Thursday, April 22, 2010

Poorly Conducted Trailing and Scent Identification Procedures Lead to Wrongful Accusation

In a case arising in Texas in 2006, police, largely on the basis of evidence provided by a dog handler named Keith Pikett, concluded that Michael Buchanek had murdered Sally Blackwell. Several months after the investigation began another individual, Jeffrey Grimsinger, confessed to having killed Blackwell. Had it not been for this confession, the canine evidence provided by Pikett might have put Buchanek, who did a tour training Iraqi police, in prison. Grimsinger was suggested as a suspect early in the investigation but this seems to have been ignored by the officer in charge of the case, Sam Eyre.

Blackwell was abducted from her home on March 13. Her body was found two days later on March 15, 2006. Before the body was found, dogs and handlers from the Texas Department of Criminal Justice (TDCJ) had been brought to help search the area where Blackwell’s cell phone and purse were found. The dogs picked up no trail at the location and the handlers believed that the phone and purse had been thrown from a car. (The Texas Department of Criminal Justice has 47 kennels throughout its system, with 101 kennel staff (as of 2006). About half the kennels hold tracking dogs used primarily to track escaped convicts, but also to help law enforcement. There are also scent-specific canines that include drug detection dogs in the TDCJ system.)

The Police Department of Victoria, Texas, employed Keith Pikett, a handler of trailing and scent identification dogs, to help in the investigation. An affidavit sworn to by Eyre stated that Pikett’s dogs trailed from the place where Blackwell’s body was found to Buchanek’s home. This was false, as other evidence established that the dogs trailed to the home of the victim, and only after reaching Blackwell’s house was Pikett told that a “person of interest” lived at a different location. The dogs were then taken to a street near that location and re-scented, after which they went to Buchanek’s home. The dogs would have passed this house when they were trailing from Blackwell’s body to her home, but if they did so they did not alert at the time.

There was another problem with the tracking in that the investigators believed that Blackwell’s body had been taken to the place where it was found in a vehicle. Thus, trailing from the location where the body was found to either house would have involved dog’s following a body in a car. Pikett argued that this was in fact possible, but the TDCJ handlers doubted it could be done. Some of those handlers, according to Buchanek’s complaint, expressed the belief that Pikett was “full of b.s.” The federal district court observed that the route followed by the dogs from the location of the body to the victim’s home would have required trailing the car transporting the body over five miles.

Pikett cited a piece in the FBI’s Forensic Science Communications noting that ventilation systems in cars move fresh air through the interior of a vehicle and into the external atmosphere, conceivably laying a scent trail a dog could follow. (Stockham, R.A., Slavin, D.L., and Kift, W. Specialized Use of Human Scent in Criminal Investigations. Forensic Science Communications, July 2004, vol. 6(3)). Research might support this possibility for short distances, but the five miles involved in this case and several major highways would truly be exceptional.

When Pikett and his dogs arrived in the vicinity of Buchanek’s home (after the dogs supposedly trailed to Blackwell’s home), an officer asked a man walking his dogs to remove them from the street because police dogs were coming. The court said this may have indicated that Pikett expected his dogs to trail to Buchanek’s house. The court also noted that the officer was preceding the dogs and it may have been his scent they were following. Even worse, it may have meant that Pikett was leading his dogs, rather than the other way around.

Pikett also conducted a scent lineup. The scent for Buchanek used in the lineup appears to have been a legal document Buchanek may not have touched for two years and may have been more recently handled by law enforcement officers. The other scents in the lineup may have come from “filler scents” that Pikett kept for such purposes, though this was not definitely established. (Another case involving Pikett’s services describes Pikett testifying that he has “a large selection of scent samples” for scent lineups, and that he has separated his samples by race and gender. Winfrey v. State, 2009 WL 1636849 (2009).) If filler samples were used here, the scents were not prepared in the same way or at the same time, which would be a serious defect in a scent lineup. There was also no evidence as to whether the dogs were on leads or not, or as to how Pikett was kept blind as to the placement of the suspect’s scent (if he was indeed blind). The lineup was performed in a police department parking lot. Officers who re-arranged the bags containing the scents in a row of cans may have touched the bags in this process, creating the possibility, if not the likelihood, of contamination. There were probably other problems with the scent lineup but there is no detailed description of how it was conducted.

In the two decisions of the federal district court, motions for summary judgment have been denied, clearing the way for a trial. Buchanek v. City of Victoria, 2009 WL 500564, 2010 WL 1268069 (S.D. Tex 2010). The case has received national attention when the first decision was released. John Schwartz, “Picked From a Lineup, on a Whiff of Evidence,” New York Times, November 11, 2009, p. 1.

Scent lineups have come to be called “junk science” by those who believe that wrongful convictions as a result of these lineups are all too common. This case may be unique in that a single handler may have turned trailing as well as scent identification into “junk science” in a single investigation.

Addendum. Winfrey has been reversed by the Texas Court of Criminal Appeals. This court noted that no eyewitnesses put the defendant at the crime scene, the state was unable to match the defendant to a fingerprint and footprints found at the crime scene, and the defendant did not match the DNA profile obtained from the crime scene. The evidence convicting Winfrey consisted of his belief that he was the number one suspect when he wasn't and the fact he told a cellmate things he had heard about the crime but said nothing about his own involvement to the cellmate. The court found this "legally insufficient to support a conviction of murder beyond a reasonable doubt." The court cited other state courts that had held that dog-scent evidence was insufficient, standing alone, to support a conviction, and cited Taslitz (42 Hastings Law Journal 15 (1990)) in holding that scent-discrimination lineups were to be regarded as "separate and distinct from dog-scent tracking evidence." The court said that the scent lineup evidence could raise a "strong suspicion" of the defendant's guilt but could not convict him. A judgment of acquittal was entered by the appellate court. Winfrey v. Texas, 2010 WL 3656064 (Tex. Crim. App. 2010). Thanks to Gail Kikawa McConnell of the Fort Bend County, Texas, District Attorney's office for emailing me about the reversal.

Tuesday, April 20, 2010

Scent Lineup Admitted into Evidence by California Court Despite Numerous Deficiencies

A recent California case demonstrates why scent lineups continue to be attacked as “junk science” by the defense bar, and why better procedures must be imposed by courts and law enforcement agencies to overcome such criticisms. People v. White, 2009 WL 3111677 (Cal.App. 2Dist. 2009)

The crime occurred on June 7, 2006, in the early evening in Compton, California. Five teenagers were skateboarding when two of them were shot by an assailant. One died. Shell casings were recovered from the crime scene and placed in manila envelopes. Problem one: manila envelopes, as noted by an expert witness for the defense, Lawrence Myers (Auburn University College of Veterinary Medicine), are porous, which allows for the possibility of contamination. Scent pads were later wrapped around the casings for about 10 minutes and the pads were stored in Ziploc bags. Problem two (minor): glass containers would have been better. (See Hudson, D.T., Curran, A.M., and Furton, K.G., The Stability of Collected Human Scent under Various Environmental Conditions. Journal of Forensic Sciences, 2009, vol. 54(6): 1270-1277, noting glass containers were preferable to plastic.) The Ziploc bags were put in an evidence locker for two days. Problem three: The scent samples should have been refrigerated, as noted by the defense expert.

Two officers collected scent from four individuals, including the arms of the defendant, put the scents in boxes and arranged the boxes in a diamond format. Problem four (perhaps minor): If by collecting scent from the arm of the defendant, the court means that it was collected by some means such as rubbing a scent pad along the arm, this would mean that scent was collected from a slightly different body part that the fingers that probably handled the casings collected at the crime scene (while the weapon was being loaded). Some evidence has suggested that scent from different parts of the body may be difficult for a dog to distinguish. (See Brisbin, I.L. and Austad, S.N., Testing the Individual Odour Theory of Canine Olfaction. Animal Behaviour, 1991, vol. 42, 63-69.) Another study contradicted this result, however. (See Schoon, G.A.A., and de Bruin, J.C. The Ability of Dogs to Recognize and Cross-Match Human Odours. Forensic Science International, 1994, vol. 69: 111-118.)

The dog was taken into the middle of the diamond and scented to a pad that had been wrapped around a shell casing. The dog lay down next to the box containing the defendant’s scent. Problem five: There was no discussion of any control trials. European protocols generally call for several control trials to determine if the dog is willing to work on the day of the tests and to verify that the dog is alerting correctly where the “perpetrator” and the target are known to be the same. Also, it may be necessary to determine if the dog has an “attraction” to the scent of a suspect. Problem six: Almost all protocols developed in the Netherlands, Poland, and other law enforcement centers in recent years provide a dog with at least five, and often six or seven choices, sometimes in two lineups simultaneously (See Schoon G.A.A. Scent identification lineups by dogs (Canis familiaris): experimental design and forensic application. Applied Animal Behaviour Science, 1999, vol. 49, 257-267; Schoon G.A.A. A first assessment of the reliability of an improved scent identification line-up. Journal of Forensic Sciences, 1998, vol. 43, 1, 70-75.) In some control lineups, it is often advisable to include “zero trials” where the correct response of the dog is not to alert at all because the item the dog is scented to does not correlate with any of the scents in the lineup.

The handler was unaware of which box contained the scent of the defendant, but while the lineup was being conducted, the two officers who set it up watched from a picnic table 40 to 50 feet away from the diamond formation. Problem seven: No responsible lineup procedure allows the handler to see an experimenter or anyone who knows the location of the target or the suspect’s scent. The “clever Hans” effect can arise even if the handler is blind to the location of the target. This issue was raised by the defense expert (Myers) but dismissed

The defense expert could not interpret the handler’s training logs and could therefore not assess the dog’s reliability. Problem eight: It is the author’s opinion that training logs should always be available to counsel for the defense (or the prosecution in the case of an exonerating scent lineup). The court found this problem was irrelevant because the handler testified as to the dog’s training and was subject to cross-examination. The handler testified that he was not aware that his dog had made any mistakes in 171 criminal lineups. Courts should cease allowing handlers to be the sole means by which major foundational requirements are satisfied in scent lineups. (See the records procedures recommended by the Scientific Working Group on Dog and Orthological Detector Guidelines (SWGDOG), Human Scent Dogs: Scent Identification Lineups (posted at www.swgdog.org). SWGDOG is a forum, with participants from the FBI, other law enforcement agencies, and research facilities involved in canine forensic research.)

The court concluded that the defense expert had done no more than make suggestions for improving the lineup procedure that was used, but that his testimony had not shown the evidence to be unreliable. There was, admittedly, good corroborating evidence here, including the testimony of two eyewitnesses. Part of the case concerned the reliability of visual lineups for eyewitnesses, another area that is not as simple as it once seemed to be. (See, e.g., Wright, D.B. and McDaid, A.T. Comparing System and Estimator Variables Using Data from Real Line-ups. Applied Cognitive Psychology, 1996, vol. 10(1), 75-84.)

Twenty years ago, Professor Andrew Taslitz argued that scent lineups should not be admitted as evidence of guilt in a criminal trial. He noted that research being conducted by Jan de Bruin of the Dutch police might reach a level to be satisfactory for judicial use but noted the research had not yet been published. Adee Schoon, cited in several parentheses above, continued that work. Some convictions obtained in part by scent lineup results have been overturned, and it is not clear if the public still has a “mythic belief” in the scenting powers of dogs. (Taslitz, A. Does the Cold Nose Know? The Unscientific Myth of the Dog Scent Lineup. Hastings Law Journal,1990, vol 42, 15-134.)

The problem with cases like People v. White is that they risk opening use of a poorly conducted scent lineup to the charge of “junk science.” Almost all of the defects of this lineup would be obvious to the researchers in this area, and most of them could have been easily cured. Some might require that additional dogs be available to confirm the result. Protocols developed in Polish research have demonstrated that requiring at least two dogs—ideally three dogs—to identify the suspect in a scent lineup substantially diminishes the possibility of a misidentification. (Jezierski, T., personal communication, 2010) A proper testing environment and equipment might require additional expense. Nevertheless, in cases where witnesses are fearful or disappear, the scent lineup, if properly conducted at the levels required by the more advanced forensic research centers, may be an appropriate link in the case.

Saturday, April 17, 2010

Dogs Distinguish Types of Warning Growls


Animals have been shown to make sounds that are context specific, such as alerting other members of the group to a predator, even a specific type of predator. Dogs, like wolves, bark, but dogs bark more frequently and in more variable contexts than is the case with wolves. Both dogs and wolves growl in three situations that were studied in a recent paper: (1) offensive threatening in social conflict, (2) guarding food, and (3) during social play. Calls are described as functionally referential if three criteria are satisfied:

1. The signals displayed in different contexts have different acoustic structures.
2. Prerecorded signals evoke different behavioral responses in a playback experiment.
3. The manipulation of the referent alters the signal production.

A group of Hungarian and Austrian scientists recorded growls during three situations: (1) a dog guarding his food from a strange dog (a food-guarding growl, FG); (2) a strange human approaches a dog threateningly (threatening growl, TS); and (3) the dog plays tug-of-war with his owner (play growl, PL). The play growls were less than half as long as the two other types of growls. In a playback experiment, the researchers looked to see whether a food-guarding growl would cause another dog to retreat despite the physical absence of the dog making the sound. The threatening growl was also used in the same context to see if it would have an equally strong deterrent effect. The expressions of the dogs were different in the contexts of the growls. Thus, dogs growled at a threatening human with a closed mouth, but while defending food against another dog, they showed their teeth and pulled back their lips.

In the experiment, an owner and dog entered a room with a bone, which the dog was allowed to sniff but not touch. After a knocking signal from the experimenter, the owner released the dog. The owner was told not to talk, touch, or look at his dog. If the dog got within 5 cm of the bone, a pre-recorded growl was played. If the dog withdrew, the growl recording stopped paying. If the dog approached the bone again, the growl was played again.

The results were dramatic. After hearing the growl for the first time, 11 of 12 dogs in the FG group withdrew from the bone within 15 seconds, whereas only 2 of 12 dogs in the TS group and 4 of 12 dogs in the PL group withdrew in that period. Seven dogs in the FG group did not approach the bone again within the next 90 seconds, while only one dog in the TS group and one in the PL group stayed away. It was clear that FG growls had a stronger deterrent effect than TS growls though both were in the “agonistic” category of sounds. The researchers also found that the growls of bigger dogs had no measurably greater deterrent effect than the growls made by small dogs, confirming at least in part the common belief that dogs don’t think much about size. Farago, T., Pongracz, P., Range, F., Viranyi, Z, and Miklosi, A., ‘The Bone is Mine’: Affective and Referential Aspects of Dog Growls, Animal Behaviour vol. 79, 917-925 (2010).

Aerodynamics of the Wake of Walking Person

Trailing dogs are generally thought to follow a path based on their smell of proteins and/or bacteria in desiccated epithelial cells that flow off us at the rate of thousands per minute. This raises the question of how the air flow around a person affects the pattern by which these cells swirl and fall in our wake. Three engineering professors at Pennsylvania State University studied the aerodynamics of the human wake in an article published in the Journal of Fluids Engineering. The authors noted that prior studies of the aerodynamics of moving cylinders had not always taken into account the gap between the legs of a person. The authors state that “the leg separation has a significant effect on the structure of the wake for flows in the primary walking direction.” Swinging the arms, on the other hand, “has little significance in the formation of the wake.” The authors note that the results share many bulk flow structures with smoke visualization experiments. The figure shows the “significant physics” observed in flow visualization experiments. The results indicate two distinct wake regions behind a walking person, one behind the torso and one behind the legs. A significant downwash occurs behind the body which spreads the lower portions of the wake. Edge, B.A., Paterson, E.G., and Settles, G.S., Computational Study of the Wake and Contaminant Transport of a Walking Human, J. Fluids Engineering, 127, 967-977 (September 2005).

Wednesday, April 14, 2010

Middle Eastern Origin of Dogs Argued in New Genome Study

A study of the dog genome published in Nature, the pre-eminent publication in biology, reached some surprising and immediately controversial results. For one, disagreeing with prior studies, the new research concludes that “dog breeds share a higher proportion of multi-locus haplotypes unique to grey wolves from the Middle East….” (vonHoldt et al., Genome-wide SNP and Haplotype Analyses Reveal a Rich History Underying Dog Domestication, Nature, vol. 464(8), April 2010). This takes issue with mitochondrial DNA studies that suggested that canine domestication first occurred in east Asia (see Savolainen et al., Science vol. 298, 1610-1613, 2002).

The new research does accept that interbreeding with local wolf populations “clearly occurred elsewhere in the early history of specific lineages.” The researchers sampled 912 dogs from 85 breeds, as well as taking samples from 225 grey wolves from 11 populations around the world. Certain breeds were found to have demonstrated admixture with wolves, particularly breeds with ancient origins that are highly divergent from other dog breeds. Breeds in this group include, not surprisingly, Akitas, Alaskan malamutes, Siberian huskies, and American Eskimo dogs. However, wolf admixture is also found in basenjis, Afghan hounds, Samoyeds, salukis, Canaan dogs, New Guinea singing dogs, dingoes, Chow chows, and Chinese Shar Peis. Some breeds share more with Chinese wolves, some with European wolves, and some with Middle Eastern wolves. Akitas, for instance, show admixture with Chinese wolves, Staffordshire bull terriers with Northern wolves, and basenjis with Middle Eastern wolves. In some cases, the admixture must have been ancient. Dingoes have been isolated from any wolf population for perhaps 4,000 years.

Towards the end of the paper, the researches say that “some ancient east Asian breeds show affinity with Chinese wolves, which suggests that they were derived from Chinese wolves or admixed with them after domestication." This challenges the assumption that has prevailed for some years that dogs were domesticated only once, and harks back to Darwin and others who suggested that canine domestication may have occurred more than once from different populations of canids. The researchers also note that wolves have changed as a result of dogs. The mutation responsible for black coat color was transferred from dogs to grey wolves.

The researchers found “distinct genetic clusters” even within modern dogs based on phenotype and function, including spaniels, scent hounds, mastiff-like breeds, small terriers, retrievers, herding dogs, and sight hounds. Scent hounds include Basset hounds, beagles, and dachshunds. The research behind this six-page paper (supplementary information of 38 pages) is probably only the beginning of a series of papers one can expect from these 36 authors. That’s right! 36 authors.

Supplemental note. I received a comment arguing that American Eskimo dogs do not belong in the category of ancient breeds that show admixture with wolves. The supplemental information to the paper in Nature lists three somewhat distinct groups of dogs in this category. The first consists of basenjis, Afghan hounds, Samoyeds, Salukis, and Canaan dogs. The second consists of New Guinea singing dogs, dingoes, Chow chows, Shar peis, and Akitas. The third consists of Alaskan malamutes, Siberian huskies and American Eskimo dogs. Looking at Supplemental Figure 1, the research indicates that the first group demonstrates significant gene sharing with Middle Eastern wolves, the second group with Chinese wolves, and the third group with Northern wolves. The Inuit Sled Dog (sometimes confused with the American Eskimo dog) is generally thought of as an ancient breed, but the American Eskimo Dog is described by the American Kennel Club as a member of the Spitz family related to the spitz breeds and the white Pomeranian. See Sue Hamilton, "Replica or the Real Deal," The Fan Hitch 4(3) (May 2002). The Nature article does, however, list "Ancient and spitz breeds" as one category so I have to assume, until I hear from one of the Nature article's authors, that they do in fact mean the American Eskimo dog.