) as falling in the quadrillionths. Although I have not consulted allele frequency tables, it is a safe bet that similarly small probabilities would pertain to the profiles with the (13,13) and (10,13) genotypes. The random-match probability for the profile with the tri-allelic pattern would be even smaller. (When asked by the defense, Ms. Lambatos testified that a tri-allelic male was not a real possibility.) Therefore, I would predict that the correct computation would not differ from the number given to the judge by more than an order of magnitude. Hence, the witness’ failure to clarify or correct the prosecutor in her questioning affected the probative value of the evidence minimally.
Nevertheless, for the expert to present such oversimplified testimony without any qualification seems problematic to me. When confronted with the omissions on cross-examination, the expert owned up to them, but did she not ask the prosecutor to present the expert's reasoning accurately in the first place? And if she did, why did the prosecutor not do it?
References
1. G. Mertensemail, S. Rand, E. Jehaes et al., Observation of Tri-allelic Patterns in Autosomal STRs During Routine Casework, 2 Forensic Sci. Int’l: Genetics Supplement Series 38-40 (2009).
2. NIST STR-base, Tri-Allelic Patterns, June 2, 2011, http://www.cstl.nist.gov/strbase/var_D5S818.htm#Tri
3. American Society of Crime Laboratory Directors Laboratory Accreditation Board, ASCLD/LAB Guiding Principles of Professional Responsibility for Crime Laboratories and Forensic Scientists, Principle 19, Version 1.1, 2009.
Cross posted from The Double Helix Law Blog, 15 Dec. 2011
Subscribe to:
Posts (Atom)