What the Dog Knows

by Cat Warren

  I will never know for certain, when Solo false alerts, whether he distinguishes between lying and bullshitting. I believe Solo tries to be honest. Mike Baker once called him one of the most honest dogs he knows, partly because it is so easy to read his body language before an alert.

  Nonetheless, Solo will false alert. Infrequently, but it goes into my training and search records. Every alert gets counted. It mostly happens when he’s in scent, but not as close as he could and should get, but he decides that’s good enough for him. Or if I’m handling him badly, as I did in the warehouse. Sometimes I’ll never know why. Sometimes I can speculate why he alerted. Not all of them are false. If we’re searching a junkyard of wrecked cars, and Solo alerts on a front seat where an airbag has deployed or the windshield is shattered? I’ll reward him for that one, even if there’s no body in the trunk. Blood can stick around for years.

  If there are five or six cops standing and staring at something redolent lying on the ground, say, a bag with a dead dog in it? Solo will look at it, too, look around, gauge everyone’s expression, and think, Hey, maybe that’s something worth alerting on. They’re into it, right? In that case, no reward. Move on. During searches, if people want me to check garbage bags that look suspicious, or particular bones, I politely ask if they can keep a short distance away. On a recent search, I was having Solo check what I was sure was a pile of sand deposited by a recent creek flood, though one careful investigator thought it looked gravelike. I simply asked the searchers to stand away while I ran Solo through the area. They all obediently stepped well away and turned their backs, but they couldn’t help looking over their shoulders to see what he would do. Solo sniffed and moved on. Over the years, he has been increasingly proofed off dead animals. More important, he’s been increasingly proofed off the human gaze.

  It’s hardly dogs’ fault: We humans hardwired them through selective breeding to be utterly responsive to us. With working dogs, we take it a step further. We ask them to be both deeply bonded with their handlers and to act independently. They need to be both obedient and to think for themselves. We train them to ignore us and go get their work done. Push that door. Don’t look at me to do it. Open that gate. Find that body. Stop watching me and do your job. The game is to be together and separate. Bonded and independent. For some breeds, and for some dogs, it’s easier than for others.

  It’s partly why one of the hardest exercises in advanced obedience can be the “go out.” An obedience dog is used to being fed treats in exchange for gazing lovingly, unstintingly into the handler’s eyes on the heel and recall. Then the handler asks the dog to run enthusiastically straight away. If it’s not taught with the proper chain of treats and rewards, you can witness an otherwise fine obedience dog walk away slowly, sulkily, gazing back at his owner: You don’t love me anymore. You want me to go away.

  We humans are hardwired, too. We are attached to our dogs. The handler or even the helpful trainer can unconsciously play an unhelpful role in creating dependency, wanting the dog’s success. That’s why a handler training a sniffer dog should start training on blind problems, where the handler doesn’t know the location of the hide and can’t help the dog cheat. Then she should go on to double-blind problems, where the trainer flanking her doesn’t know the hide’s location, either. That’s why it was good when Nancy Hook started forgetting where she put the hides in the fields and woods early in Solo’s training. She was unconsciously providing all three of us with double-blind trainings. She was helping us avoid bullshit.

  • • •

  So what’s the harm with a cadaver handler here and there saying, with some bravado among friends at a seminar, that her dog never false alerts? Or a bloodhound handler bragging that his dog can follow a two-month-old track, or trail someone driving in a car for miles? Aren’t they the harmless equivalent of big fish tales?

  No. Bragging on your dog provides a tiny contribution to the general spreading of bullshit about working dogs. It creates a wishful blindness that doesn’t just end up hurting the training of a particular dog. It also helps create a filmy fiction about working dogs in general—a kind of milky, soft-focus portrait that helps us practice Hero Dog Worship.

  False claims, repeated often enough—as tracking trainer Tracy Bowling pointed out—reach the level of legitimacy. From there, one can trace the real and obvious harm those lies create. They undermine truthful handlers who don’t overreach. They keep people from training their dogs to the necessary level. They can make the work of a good dependable dog look obvious and simple when, in reality, it’s enormously difficult.

  The exaggerations send the media into a tailspin about the wonder of dogs, then a counter-tailspin when the inevitable cautionary tales emerge. The bed-bug backlash is a nice example. The New York Times’s honeymoon with bed-bug-detector dogs ended in less than a year. Its first March 2010 article had nary a doubt about the effectiveness of canine versus bug: “Bedbug-sniffing dogs, adorable yet stunningly accurate—entomology researchers at the University of Florida report that well-trained dogs can detect a single live bug or egg with 96 percent accuracy—are the new and furry front line in an escalating and confounding domestic war.” Dogs in those accounts seemed to work alone, without handlers. They just took a taxi to inspect Upper East Side hotels by themselves.

  Eight months later, the Times’s tone had changed: “Doubts Rise on Bedbug-Sniffing Dogs”: “But as the number of reported infestations rises and the demand for the dogs soars, complaints from people who say dogs have inaccurately detected bedbugs are also climbing.”

  Cautionary scientific studies are starting to appear, to the great consternation and sometimes rancorous objections of some dog handlers and organizations, who can pick holes in a study faster than an army of termites and claim that they are “shocked, shocked” to find there is anything amiss in the magical kingdom of working dogs. The feigned shock is another nice example of what Frankfurt calls bullshit. But it’s inevitable and, at one level, understandable. The canine legal arena has become incredibly complex and contentious—the Supreme Court took on its first two Florida dog-sniff cases in 2013. The cases were based on the Fourth Amendment’s protection against unreasonable searches. While the court sided with the dog and handler during a traffic stop in one case, the other case, which involved suspected marijuana growing in a private home, had a different outcome. The police used a drug dog’s sniff at the door of the house to establish probable cause to get a search warrant. The majority ruled that K9 noses aren’t that different from prying government eyes. Your right to privacy extends to keeping a K9 nose away from your home. That decision will have repercussions on how sniffer dogs’ noses can be used.

  So when a scientific study casts any doubt on the invincibility of working dogs, handlers and trainers react with alacrity. That was certainly the case for a 2011 study in Animal Cognition by researcher and former detection-dog handler Lisa Lit and two colleagues at the University of California, Davis. The study showed that law enforcement K9 handlers, when they expect to find gunpowder or marijuana in a certain place, either will state or will actually believe that their dogs have found the substance—even if there’s nothing there.

  It was a simple study at one level: no drugs or explosives planted anywhere. Instead, the researchers placed little pieces of red construction paper and told the handlers that those markers indicated either drugs or explosives. When handlers saw a red piece of paper falsely marking a scent location, they were much more likely to say that their dogs had signaled an alert. Conscious? Unconscious? Perhaps both. What was interesting was that the handlers were more distracted by those misleading pieces of red paper than their dogs were by the Slim Jims and tennis balls stashed in various corners. Lit and her colleagues recorded literally hundreds of false alerts.

  Lit’s study was not a dog study; it was a human-nature-with-dogs-added study, and it highlighted the need for a strengthened training regime. It also pointed to the problem of expectations. If we go in expecti
ng to find something, the chances are higher that we will. We all have confirmation bias. How much better it is when we can get our dogs to confirm that bias as well.

  After reading Lit’s study, I began incorporating more negative searches in my training with Solo. The first few times I did it—searching an entire abandoned aviation building without a single hide—Solo yowled in protest, trying to get at the tug toy in my pocket on the way out. He was mad. Here he was, surrounded by sheriff K9s and guys in uniforms who love to play tug. And no hides? My pocket got stained with resentful saliva on the way out, my thigh a bit bruised. He didn’t false alert. It was a start.

  The next time, the negative search would have to be blind: I wouldn’t know there weren’t hides out. Then double blind, where the person with me wouldn’t know if there were hides or not. At some point, I might graduate to the equivalent of pieces of red construction paper and not react in a knee-jerk fashion. One step at a time. I did call home as we left the training area. David put a cadaver hide out in the yard, so as Solo and I walked from the car to the house, Solo’s head flipped. He ran toward the scent. Look! Cadaver after all! Give me my toy now. He was pleased. I was pleased that I hadn’t needed to give a specific command for him to find the hide. Mike Baker had told me at the beginning of training that Solo should be prepared at any time to define the game without waiting for my specific command.

  The study by Lisa Lit and her colleagues is not the only detection study that has shown less-than-stellar results and pointed to the need for strengthened training regimes. Larry Myers of Auburn University did an extensive, not-yet-published study of twelve dog-and-handler teams who work full time at detection.

  “It was a simple test,” Myers said in his straightforward way. “I was afraid it was going to be too simple.” He randomly placed scent samples in brand-new pizza boxes. He stayed out of the room where the boxes were placed so he wouldn’t unconsciously cue the handlers. The reliability of the dog-and-handler teams ranged from one team’s dismal score of 30 percent reliability to another handler who had 97 percent reliability. That highest score belonged to an experienced trainer and handler who, Myers noted, works double-blind problems constantly.

  A big clump of teams came in between 60 and 85 percent reliability. Eighty-five percent is quite respectable. Sixty percent, not so much. That starts to get closer to chance. Thirty percent reliability should make you think about changing the dog or the handler or the entire training regimen.

  “It was interesting to see how bloody awfully a lot of people did who thought they were hot shit,” Myers commented. “I have given up being amazed at how people can think they’re honestly doing something right, and how self-deluding they can be.”

  • • •

  He has testified under oath, for example, that even though he does not keep detailed records of his activities he knows that his dogs have almost never been wrong. According to [Keith] Pikett, as of 2009 his dog “Clue” had been wrong once out of 1,659 lineups. “James Bond” had been wrong once out of 2,266 times. “Quincy” had only been proven wrong three times in 2,831 lineups.

  —Innocence Project of Texas report, 2009

  Although it’s rare, extreme canine versions of Clever Hans have appeared in America’s courtrooms, with verdicts of guilt or innocence at stake. Math tricks played for the amusement of crowds can morph into dog tricks played in front of gullible juries, sending innocent people to prison. When handlers lie about or exaggerate their dogs’ capabilities under oath, it poisons the well for handler testimony and the credibility of the dog’s nose.

  This is the kind of testimony that exercises Roger Titus, vice president of the National Police Bloodhound Association. Over the past decades, he has worked trails with his many bloodhounds that made him proud. That work has helped put guilty people in prison. When his dogs are able to follow three- and four-day-old trails, he’s incredibly pleased. What undermines the work are the lies he hears in training and on the witness stand. The stories can become albatrosses around the necks of conscientious trainers and handlers. “On occasion, it has become outrageous,” Roger said of handlers’ claims. “Four months old? Impossible. People who put trails out in January to run in May are full of it.”

  The danger signals are clear, Roger said. “It’s the handler who wants to be a legend in his own mind.” Yet such legends end up as sworn evidence in the courtroom and cautionary tales in law journals. Scent evidence, or a dog’s sniff, should be one piece of many pieces of evidence in a case, but sometimes it’s the major evidence. That’s a problem.

  Now-deceased Pennsylvania State Trooper John Preston was one such legend. His fraudulent claims of his dog’s ability to track scent led to as many as sixty people being convicted solely or partially on his false testimony, according to the Innocence Project. Preston claimed what tracking-dog experts say is impossible—that his dog could smell human traces months or even years after a suspect walked over the ground or on heavily trafficked streets. One man was freed in 2009 after spending twenty-six years in prison. Florida prosecutors hadn’t bothered reviewing Preston’s cases after he was exposed as a fraud in the mid-1980s. In 2008, Florida State Attorney Norman Wolfinger ordered a review of murder and sexual battery cases where Preston testified, although the local newspaper editorialized that an independent investigation was needed. That didn’t happen. Preston died in 2008.

  Keith Pikett, a now-retired Fort Bend County, Texas, sheriff’s deputy, is a more recent, still-living legend. His claims about his bloodhounds’ scenting abilities resulted in what the Texas Innocence Project told the New York Times amounts to fifteen to twenty people in prison “based on virtually nothing but Pikett’s testimony.”

  Pikett had been involved in helping indict more than 1,000 suspects nationwide. His specialty was the scent lineup. A scent lineup starts with collecting scent from a crime scene, then collecting scent from a suspect. The dog’s job is to “match” the scent from the crime scene with the scent of the suspect. For scent matching to be valid, it needs to be done under pristine circumstances, double-blind, with careful preservation methods. In the Netherlands—where the courts accept scent lineup but only as corroborating evidence—they use more than one dog, and the work is done in a sterile room without handlers present. In other words, no cross-contamination and no possibility of Clever Hans. That’s not the way Keith Pikett did it in Texas.

  Ultimately, the police evidence videos showing Pikett and his dogs running scent lineups “cooked him,” Roger said succinctly. I watched them online. Paint cans with numbers were placed on the grass in a line. An investigator pulled gauze pads in plastic bags out of one can and put them in another with bare hands. If there ever were an uncontaminated scent object with the suspect’s scent on it, that scent was now possibly in several cans. Pikett then ran his bloodhounds on leashes down the line of paint cans. The dogs would look up, bay, stop when Pikett stopped. They would shake their heads, slobber flying, and bay again. They avoided some of the cans. Pikett stopped one dog at a can with the leash, and the dog stood there. Another dog paused squarely between two cans, and Pikett said the dog had alerted on one of them. One dog bayed and ran past two cans, and Pikett said the dog had alerted on one of them. Head shakes, barks, and pauses were all alerts, according to Pikett. The bloodhounds were doing all three of those things.

  “This is the most primitive evidential police procedure I have ever witnessed,” Robert Coote, the former head of a British K9 police unit, testified after he watched the videos. “If it was not for the fact that it is a serious matter, I could have been watching a comedy.”

  The problem is that cops and prosecutors and juries across Texas bought the comedy for years. One man accused of killing three people, based greatly on Pikett’s dog evidence, was partially blind, handicapped with diabetes and bone spurs, and physically incapable of committing the murders Pikett linked him to. He spent seven months in prison before someone else confessed to the killings.

  Michael B
uchanek, a retired sheriff’s department captain, was identified by Pikett’s dogs as the prime suspect in the rape and murder of his next-door neighbor, a social worker, based on a police theory that Buchanek had put the body in his car trunk, driven five miles, and dumped the body in a field. Pikett’s dogs supposedly followed the victim’s scent in a moving car for five miles, twenty-four hours after the crime occurred. As international working-dog experts Resi Gerritsen and Ruud Haak noted with heavy irony in their book, K9 Fraud, it was “an exceptional performance that no dog can copy.”

  The police, Buchanek told the New York Times, “just kept telling me, ‘the dogs don’t lie—we know you did it.’ ” After months of living under a cloud of suspicion, Buchanek was cleared when DNA implicated another man who later confessed to the crime.

  Juries are especially vulnerable to dog testimony, Roger Titus said. “You see them look at each other,” he said. “Out of ten people, you’ve got eight that like dogs. A receptive audience.”

  Coote and Roger were not the only ones horrified. Roger’s colleague Doug Lowry, the president of the National Police Bloodhound Association, testified against Pikett, saying he was doing “a disservice to police bloodhound teams throughout the country.” It’s rare for organizations or top handlers or trainers to testify against other handlers. But these men believed that Pikett and his practices needed to be stopped. “Pikett has done a lot of damage to the veracity of dogs in the Texas system,” Andy Rebmann said.

  While Pikett is retired and no longer testifying, his cases still pop up in the news. In 2007, Megan Winfrey of East Texas was sentenced to life in prison for a murder she was charged with committing at the age of sixteen. The major evidence against her? Keith Pikett’s scent lineups. On appeal, her father was exonerated for the same murder. Her brother was tried for the murder as well, but his attorney argued strenuously against the scientific validity of Pikett’s scent lineups; her brother’s jury deliberated thirteen minutes before finding him not guilty. Megan Winfrey appealed to the Texas Court of Criminal Appeals in April 2012. On February 27, 2013, she was acquitted of all the charges against her, but the prosecutor in her case requested a rehearing. She was finally released on April 19.